DataAnalysisWithHistogram.py

from datasets import load_dataset
import numpy as np
import json
import matplotlib.pyplot as plt

data = load_dataset("rikeshsilwalekg/imeaudio", revision = "v3-output-audio-features", split="train")

duration_values = [example['audio_features']['duration'] for example in data]



###########----------------__FOR DURATION__--------------------------------------########################



# Define bin width and bins for 0.5 seconds interval between 0 and 5 seconds
bin_width = 0.5
min_duration = 25
max_duration = 30
bins = np.arange(min_duration, max_duration + bin_width, bin_width)

# Filter data to only include durations between 0 and 5 seconds
filtered_durations = duration_values[(duration_values >= min_duration) & (duration_values <= max_duration)]

# Compute histogram data
counts, bin_edges = np.histogram(filtered_durations, bins=bins)

# Prepare dictionary with string keys for each bin range
hist_dict = {}
for i in range(len(counts)):
    key = f"{bin_edges[i]}-{bin_edges[i+1]}"
    hist_dict[key] = int(counts[i])

# Save histogram data to JSON file

json_filename = "audio_duration_histtogram_25_30_seconds.json"
with open(json_filename, "w") as f:
    json.dump(hist_dict, f, indent=2)

# Plot histogram
plt.hist(filtered_durations, bins=bins, color='purple', edgecolor='black')
plt.title('Distribution of Audio Durations (25-30 seconds)')
plt.xlabel('Duration in seconds')
plt.ylabel('Count')
plt.xlim(min_duration, max_duration)
plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/audio_duration/AudioDurations0-5.jpg")
plt.show()


import numpy as np
import json
import matplotlib.pyplot as plt


# Define bin width and bins for 0.5 seconds interval between 0 and 5 seconds
bin_width = 0.5
min_duration = 0
max_duration = 5
bins = np.arange(min_duration, max_duration + bin_width, bin_width)

# Filter data to only include durations between 0 and 5 seconds
filtered_durations = duration_values[(duration_values >= min_duration) & (duration_values <= max_duration)]

# Compute histogram data
counts, bin_edges = np.histogram(filtered_durations, bins=bins)

# Prepare dictionary with string keys for each bin range
hist_dict = {}
for i in range(len(counts)):
    key = f"{bin_edges[i]}-{bin_edges[i+1]}"
    hist_dict[key] = int(counts[i])

# Save histogram data to JSON file
json_filename = "audio_duration_histtogram_0_51_seconds.json"
with open(json_filename, "w") as f:
    json.dump(hist_dict, f, indent=2)

# Plot histogram
plt.hist(filtered_durations, bins=bins, color='purple', edgecolor='black')
plt.title('Distribution of Audio Durations (0-5 seconds)')
plt.xlabel('Duration in seconds')
plt.ylabel('Count')
plt.xlim(min_duration, max_duration)
plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/audio_duration/AudioDurations0-5.jpg")
plt.show()


import numpy as np
import json
import matplotlib.pyplot as plt


# Define bin width and bins for 0.5 seconds interval between 5 and 10 seconds
bin_width = 0.5
min_duration = 5
max_duration = 10
bins = np.arange(min_duration, max_duration + bin_width, bin_width)

# Ensure duration_values is a NumPy array
duration_values = np.array(duration_values)

# Filter data to only include durations between 5 and 10 seconds
filtered_durations = duration_values[(duration_values >= min_duration) & (duration_values <= max_duration)]

# Compute histogram data
counts, bin_edges = np.histogram(filtered_durations, bins=bins)

# Prepare dictionary with string keys for each bin range
hist_dict = {}
for i in range(len(counts)):
    key = f"{bin_edges[i]}-{bin_edges[i+1]}"
    hist_dict[key] = int(counts[i])

# Save histogram data to JSON file
json_filename = "audio_duration_histogram_5_10_seconds.json"
with open(json_filename, "w") as f:
    json.dump(hist_dict, f, indent=2)

# Plot histogram
plt.hist(filtered_durations, bins=bins, color='purple', edgecolor='black')
plt.title('Distribution of Audio Durations (5-10 seconds)')
plt.xlabel('Duration in seconds')
plt.ylabel('Count')
plt.xlim(min_duration, max_duration)
plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/audio_duration/AudioDurations5-10.jpg")
plt.show()


import numpy as np
import json
import matplotlib.pyplot as plt


# Define bin width and bins for 0.5 seconds interval between 5 and 10 seconds
bin_width = 0.5
min_duration = 10
max_duration = 15
bins = np.arange(min_duration, max_duration + bin_width, bin_width)

# Ensure duration_values is a NumPy array
duration_values = np.array(duration_values)

# Filter data to only include durations between 5 and 10 seconds
filtered_durations = duration_values[(duration_values >= min_duration) & (duration_values <= max_duration)]

# Compute histogram data
counts, bin_edges = np.histogram(filtered_durations, bins=bins)

# Prepare dictionary with string keys for each bin range
hist_dict = {}
for i in range(len(counts)):
    key = f"{bin_edges[i]}-{bin_edges[i+1]}"
    hist_dict[key] = int(counts[i])

# Save histogram data to JSON file
json_filename = "audio_duration_histogram_10_15_seconds.json"
with open(json_filename, "w") as f:
    json.dump(hist_dict, f, indent=2)

# Plot histogram
plt.hist(filtered_durations, bins=bins, color='purple', edgecolor='black')
plt.title('Distribution of Audio Durations (10-15seconds)')
plt.xlabel('Duration in seconds')
plt.ylabel('Count')
plt.xlim(min_duration, max_duration)
plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/audio_duration/AudioDurations10-15.jpg")
plt.show()


import numpy as np
import json
import matplotlib.pyplot as plt

# Example duration_values data (replace with your actual data)
# duration_values = ...

# Define bin width and bins for 0.5 seconds interval between 5 and 10 seconds
bin_width = 0.5
min_duration = 15
max_duration = 20
bins = np.arange(min_duration, max_duration + bin_width, bin_width)

# Ensure duration_values is a NumPy array
duration_values = np.array(duration_values)

# Filter data to only include durations between 10 and 15 seconds
filtered_durations = duration_values[(duration_values >= min_duration) & (duration_values <= max_duration)]

# Compute histogram data
counts, bin_edges = np.histogram(filtered_durations, bins=bins)

# Prepare dictionary with string keys for each bin range
hist_dict = {}
for i in range(len(counts)):
    key = f"{bin_edges[i]}-{bin_edges[i+1]}"
    hist_dict[key] = int(counts[i])

# Save histogram data to JSON file
json_filename = "audio_duration_histogram_15_20_seconds.json"
with open(json_filename, "w") as f:
    json.dump(hist_dict, f, indent=2)

# Plot histogram
plt.hist(filtered_durations, bins=bins, color='purple', edgecolor='black')
plt.title('Distribution of Audio Durations (15-20 seconds)')
plt.xlabel('Duration in seconds')
plt.ylabel('Count')
plt.xlim(min_duration, max_duration)
plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/audio_duration/AudioDurations15-20.jpg")
plt.show()


import numpy as np
import json
import matplotlib.pyplot as plt

# Example duration_values data (replace with your actual data)
# duration_values = ...

# Define bin width and bins for 0.5 seconds interval between 5 and 10 seconds
bin_width = 0.5
min_duration = 20
max_duration = 25
bins = np.arange(min_duration, max_duration + bin_width, bin_width)

# Ensure duration_values is a NumPy array
duration_values = np.array(duration_values)

# Filter data to only include durations between 10 and 15 seconds
filtered_durations = duration_values[(duration_values >= min_duration) & (duration_values <= max_duration)]

# Compute histogram data
counts, bin_edges = np.histogram(filtered_durations, bins=bins)

# Prepare dictionary with string keys for each bin range
hist_dict = {}
for i in range(len(counts)):
    key = f"{bin_edges[i]}-{bin_edges[i+1]}"
    hist_dict[key] = int(counts[i])

# Save histogram data to JSON file
json_filename = "audio_duration_histogram_20_25_seconds.json"
with open(json_filename, "w") as f:
    json.dump(hist_dict, f, indent=2)

# Plot histogram
plt.hist(filtered_durations, bins=bins, color='purple', edgecolor='black')
plt.title('Distribution of Audio Durations (20-25 seconds)')
plt.xlabel('Duration in seconds')
plt.ylabel('Count')
plt.xlim(min_duration, max_duration)
plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/audio_duration/AudioDurations20-25.jpg")
plt.show()





# Define bin width and bins for 0.5 seconds interval between 5 and 10 seconds
bin_width = 0.5
min_duration = 25
max_duration = 30
bins = np.arange(min_duration, max_duration + bin_width, bin_width)

# Ensure duration_values is a NumPy array
duration_values = np.array(duration_values)

# Filter data to only include durations between 10 and 15 seconds
filtered_durations = duration_values[(duration_values >= min_duration) & (duration_values <= max_duration)]

# Compute histogram data
counts, bin_edges = np.histogram(filtered_durations, bins=bins)

# Prepare dictionary with string keys for each bin range
hist_dict = {}
for i in range(len(counts)):
    key = f"{bin_edges[i]}-{bin_edges[i+1]}"
    hist_dict[key] = int(counts[i])

# Save histogram data to JSON file
json_filename = "audio_duration_histogram_25-30_seconds.json"
with open(json_filename, "w") as f:
    json.dump(hist_dict, f, indent=2)

# Plot histogram
plt.hist(filtered_durations, bins=bins, color='purple', edgecolor='black')
plt.title('Distribution of Audio Durations (20-25 seconds)')
plt.xlabel('Duration in seconds')
plt.ylabel('Count')
plt.xlim(min_duration, max_duration)
# plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/audio_duration//DynamicRangeLRA0-5.jpg")
plt.show()


###########-----------------------------------------__FOR DYNAMIC_RANGE_LRA__-------------------------------------------------------------------########################


dynamic_range_lra = [example['audio_features']['dynamic_range_lra'] for example in data]


bin_width = 0.5
min_duration = 0
max_duration = 5

# Ensure arrays are numpy arrays
duration_values = np.array(duration_values)
dynamic_range_lra_values = np.array(dynamic_range_lra)

# Filter dynamic_range_lra for files with duration between 0 and 5 seconds
filtered_dynamic_range_lra = dynamic_range_lra_values[
    (duration_values >= min_duration) & (duration_values <= max_duration)
]

# Define bins for dynamic_range_lra (0 to 30 in 0.5 steps)
lra_bins = np.arange(0, 5 + bin_width, bin_width)

# Compute histogram for filtered dynamic_range_lra
counts, bin_edges = np.histogram(filtered_dynamic_range_lra, bins=lra_bins)

# Prepare JSON output in requested format
hist_dict = {}
for i in range(len(counts)):
    key = f"{bin_edges[i]}-{bin_edges[i+1]}"
    hist_dict[key] = int(counts[i])

# Save to JSON
json_filename = "dynamic_range_lra_histogram_0_5_seconds.json"
with open(json_filename, "w") as f:
    json.dump(hist_dict, f, indent=2)

# Plot
plt.hist(filtered_dynamic_range_lra, bins=lra_bins, color='purple', edgecolor='black')
plt.title('Distribution of Dynamic Range LRA (0-5 seconds duration)')
plt.xlabel('Dynamic Range LRA')
plt.ylabel('Count')
plt.xlim(0, 5)
plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/audio_duration//DynamicRangeLRA0-5.jpg")
plt.show()




# Define bin width and range for dynamic range LRA between 5-10
bin_width = 0.5
min_lra = 5
max_lra = 10
bins = np.arange(min_lra, max_lra + bin_width, bin_width)

# Ensure data is in numpy arrays
dynamic_range_lra_values = np.array(dynamic_range_lra)

# Filter to only include dynamic range LRA values between 5-10
filtered_dynamic_range_lra = dynamic_range_lra_values[
    (duration_values >= min_duration) & (duration_values <= max_duration)]

# Compute histogram for the filtered values
counts, bin_edges = np.histogram(filtered_dynamic_range_lra, bins=bins)

# Prepare JSON with string keys
hist_dict = {}
for i in range(len(counts)):
    key = f"{bin_edges[i]}-{bin_edges[i+1]}"
    hist_dict[key] = int(counts[i])

# Save to JSON file
json_filename = "dynamic_range_lra_histogram_5_10.json"
with open(json_filename, "w") as f:
    json.dump(hist_dict, f, indent=2)

# Plot histogram
plt.hist(filtered_dynamic_range_lra, bins=bins, color='purple', edgecolor='black')
plt.title('Distribution of Dynamic Range LRA (5-10)')
plt.xlabel('Dynamic Range LRA')
plt.ylabel('Count')
plt.xlim(min_lra, max_lra)
plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/audio_duration//DynamicRangeLRA5-10.jpg")
plt.show()


# Define bin width and range for dynamic range LRA between 10-15
bin_width = 0.5
min_lra = 10
max_lra = 15
bins = np.arange(min_lra, max_lra + bin_width, bin_width)

# Ensure data is in numpy arrays
dynamic_range_lra_values = np.array(dynamic_range_lra)

# Filter to only include dynamic range LRA values between 10-15
filtered_dynamic_range_lra = dynamic_range_lra_values[
    (duration_values >= min_duration) & (duration_values <= max_duration)]

# Compute histogram for the filtered values
counts, bin_edges = np.histogram(filtered_dynamic_range_lra, bins=bins)

# Prepare JSON with string keys
hist_dict = {}
for i in range(len(counts)):
    key = f"{bin_edges[i]}-{bin_edges[i+1]}"
    hist_dict[key] = int(counts[i])

# Save to JSON file
json_filename = "dynamic_range_lra_histogram_10_15.json"
with open(json_filename, "w") as f:
    json.dump(hist_dict, f, indent=2)

# Plot histogram
plt.hist(filtered_dynamic_range_lra, bins=bins, color='purple', edgecolor='black')
plt.title('Distribution of Dynamic Range LRA (10-15)')
plt.xlabel('Dynamic Range LRA')
plt.ylabel('Count')
plt.xlim(min_lra, max_lra)
plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/audio_duration//DynamicRangeLRA10-15.jpg")
plt.show()




# Define bin width and range for dynamic range LRA between 10-15
bin_width = 0.5
min_lra = 15
max_lra = 20
bins = np.arange(min_lra, max_lra + bin_width, bin_width)

# Ensure data is in numpy arrays
dynamic_range_lra_values = np.array(dynamic_range_lra)

# Filter to only include dynamic range LRA values between 10-15
filtered_dynamic_range_lra = dynamic_range_lra_values[
    (duration_values >= min_duration) & (duration_values <= max_duration)]

# Compute histogram for the filtered values
counts, bin_edges = np.histogram(filtered_dynamic_range_lra, bins=bins)

# Prepare JSON with string keys
hist_dict = {}
for i in range(len(counts)):
    key = f"{bin_edges[i]}-{bin_edges[i+1]}"
    hist_dict[key] = int(counts[i])

# Save to JSON file
json_filename = "dynamic_range_lra_histogram_15_20.json"
with open(json_filename, "w") as f:
    json.dump(hist_dict, f, indent=2)

# Plot histogram
plt.hist(filtered_dynamic_range_lra, bins=bins, color='purple', edgecolor='black')
plt.title('Distribution of Dynamic Range LRA (15-20)')
plt.xlabel('Dynamic Range LRA')
plt.ylabel('Count')
plt.xlim(min_lra, max_lra)
plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/audio_duration//DynamicRangeLRA15-20.jpg")
plt.show()




# Define bin width and range for dynamic range LRA between 10-15
bin_width = 0.5
min_lra = 20
max_lra = 25
bins = np.arange(min_lra, max_lra + bin_width, bin_width)

# Ensure data is in numpy arrays
dynamic_range_lra_values = np.array(dynamic_range_lra)

# Filter to only include dynamic range LRA values between 10-15
filtered_dynamic_range_lra = dynamic_range_lra_values[
    (duration_values >= min_duration) & (duration_values <= max_duration)]

# Compute histogram for the filtered values
counts, bin_edges = np.histogram(filtered_dynamic_range_lra, bins=bins)

# Prepare JSON with string keys
hist_dict = {}
for i in range(len(counts)):
    key = f"{bin_edges[i]}-{bin_edges[i+1]}"
    hist_dict[key] = int(counts[i])

# Save to JSON file
json_filename = "dynamic_range_lra_histogram_20_25.json"
with open(json_filename, "w") as f:
    json.dump(hist_dict, f, indent=2)

# Plot histogram
plt.hist(filtered_dynamic_range_lra, bins=bins, color='purple', edgecolor='black')
plt.title('Distribution of Dynamic Range LRA (20-25)')
plt.xlabel('Dynamic Range LRA')
plt.ylabel('Count')
plt.xlim(min_lra, max_lra)
plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/audio_duration//DynamicRangeLRA20-25.jpg")
plt.show()




# Define bin width and range for dynamic range LRA between 10-15
bin_width = 0.5
min_lra = 25
max_lra = 30
bins = np.arange(min_lra, max_lra + bin_width, bin_width)

# Ensure data is in numpy arrays
dynamic_range_lra_values = np.array(dynamic_range_lra)

# Filter to only include dynamic range LRA values between 10-15
filtered_dynamic_range_lra = dynamic_range_lra_values[
    (duration_values >= min_duration) & (duration_values <= max_duration)]

# Compute histogram for the filtered values
counts, bin_edges = np.histogram(filtered_dynamic_range_lra, bins=bins)

# Prepare JSON with string keys
hist_dict = {}
for i in range(len(counts)):
    key = f"{bin_edges[i]}-{bin_edges[i+1]}"
    hist_dict[key] = int(counts[i])

# Save to JSON file
json_filename = "dynamic_range_lra_histogram_25-30.json"
with open(json_filename, "w") as f:
    json.dump(hist_dict, f, indent=2)

# Plot histogram
plt.hist(filtered_dynamic_range_lra, bins=bins, color='purple', edgecolor='black')
plt.title('Distribution of Dynamic Range LRA (25-30)')
plt.xlabel('Dynamic Range LRA')
plt.ylabel('Count')
plt.xlim(min_lra, max_lra)
plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/audio_duration//DynamicRangeLRA25-30.jpg")
plt.show()


##################-------------------------------------FOR LOUDNESS LUFFS----------------------------------------------------------------------####################


loudness_lufs = [example['audio_features']['loudness_lufs'] for example in data]

# Convert to numpy arrays
duration_array = np.array(duration_values)
loudness_array = np.array(loudness_lufs)

# Define duration bins (0 to 5 seconds, 0.5s intervals)
bin_width = 0.5
min_duration = 0
max_duration = 5
bins = np.arange(min_duration, max_duration + bin_width, bin_width)

# Prepare results
loudness_per_duration_bin = {}
x_labels = []
y_values = []

# Loop through bins and compute average loudness
for i in range(len(bins) - 1):
    start = bins[i]
    end = bins[i + 1]
    
    # Mask for samples in the duration bin
    mask = (duration_array >= start) & (duration_array < end)
    loudness_in_bin = loudness_array[mask]
    
    # Remove NaNsplt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/audio_duration/Dynamic-Range-LRA0-5.jpg")
    loudness_in_bin = loudness_in_bin[~np.isnan(loudness_in_bin)]

    # Calculate mean loudness
    if len(loudness_in_bin) > 0:
        avg_loudness = float(np.mean(loudness_in_bin))
    else:
        avg_loudness = None

    # Save results
    key = f"{start}-{end}"
    loudness_per_duration_bin[key] = avg_loudness
    x_labels.append(key)
    y_values.append(avg_loudness if avg_loudness is not None else 0)

# Save to JSON
json_filename = "avg_loudness_lufs_per_duration_bin.json"
with open(json_filename, "w") as f:
    json.dump(loudness_per_duration_bin, f, indent=2)

# Plot histogram-style bar chart
plt.figure(figsize=(10, 6))
plt.bar(x_labels, y_values, color='teal', edgecolor='black')
plt.xticks(rotation=45)
plt.xlabel("Duration Bin (seconds)")
plt.ylabel("Average Loudness (LUFS)")
plt.title("Average Loudness per Duration Bin (0–5 sec)")
plt.grid(axis='y')
plt.tight_layout()
plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/audio_duration/loudness0-5.jpg")
plt.show()






# Convert to numpy arrays
duration_array = np.array(duration_values)
loudness_array = np.array(loudness_lufs)

# Define duration bins (0 to 5 seconds, 0.5s intervals)
bin_width = 0.5
min_duration = 5
max_duration = 10
bins = np.arange(min_duration, max_duration + bin_width, bin_width)

# Prepare results
loudness_per_duration_bin = {}
x_labels = []
y_values = []

# Loop through bins and compute average loudness
for i in range(len(bins) - 1):
    start = bins[i]
    end = bins[i + 1]
    
    # Mask for samples in the duration bin
    mask = (duration_array >= start) & (duration_array < end)
    loudness_in_bin = loudness_array[mask]
    
    # Remove NaNs
    loudness_in_bin = loudness_in_bin[~np.isnan(loudness_in_bin)]

    # Calculate mean loudness
    if len(loudness_in_bin) > 0:
        avg_loudness = float(np.mean(loudness_in_bin))
    else:
        avg_loudness = None

    # Save results
    key = f"{start}-{end}"
    loudness_per_duration_bin[key] = avg_loudness
    x_labels.append(key)
    y_values.append(avg_loudness if avg_loudness is not None else 0)

# Save to JSON
json_filename = "avg_loudness_lufs_per_duration5-10_bin.json"
with open(json_filename, "w") as f:
    json.dump(loudness_per_duration_bin, f, indent=2)

# Plot histogram-style bar chart
plt.figure(figsize=(10, 6))
plt.bar(x_labels, y_values, color='teal', edgecolor='black')
plt.xticks(rotation=45)
plt.xlabel("Duration Bin (seconds)")
plt.ylabel("Average Loudness (LUFS)")
plt.title("Average Loudness per Duration Bin (5-10 sec)")
plt.grid(axis='y')
plt.tight_layout()
plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/audio_duration/loudness5-10.jpg")
plt.show()





# Convert to numpy arrays
duration_array = np.array(duration_values)
loudness_array = np.array(loudness_lufs)

# Define duration bins (0 to 5 seconds, 0.5s intervals)
bin_width = 0.5
min_duration = 10
max_duration = 15
bins = np.arange(min_duration, max_duration + bin_width, bin_width)

# Prepare results
loudness_per_duration_bin = {}
x_labels = []
y_values = []

# Loop through bins and compute average loudness
for i in range(len(bins) - 1):
    start = bins[i]
    end = bins[i + 1]
    
    # Mask for samples in the duration bin
    mask = (duration_array >= start) & (duration_array < end)
    loudness_in_bin = loudness_array[mask]
    
    # Remove NaNs
    loudness_in_bin = loudness_in_bin[~np.isnan(loudness_in_bin)]

    # Calculate mean loudness
    if len(loudness_in_bin) > 0:
        avg_loudness = float(np.mean(loudness_in_bin))
    else:
        avg_loudness = None

    # Save results
    key = f"{start}-{end}"
    loudness_per_duration_bin[key] = avg_loudness
    x_labels.append(key)
    y_values.append(avg_loudness if avg_loudness is not None else 0)

# Save to JSON
json_filename = "avg_loudness_lufs_per_duration10-15_bin.json"
with open(json_filename, "w") as f:
    json.dump(loudness_per_duration_bin, f, indent=2)

# Plot histogram-style bar chart
plt.figure(figsize=(10, 6))
plt.bar(x_labels, y_values, color='teal', edgecolor='black')
plt.xticks(rotation=45)
plt.xlabel("Duration Bin (seconds)")
plt.ylabel("Average Loudness (LUFS)")
plt.title("Average Loudness per Duration Bin (10-15 sec)")
plt.grid(axis='y')
plt.tight_layout()
plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/audio_duration/loudness10-15.jpg")
plt.show()




# Convert to numpy arrays
duration_array = np.array(duration_values)
loudness_array = np.array(loudness_lufs)

# Define duration bins (0 to 5 seconds, 0.5s intervals)
bin_width = 0.5
min_duration = 15
max_duration = 20
bins = np.arange(min_duration, max_duration + bin_width, bin_width)

# Prepare results
loudness_per_duration_bin = {}
x_labels = []
y_values = []

# Loop through bins and compute average loudness
for i in range(len(bins) - 1):
    start = bins[i]
    end = bins[i + 1]
    
    # Mask for samples in the duration bin
    mask = (duration_array >= start) & (duration_array < end)
    loudness_in_bin = loudness_array[mask]
    
    # Remove NaNs
    loudness_in_bin = loudness_in_bin[~np.isnan(loudness_in_bin)]

    # Calculate mean loudness
    if len(loudness_in_bin) > 0:
        avg_loudness = float(np.mean(loudness_in_bin))
    else:
        avg_loudness = None

    # Save results
    key = f"{start}-{end}"
    loudness_per_duration_bin[key] = avg_loudness
    x_labels.append(key)
    y_values.append(avg_loudness if avg_loudness is not None else 0)

# Save to JSON
json_filename = "avg_loudness_lufs_per_duration15-20_bin.json"
with open(json_filename, "w") as f:
    json.dump(loudness_per_duration_bin, f, indent=2)

# Plot histogram-style bar chart
plt.figure(figsize=(10, 6))
plt.bar(x_labels, y_values, color='teal', edgecolor='black')
plt.xticks(rotation=45)
plt.xlabel("Duration Bin (seconds)")
plt.ylabel("Average Loudness (LUFS)")
plt.title("Average Loudness per Duration Bin (15-20sec)")
plt.grid(axis='y')
plt.tight_layout()
plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/audio_duration/loudness15-20.jpg")
plt.show()




# Convert to numpy arrays
duration_array = np.array(duration_values)
loudness_array = np.array(loudness_lufs)

# Define duration bins (0 to 5 seconds, 0.5s intervals)
bin_width = 0.5
min_duration = 20
max_duration = 25
bins = np.arange(min_duration, max_duration + bin_width, bin_width)

# Prepare results
loudness_per_duration_bin = {}
x_labels = []
y_values = []

# Loop through bins and compute average loudness
for i in range(len(bins) - 1):
    start = bins[i]
    end = bins[i + 1]
    
    # Mask for samples in the duration bin
    mask = (duration_array >= start) & (duration_array < end)
    loudness_in_bin = loudness_array[mask]
    
    # Remove NaNs
    loudness_in_bin = loudness_in_bin[~np.isnan(loudness_in_bin)]

    # Calculate mean loudness
    if len(loudness_in_bin) > 0:
        avg_loudness = float(np.mean(loudness_in_bin))
    else:
        avg_loudness = None

    # Save results
    key = f"{start}-{end}"
    loudness_per_duration_bin[key] = avg_loudness
    x_labels.append(key)
    y_values.append(avg_loudness if avg_loudness is not None else 0)

# Save to JSON
json_filename = "avg_loudness_lufs_per_duration20-25_bin.json"
with open(json_filename, "w") as f:
    json.dump(loudness_per_duration_bin, f, indent=2)

# Plot histogram-style bar chart
plt.figure(figsize=(10, 6))
plt.bar(x_labels, y_values, color='teal', edgecolor='black')
plt.xticks(rotation=45)
plt.xlabel("Duration Bin (seconds)")
plt.ylabel("Average Loudness (LUFS)")
plt.title("Average Loudness per Duration Bin (20-25sec)")
plt.grid(axis='y')
plt.tight_layout()
plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/audio_duration/loudness20-25.jpg")
plt.show()







# Convert to numpy arrays
duration_array = np.array(duration_values)
loudness_array = np.array(loudness_lufs)

# Define duration bins (0 to 5 seconds, 0.5s intervals)
bin_width = 0.5
min_duration = 25
max_duration = 30
bins = np.arange(min_duration, max_duration + bin_width, bin_width)

# Prepare results
loudness_per_duration_bin = {}
x_labels = []
y_values = []

# Loop through bins and compute average loudness
for i in range(len(bins) - 1):
    start = bins[i]
    end = bins[i + 1]
    
    # Mask for samples in the duration bin
    mask = (duration_array >= start) & (duration_array < end)
    loudness_in_bin = loudness_array[mask]
    
    # Remove NaNs
    loudness_in_bin = loudness_in_bin[~np.isnan(loudness_in_bin)]

    # Calculate mean loudness
    if len(loudness_in_bin) > 0:
        avg_loudness = float(np.mean(loudness_in_bin))
    else:
        avg_loudness = None

    # Save results
    key = f"{start}-{end}"
    loudness_per_duration_bin[key] = avg_loudness
    x_labels.append(key)
    y_values.append(avg_loudness if avg_loudness is not None else 0)

# Save to JSON
json_filename = "avg_loudness_lufs_per_duration25-30_bin.json"
with open(json_filename, "w") as f:
    json.dump(loudness_per_duration_bin, f, indent=2)

# Plot histogram-style bar chart
plt.figure(figsize=(10, 6))
plt.bar(x_labels, y_values, color='teal', edgecolor='black')
plt.xticks(rotation=45)
plt.xlabel("Duration Bin (seconds)")
plt.ylabel("Average Loudness (LUFS)")
plt.title("Average Loudness per Duration Bin (25-30sec)")
plt.grid(axis='y')
plt.tight_layout()
plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/audio_duration/loudness25-30.jpg")
plt.show()





##########-------------------------AVERAGE MEAN f0--------------------------------------------__#############################################


mean_f0_hz = [example['audio_features']['mean_f0_hz'] for example in data]


# Clean your data: convert to float or np.nan
cleaned_f0 = [float(x) if isinstance(x, (int, float)) and x is not None else np.nan for x in mean_f0_hz]
f0_array = np.array(cleaned_f0, dtype=np.float32)

# Also clean durations to match
cleaned_durations = [float(x) if isinstance(x, (int, float)) and x is not None else np.nan for x in duration_values]
duration_array = np.array(cleaned_durations, dtype=np.float32)

# Define duration bins
bin_width = 0.5
min_duration = 0
max_duration = 5
bins = np.arange(min_duration, max_duration + bin_width, bin_width)

# Prepare results
f0_per_duration_bin = {}
x_labels = []
y_values = []

# Loop through each bin
for i in range(len(bins) - 1):
    start = bins[i]
    end = bins[i + 1]
    
    # Mask valid values only
    mask = (duration_array >= start) & (duration_array < end) & ~np.isnan(f0_array)
    f0_in_bin = f0_array[mask]
    
    avg_f0 = float(np.mean(f0_in_bin)) if len(f0_in_bin) > 0 else None

    key = f"{start}-{end}"
    f0_per_duration_bin[key] = avg_f0
    x_labels.append(key)
    y_values.append(avg_f0 if avg_f0 is not None else 0)

# Save to JSON
json_filename = "avg_mean_f0_hz_per_duration0-5_bin.json"
with open(json_filename, "w") as f:
    json.dump(f0_per_duration_bin, f, indent=2)

# Plot
plt.figure(figsize=(10, 6))
plt.bar(x_labels, y_values, color='salmon', edgecolor='black')
plt.xticks(rotation=45)
plt.xlabel("Duration Bin (seconds)")
plt.ylabel("Average Mean F0 (Hz)")
plt.title("Average Mean F0 per Duration Bin (0–5 sec)")
plt.grid(axis='y')
plt.tight_layout()
plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/audio_duration/AvgMeanF0:0-5.jpg")
plt.show()








# Clean your data: convert to float or np.nan
cleaned_f0 = [float(x) if isinstance(x, (int, float)) and x is not None else np.nan for x in mean_f0_hz]
f0_array = np.array(cleaned_f0, dtype=np.float32)

# Also clean durations to match
cleaned_durations = [float(x) if isinstance(x, (int, float)) and x is not None else np.nan for x in duration_values]
duration_array = np.array(cleaned_durations, dtype=np.float32)

# Define duration bins
bin_width = 0.5
min_duration = 5
max_duration = 10
bins = np.arange(min_duration, max_duration + bin_width, bin_width)

# Prepare results
f0_per_duration_bin = {}
x_labels = []
y_values = []

# Loop through each bin
for i in range(len(bins) - 1):
    start = bins[i]
    end = bins[i + 1]
    
    # Mask valid values only
    mask = (duration_array >= start) & (duration_array < end) & ~np.isnan(f0_array)
    f0_in_bin = f0_array[mask]
    
    avg_f0 = float(np.mean(f0_in_bin)) if len(f0_in_bin) > 0 else None

    key = f"{start}-{end}"
    f0_per_duration_bin[key] = avg_f0
    x_labels.append(key)
    y_values.append(avg_f0 if avg_f0 is not None else 0)

# Save to JSON
json_filename = "avg_mean_f0_hz_per_duration5-10_bin.json"
with open(json_filename, "w") as f:
    json.dump(f0_per_duration_bin, f, indent=2)

# Plot
plt.figure(figsize=(10, 6))
plt.bar(x_labels, y_values, color='salmon', edgecolor='black')
plt.xticks(rotation=45)
plt.xlabel("Duration Bin (seconds)")
plt.ylabel("Average Mean F0 (Hz)")
plt.title("Average Mean F0 per Duration Bin (5-10 sec)")
plt.grid(axis='y')
plt.tight_layout()
plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/audio_duration/AvgMeanF0:5-10.jpg")
plt.show()







# Clean your data: convert to float or np.nan
cleaned_f0 = [float(x) if isinstance(x, (int, float)) and x is not None else np.nan for x in mean_f0_hz]
f0_array = np.array(cleaned_f0, dtype=np.float32)

# Also clean durations to match
cleaned_durations = [float(x) if isinstance(x, (int, float)) and x is not None else np.nan for x in duration_values]
duration_array = np.array(cleaned_durations, dtype=np.float32)

# Define duration bins
bin_width = 0.5
min_duration = 10
max_duration = 15
bins = np.arange(min_duration, max_duration + bin_width, bin_width)

# Prepare results
f0_per_duration_bin = {}
x_labels = []
y_values = []

# Loop through each bin
for i in range(len(bins) - 1):
    start = bins[i]
    end = bins[i + 1]
    
    # Mask valid values only
    mask = (duration_array >= start) & (duration_array < end) & ~np.isnan(f0_array)
    f0_in_bin = f0_array[mask]
    
    avg_f0 = float(np.mean(f0_in_bin)) if len(f0_in_bin) > 0 else None

    key = f"{start}-{end}"
    f0_per_duration_bin[key] = avg_f0
    x_labels.append(key)
    y_values.append(avg_f0 if avg_f0 is not None else 0)

# Save to JSON
json_filename = "avg_mean_f0_hz_per_duration10-15_bin.json"
with open(json_filename, "w") as f:
    json.dump(f0_per_duration_bin, f, indent=2)

# Plot
plt.figure(figsize=(10, 6))
plt.bar(x_labels, y_values, color='salmon', edgecolor='black')
plt.xticks(rotation=45)
plt.xlabel("Duration Bin (seconds)")
plt.ylabel("Average Mean F0 (Hz)")
plt.title("Average Mean F0 per Duration Bin (10-15 sec)")
plt.grid(axis='y')
plt.tight_layout()
plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/audio_duration/AvgMeanF0:10-15.jpg")
plt.show()




# Clean your data: convert to float or np.nan
cleaned_f0 = [float(x) if isinstance(x, (int, float)) and x is not None else np.nan for x in mean_f0_hz]
f0_array = np.array(cleaned_f0, dtype=np.float32)

# Also clean durations to match
cleaned_durations = [float(x) if isinstance(x, (int, float)) and x is not None else np.nan for x in duration_values]
duration_array = np.array(cleaned_durations, dtype=np.float32)

# Define duration bins
bin_width = 0.5
min_duration = 15
max_duration = 20
bins = np.arange(min_duration, max_duration + bin_width, bin_width)

# Prepare results
f0_per_duration_bin = {}
x_labels = []
y_values = []

# Loop through each bin
for i in range(len(bins) - 1):
    start = bins[i]
    end = bins[i + 1]
    
    # Mask valid values only
    mask = (duration_array >= start) & (duration_array < end) & ~np.isnan(f0_array)
    f0_in_bin = f0_array[mask]
    
    avg_f0 = float(np.mean(f0_in_bin)) if len(f0_in_bin) > 0 else None

    key = f"{start}-{end}"
    f0_per_duration_bin[key] = avg_f0
    x_labels.append(key)
    y_values.append(avg_f0 if avg_f0 is not None else 0)

# Save to JSON
json_filename = "avg_mean_f0_hz_per_duration_bin15-20.json"
with open(json_filename, "w") as f:
    json.dump(f0_per_duration_bin, f, indent=2)

# Plot
plt.figure(figsize=(10, 6))
plt.bar(x_labels, y_values, color='salmon', edgecolor='black')
plt.xticks(rotation=45)
plt.xlabel("Duration Bin (seconds)")
plt.ylabel("Average Mean F0 (Hz)")
plt.title("Average Mean F0 per Duration Bin (15-20 sec)")
plt.grid(axis='y')
plt.tight_layout()
plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/audio_duration/AvgMeanF0:15-20.jpg")
plt.show()






# Clean your data: convert to float or np.nan
cleaned_f0 = [float(x) if isinstance(x, (int, float)) and x is not None else np.nan for x in mean_f0_hz]
f0_array = np.array(cleaned_f0, dtype=np.float32)

# Also clean durations to match
cleaned_durations = [float(x) if isinstance(x, (int, float)) and x is not None else np.nan for x in duration_values]
duration_array = np.array(cleaned_durations, dtype=np.float32)

# Define duration bins
bin_width = 0.5
min_duration = 20
max_duration = 25
bins = np.arange(min_duration, max_duration + bin_width, bin_width)

# Prepare results
f0_per_duration_bin = {}
x_labels = []
y_values = []

# Loop through each bin
for i in range(len(bins) - 1):
    start = bins[i]
    end = bins[i + 1]
    
    # Mask valid values only
    mask = (duration_array >= start) & (duration_array < end) & ~np.isnan(f0_array)
    f0_in_bin = f0_array[mask]
    
    avg_f0 = float(np.mean(f0_in_bin)) if len(f0_in_bin) > 0 else None

    key = f"{start}-{end}"
    f0_per_duration_bin[key] = avg_f0
    x_labels.append(key)
    y_values.append(avg_f0 if avg_f0 is not None else 0)

# Save to JSON
json_filename = "avg_mean_f0_hz_per_duration20-25_bin.json"
with open(json_filename, "w") as f:
    json.dump(f0_per_duration_bin, f, indent=2)

# Plot
plt.figure(figsize=(10, 6))
plt.bar(x_labels, y_values, color='salmon', edgecolor='black')
plt.xticks(rotation=45)
plt.xlabel("Duration Bin (seconds)")
plt.ylabel("Average Mean F0 (Hz)")
plt.title("Average Mean F0 per Duration Bin (20-25 sec)")
plt.grid(axis='y')
plt.tight_layout()
plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/audio_duration/AvgMeanF0:20-25.jpg")
plt.show()








# Clean your data: convert to float or np.nan
cleaned_f0 = [float(x) if isinstance(x, (int, float)) and x is not None else np.nan for x in mean_f0_hz]
f0_array = np.array(cleaned_f0, dtype=np.float32)

# Also clean durations to match
cleaned_durations = [float(x) if isinstance(x, (int, float)) and x is not None else np.nan for x in duration_values]
duration_array = np.array(cleaned_durations, dtype=np.float32)

# Define duration bins
bin_width = 0.5
min_duration = 25
max_duration = 30
bins = np.arange(min_duration, max_duration + bin_width, bin_width)

# Prepare results
f0_per_duration_bin = {}
x_labels = []
y_values = []

# Loop through each bin
for i in range(len(bins) - 1):
    start = bins[i]
    end = bins[i + 1]
    
    # Mask valid values only
    mask = (duration_array >= start) & (duration_array < end) & ~np.isnan(f0_array)
    f0_in_bin = f0_array[mask]
    
    avg_f0 = float(np.mean(f0_in_bin)) if len(f0_in_bin) > 0 else None

    key = f"{start}-{end}"
    f0_per_duration_bin[key] = avg_f0
    x_labels.append(key)
    y_values.append(avg_f0 if avg_f0 is not None else 0)

# Save to JSON
json_filename = "avg_mean_f0_hz_per_duration25-30_bin.json"
with open(json_filename, "w") as f:
    json.dump(f0_per_duration_bin, f, indent=2)

# Plot
plt.figure(figsize=(10, 6))
plt.bar(x_labels, y_values, color='salmon', edgecolor='black')
plt.xticks(rotation=45)
plt.xlabel("Duration Bin (seconds)")
plt.ylabel("Average Mean F0 (Hz)")
plt.title("Average Mean F0 per Duration Bin (25-30 sec)")
plt.grid(axis='y')
plt.tight_layout()
plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/audio_duration/AvgMeanF0:25-30.jpg")
plt.show()







########################--------------------------------__MEAN SPECTRAL_CENTROID--------------------------------------------------------____#########################3

mean_spectral_centroid_hz = [example['audio_features']['mean_spectral_centroid_hz'] for example in data]

# Clean spectral centroid values
cleaned_centroid = [
    float(x) if isinstance(x, (int, float)) and x is not None else np.nan 
    for x in mean_spectral_centroid_hz
]
centroid_array = np.array(cleaned_centroid, dtype=np.float32)

# Clean duration values
cleaned_durations = [
    float(x) if isinstance(x, (int, float)) and x is not None else np.nan 
    for x in duration_values
]
duration_array = np.array(cleaned_durations, dtype=np.float32)

# Define duration bins (0–5s with 0.5s step)
bin_width = 0.5
min_duration = 0
max_duration = 5
bins = np.arange(min_duration, max_duration + bin_width, bin_width)

# Compute average per bin
centroid_per_bin = {}
x_labels = []
y_values = []

for i in range(len(bins) - 1):
    start = bins[i]
    end = bins[i + 1]

    # Filter valid entries
    mask = (duration_array >= start) & (duration_array < end) & ~np.isnan(centroid_array)
    values_in_bin = centroid_array[mask]

    avg_centroid = float(np.mean(values_in_bin)) if len(values_in_bin) > 0 else None

    key = f"{start}-{end}"
    centroid_per_bin[key] = avg_centroid
    x_labels.append(key)
    y_values.append(avg_centroid if avg_centroid is not None else 0)

# Save to JSON
json_filename = "avg_mean_spectral_centroid_hz_per_duration0-5_bin.json"
with open(json_filename, "w") as f:
    json.dump(centroid_per_bin, f, indent=2)

# Plot
plt.figure(figsize=(10, 6))
plt.bar(x_labels, y_values, color='royalblue', edgecolor='black')
plt.xticks(rotation=45)
plt.xlabel("Duration Bin (seconds)")
plt.ylabel("Average Spectral Centroid (Hz)")
plt.title("Average Spectral Centroid per Duration Bin (0–5 sec)")
plt.grid(axis='y')
plt.tight_layout()
plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/audio_duration/AverageSpectral0-5.jpg")
plt.show()




# Clean spectral centroid values
cleaned_centroid = [
    float(x) if isinstance(x, (int, float)) and x is not None else np.nan 
    for x in mean_spectral_centroid_hz
]
centroid_array = np.array(cleaned_centroid, dtype=np.float32)

# Clean duration values
cleaned_durations = [
    float(x) if isinstance(x, (int, float)) and x is not None else np.nan 
    for x in duration_values
]
duration_array = np.array(cleaned_durations, dtype=np.float32)

# Define duration bins (0–5s with 0.5s step)
bin_width = 0.5
min_duration = 5
max_duration = 10
bins = np.arange(min_duration, max_duration + bin_width, bin_width)

# Compute average per bin
centroid_per_bin = {}
x_labels = []
y_values = []

for i in range(len(bins) - 1):
    start = bins[i]
    end = bins[i + 1]

    # Filter valid entries
    mask = (duration_array >= start) & (duration_array < end) & ~np.isnan(centroid_array)
    values_in_bin = centroid_array[mask]

    avg_centroid = float(np.mean(values_in_bin)) if len(values_in_bin) > 0 else None

    key = f"{start}-{end}"
    centroid_per_bin[key] = avg_centroid
    x_labels.append(key)
    y_values.append(avg_centroid if avg_centroid is not None else 0)

# Save to JSON
json_filename = "avg_mean_spectral_centroid_hz_per_duration5-10_bin.json"
with open(json_filename, "w") as f:
    json.dump(centroid_per_bin, f, indent=2)

# Plot
plt.figure(figsize=(10, 6))
plt.bar(x_labels, y_values, color='royalblue', edgecolor='black')
plt.xticks(rotation=45)
plt.xlabel("Duration Bin (seconds)")
plt.ylabel("Average Spectral Centroid (Hz)")
plt.title("Average Spectral Centroid per Duration Bin (5-10 sec)")
plt.grid(axis='y')
plt.tight_layout()
plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/audio_duration/AverageSpectral5-10.jpg")
plt.show()




# Clean spectral centroid values
cleaned_centroid = [
    float(x) if isinstance(x, (int, float)) and x is not None else np.nan 
    for x in mean_spectral_centroid_hz
]
centroid_array = np.array(cleaned_centroid, dtype=np.float32)

# Clean duration values
cleaned_durations = [
    float(x) if isinstance(x, (int, float)) and x is not None else np.nan 
    for x in duration_values
]
duration_array = np.array(cleaned_durations, dtype=np.float32)

# Define duration bins (0–5s with 0.5s step)
bin_width = 0.5
min_duration = 10
max_duration = 15
bins = np.arange(min_duration, max_duration + bin_width, bin_width)

# Compute average per bin
centroid_per_bin = {}
x_labels = []
y_values = []

for i in range(len(bins) - 1):
    start = bins[i]
    end = bins[i + 1]

    # Filter valid entrieplt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/audio_duration/AverageSpectral0-5.jpg")s
    mask = (duration_array >= start) & (duration_array < end) & ~np.isnan(centroid_array)
    values_in_bin = centroid_array[mask]

    avg_centroid = float(np.mean(values_in_bin)) if len(values_in_bin) > 0 else None

    key = f"{start}-{end}"
    centroid_per_bin[key] = avg_centroid
    x_labels.append(key)
    y_values.append(avg_centroid if avg_centroid is not None else 0)

# Save to JSON
json_filename = "avg_mean_spectral_centroid_hz_per_duration10-15_bin.json"
with open(json_filename, "w") as f:
    json.dump(centroid_per_bin, f, indent=2)

# Plot
plt.figure(figsize=(10, 6))
plt.bar(x_labels, y_values, color='royalblue', edgecolor='black')
plt.xticks(rotation=45)
plt.xlabel("Duration Bin (seconds)")
plt.ylabel("Average Spectral Centroid (Hz)")
plt.title("Average Spectral Centroid per Duration Bin (10-15) sec)")
plt.grid(axis='y')
plt.tight_layout()
plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/audio_duration/AverageSpectral10-15.jpg")
plt.show()





# Clean spectral centroid values
cleaned_centroid = [
    float(x) if isinstance(x, (int, float)) and x is not None else np.nan 
    for x in mean_spectral_centroid_hz
]
centroid_array = np.array(cleaned_centroid, dtype=np.float32)

# Clean duration values
cleaned_durations = [
    float(x) if isinstance(x, (int, float)) and x is not None else np.nan 
    for x in duration_values
]
duration_array = np.array(cleaned_durations, dtype=np.float32)

# Define duration bins (0–5s with 0.5s step)
bin_width = 0.5
min_duration = 15
max_duration = 20
bins = np.arange(min_duration, max_duration + bin_width, bin_width)

# Compute average per bin
centroid_per_bin = {}
x_labels = []
y_values = []

for i in range(len(bins) - 1):
    start = bins[i]
    end = bins[i + 1]

    # Filter valid entries
    mask = (duration_array >= start) & (duration_array < end) & ~np.isnan(centroid_array)
    values_in_bin = centroid_array[mask]

    avg_centroid = float(np.mean(values_in_bin)) if len(values_in_bin) > 0 else None

    key = f"{start}-{end}"
    centroid_per_bin[key] = avg_centroid
    x_labels.append(key)
    y_values.append(avg_centroid if avg_centroid is not None else 0)

# Save to JSON
json_filename = "avg_mean_spectral_centroid_hz_per_duration15-20_bin.json"
with open(json_filename, "w") as f:
    json.dump(centroid_per_bin, f, indent=2)

# Plot
plt.figure(figsize=(10, 6))
plt.bar(x_labels, y_values, color='royalblue', edgecolor='black')
plt.xticks(rotation=45)
plt.xlabel("Duration Bin (seconds)")
plt.ylabel("Average Spectral Centroid (Hz)")
plt.title("Average Spectral Centroid per Duration Bin (15-20) sec)")
plt.grid(axis='y')
plt.tight_layout()
plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/audio_duration/AverageSpectral15-20.jpg")
plt.show()






# Clean spectral centroid values
cleaned_centroid = [
    float(x) if isinstance(x, (int, float)) and x is not None else np.nan 
    for x in mean_spectral_centroid_hz
]
centroid_array = np.array(cleaned_centroid, dtype=np.float32)

# Clean duration values
cleaned_durations = [
    float(x) if isinstance(x, (int, float)) and x is not None else np.nan 
    for x in duration_values
]
duration_array = np.array(cleaned_durations, dtype=np.float32)

# Define duration bins (0–5s with 0.5s step)
bin_width = 0.5
min_duration = 20
max_duration = 25
bins = np.arange(min_duration, max_duration + bin_width, bin_width)

# Compute average per bin
centroid_per_bin = {}
x_labels = []
y_values = []

for i in range(len(bins) - 1):
    start = bins[i]
    end = bins[i + 1]

    # Filter valid entries
    mask = (duration_array >= start) & (duration_array < end) & ~np.isnan(centroid_array)
    values_in_bin = centroid_array[mask]

    avg_centroid = float(np.mean(values_in_bin)) if len(values_in_bin) > 0 else None

    key = f"{start}-{end}"
    centroid_per_bin[key] = avg_centroid
    x_labels.append(key)
    y_values.append(avg_centroid if avg_centroid is not None else 0)

# Save to JSON
json_filename = "avg_mean_spectral_centroid_hz_per_duration20-25_bin.json"
with open(json_filename, "w") as f:
    json.dump(centroid_per_bin, f, indent=2)

# Plot
plt.figure(figsize=(10, 6))
plt.bar(x_labels, y_values, color='royalblue', edgecolor='black')
plt.xticks(rotation=45)
plt.xlabel("Duration Bin (seconds)")
plt.ylabel("Average Spectral Centroid (Hz)")
plt.title("Average Spectral Centroid per Duration Bin (20-25) sec)")
plt.grid(axis='y')
plt.tight_layout()
plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/audio_duration/AverageSpectral20-25.jpg")
plt.show()







# Clean spectral centroid values
cleaned_centroid = [
    float(x) if isinstance(x, (int, float)) and x is not None else np.nan 
    for x in mean_spectral_centroid_hz
]
centroid_array = np.array(cleaned_centroid, dtype=np.float32)

# Clean duration values
cleaned_durations = [
    float(x) if isinstance(x, (int, float)) and x is not None else np.nan 
    for x in duration_values
]
duration_array = np.array(cleaned_durations, dtype=np.float32)

# Define duration bins (0–5s with 0.5s step)
bin_width = 0.5
min_duration = 25
max_duration = 30
bins = np.arange(min_duration, max_duration + bin_width, bin_width)

# Compute average per bin
centroid_per_bin = {}
x_labels = []
y_values = []

for i in range(len(bins) - 1):
    start = bins[i]
    end = bins[i + 1]

    # Filter valid entries
    mask = (duration_array >= start) & (duration_array < end) & ~np.isnan(centroid_array)
    values_in_bin = centroid_array[mask]

    avg_centroid = float(np.mean(values_in_bin)) if len(values_in_bin) > 0 else None

    key = f"{start}-{end}"
    centroid_per_bin[key] = avg_centroid
    x_labels.append(key)
    y_values.append(avg_centroid if avg_centroid is not None else 0)

# Save to JSON
json_filename = "avg_mean_spectral_centroid_hz_per_duration25-30_bin.json"
with open(json_filename, "w") as f:
    json.dump(centroid_per_bin, f, indent=2)

# Plot
plt.figure(figsize=(10, 6))
plt.bar(x_labels, y_values, color='royalblue', edgecolor='black')
plt.xticks(rotation=45)
plt.xlabel("Duration Bin (seconds)")
plt.ylabel("Average Spectral Centroid (Hz)")
plt.title("Average Spectral Centroid per Duration Bin (20-25) sec)")
plt.grid(axis='y')
plt.tight_layout()
plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/audio_duration/AverageSpectral25-30jpg")
plt.show()




###--------------------------------------------------SNR ESTIMATE DB----------------------------------------------------------#######

snr_estimate_db= [example['audio_features']['snr_estimate_db'] for example in data]


min_duration = 0
max_duration = 5
bin_width = 0.5
bins = np.arange(min_duration, max_duration + bin_width, bin_width)

cleaned_snr = [float(x) if isinstance(x, (int, float)) and x is not None else np.nan for x in snr_estimate_db]
snr_array = np.array(cleaned_snr, dtype=np.float32)

cleaned_durations = [float(x) if isinstance(x, (int, float)) and x is not None else np.nan for x in duration_values]
duration_array = np.array(cleaned_durations, dtype=np.float32)

snr_per_bin = {}
x_labels = []
y_values = []

for i in range(len(bins) - 1):
    mask = (duration_array >= bins[i]) & (duration_array < bins[i+1]) & ~np.isnan(snr_array)
    values = snr_array[mask]
    avg = float(np.mean(values)) if len(values) > 0 else None
    key = f"{bins[i]}-{bins[i+1]}"
    snr_per_bin[key] = avg
    x_labels.append(key)
    y_values.append(avg if avg is not None else 0)

with open("avg_snr_estimate_db_per_duration0-5_bin.json", "w") as f:
    json.dump(snr_per_bin, f, indent=2)

plt.figure(figsize=(10, 6))
plt.bar(x_labels, y_values, color='teal', edgecolor='black')
plt.xticks(rotation=45)
plt.xlabel("Duration Bin (s)")
plt.ylabel("Avg SNR (dB)")
plt.title("Average SNR Estimate per Duration Bin")
plt.grid(axis='y')
plt.tight_layout()
plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/audio_duration/AverageSNR0-5.jpg")
plt.show()








min_duration = 5
max_duration = 10
bin_width = 0.5
bins = np.arange(min_duration, max_duration + bin_width, bin_width)

cleaned_snr = [float(x) if isinstance(x, (int, float)) and x is not None else np.nan for x in snr_estimate_db]
snr_array = np.array(cleaned_snr, dtype=np.float32)

cleaned_durations = [float(x) if isinstance(x, (int, float)) and x is not None else np.nan for x in duration_values]
duration_array = np.array(cleaned_durations, dtype=np.float32)

snr_per_bin = {}
x_labels = []
y_values = []

for i in range(len(bins) - 1):
    mask = (duration_array >= bins[i]) & (duration_array < bins[i+1]) & ~np.isnan(snr_array)
    values = snr_array[mask]
    avg = float(np.mean(values)) if len(values) > 0 else None
    key = f"{bins[i]}-{bins[i+1]}"
    snr_per_bin[key] = avg
    x_labels.append(key)
    y_values.append(avg if avg is not None else 0)

with open("avg_snr_estimate_db_per_duration5-10_bin.json", "w") as f:
    json.dump(snr_per_bin, f, indent=2)

plt.figure(figsize=(10, 6))
plt.bar(x_labels, y_values, color='teal', edgecolor='black')
plt.xticks(rotation=45)
plt.xlabel("Duration Bin (s)")
plt.ylabel("Avg SNR (dB)")
plt.title("Average SNR Estimate per Duration Bin")
plt.grid(axis='y')
plt.tight_layout()
plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/audio_duration/AverageSNR5-10.jpg")
plt.show()








min_duration = 10
max_duration = 15
bin_width = 0.5
bins = np.arange(min_duration, max_duration + bin_width, bin_width)

cleaned_snr = [float(x) if isinstance(x, (int, float)) and x is not None else np.nan for x in snr_estimate_db]
snr_array = np.array(cleaned_snr, dtype=np.float32)

cleaned_durations = [float(x) if isinstance(x, (int, float)) and x is not None else np.nan for x in duration_values]
duration_array = np.array(cleaned_durations, dtype=np.float32)

snr_per_bin = {}
x_labels = []
y_values = []

for i in range(len(bins) - 1):
    mask = (duration_array >= bins[i]) & (duration_array < bins[i+1]) & ~np.isnan(snr_array)
    values = snr_array[mask]
    avg = float(np.mean(values)) if len(values) > 0 else None
    key = f"{bins[i]}-{bins[i+1]}"
    snr_per_bin[key] = avg
    x_labels.append(key)
    y_values.append(avg if avg is not None else 0)

with open("avg_snr_estimate_db_per_duration10-15_bin.json", "w") as f:
    json.dump(snr_per_bin, f, indent=2)

plt.figure(figsize=(10, 6))
plt.bar(x_labels, y_values, color='teal', edgecolor='black')
plt.xticks(rotation=45)
plt.xlabel("Duration Bin (s)")
plt.ylabel("Avg SNR (dB)")
plt.title("Average SNR Estimate per Duration Bin")
plt.grid(axis='y')
plt.tight_layout()
plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/audio_duration/AverageSNR10-15.jpg")
plt.show()






min_duration = 15
max_duration = 20
bin_width = 0.5
bins = np.arange(min_duration, max_duration + bin_width, bin_width)

cleaned_snr = [float(x) if isinstance(x, (int, float)) and x is not None else np.nan for x in snr_estimate_db]
snr_array = np.array(cleaned_snr, dtype=np.float32)

cleaned_durations = [float(x) if isinstance(x, (int, float)) and x is not None else np.nan for x in duration_values]
duration_array = np.array(cleaned_durations, dtype=np.float32)

snr_per_bin = {}
x_labels = []
y_values = []

for i in range(len(bins) - 1):
    mask = (duration_array >= bins[i]) & (duration_array < bins[i+1]) & ~np.isnan(snr_array)
    values = snr_array[mask]
    avg = float(np.mean(values)) if len(values) > 0 else None
    key = f"{bins[i]}-{bins[i+1]}"
    snr_per_bin[key] = avg
    x_labels.append(key)
    y_values.append(avg if avg is not None else 0)

with open("avg_snr_estimate_db_per_duration15-20_bin.json", "w") as f:
    json.dump(snr_per_bin, f, indent=2)

plt.figure(figsize=(10, 6))
plt.bar(x_labels, y_values, color='teal', edgecolor='black')
plt.xticks(rotation=45)
plt.xlabel("Duration Bin (s)")
plt.ylabel("Avg SNR (dB)")
plt.title("Average SNR Estimate per Duration Bin")
plt.grid(axis='y')
plt.tight_layout()
plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/audio_duration/AverageSNR15-20.jpg")
plt.show()







min_duration = 20
max_duration = 25
bin_width = 0.5
bins = np.arange(min_duration, max_duration + bin_width, bin_width)

cleaned_snr = [float(x) if isinstance(x, (int, float)) and x is not None else np.nan for x in snr_estimate_db]
snr_array = np.array(cleaned_snr, dtype=np.float32)

cleaned_durations = [float(x) if isinstance(x, (int, float)) and x is not None else np.nan for x in duration_values]
duration_array = np.array(cleaned_durations, dtype=np.float32)

snr_per_bin = {}
x_labels = []
y_values = []

for i in range(len(bins) - 1):
    mask = (duration_array >= bins[i]) & (duration_array < bins[i+1]) & ~np.isnan(snr_array)
    values = snr_array[mask]
    avg = float(np.mean(values)) if len(values) > 0 else None
    key = f"{bins[i]}-{bins[i+1]}"
    snr_per_bin[key] = avg
    x_labels.append(key)
    y_values.append(avg if avg is not None else 0)

with open("avg_snr_estimate_db_per_duration20-25_bin.json", "w") as f:
    json.dump(snr_per_bin, f, indent=2)

plt.figure(figsize=(10, 6))
plt.bar(x_labels, y_values, color='teal', edgecolor='black')
plt.xticks(rotation=45)
plt.xlabel("Duration Bin (s)")
plt.ylabel("Avg SNR (dB)")
plt.title("Average SNR Estimate per Duration Bin")
plt.grid(axis='y')
plt.tight_layout()
plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/audio_duration/AverageSNR20-25.jpg")
plt.show()






min_duration = 25
max_duration = 30
bin_width = 0.5
bins = np.arange(min_duration, max_duration + bin_width, bin_width)

cleaned_snr = [float(x) if isinstance(x, (int, float)) and x is not None else np.nan for x in snr_estimate_db]
snr_array = np.array(cleaned_snr, dtype=np.float32)

cleaned_durations = [float(x) if isinstance(x, (int, float)) and x is not None else np.nan for x in duration_values]
duration_array = np.array(cleaned_durations, dtype=np.float32)

snr_per_bin = {}
x_labels = []
y_values = []

for i in range(len(bins) - 1):
    mask = (duration_array >= bins[i]) & (duration_array < bins[i+1]) & ~np.isnan(snr_array)
    values = snr_array[mask]
    avg = float(np.mean(values)) if len(values) > 0 else None
    key = f"{bins[i]}-{bins[i+1]}"
    snr_per_bin[key] = avg
    x_labels.append(key)
    y_values.append(avg if avg is not None else 0)

with open("avg_snr_estimate_db_per_duration25-30_bin.json", "w") as f:
    json.dump(snr_per_bin, f, indent=2)

plt.figure(figsize=(10, 6))
plt.bar(x_labels, y_values, color='teal', edgecolor='black')
plt.xticks(rotation=45)
plt.xlabel("Duration Bin (s)")
plt.ylabel("Avg SNR (dB)")
plt.title("Average SNR Estimate per Duration Bin")
plt.grid(axis='y')
plt.tight_layout()
plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/audio_duration/AverageSNR25-30jpg")
plt.show()







#############-----------------------------------------------------SPEECH PERCENTAGE------------------------------------------------------_####################
speech_percentage= [example['audio_features']['speech_percentage'] for example in data]




min_duration = 0
max_duration = 5
bin_width = 0.5
bins = np.arange(min_duration, max_duration + bin_width, bin_width)

cleaned_speech_pct = [float(x) if isinstance(x, (int, float)) and x is not None else np.nan for x in speech_percentage]
speech_pct_array = np.array(cleaned_speech_pct, dtype=np.float32)

speech_per_bin = {}
x_labels = []
y_values = []

for i in range(len(bins) - 1):
    mask = (duration_array >= bins[i]) & (duration_array < bins[i+1]) & ~np.isnan(speech_pct_array)
    values = speech_pct_array[mask]
    avg = float(np.mean(values)) if len(values) > 0 else None
    key = f"{bins[i]}-{bins[i+1]}"
    speech_per_bin[key] = avg
    x_labels.append(key)
    y_values.append(avg if avg is not None else 0)

with open("avg_speech_percentage_per_duration0-5_bin.json", "w") as f:
    json.dump(speech_per_bin, f, indent=2)

plt.figure(figsize=(10, 6))
plt.bar(x_labels, y_values, color='green', edgecolor='black')
plt.xticks(rotation=45)
plt.xlabel("Duration Bin (s)")
plt.ylabel("Avg Speech %")
plt.title("Average Speech Percentage per Duration Bin")
plt.grid(axis='y')
plt.tight_layout()
plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/avg_speech/AverageSpeechPercentage0-5.jpg")
plt.show()







min_duration = 5
max_duration = 10
bin_width = 0.5
bins = np.arange(min_duration, max_duration + bin_width, bin_width)

cleaned_speech_pct = [float(x) if isinstance(x, (int, float)) and x is not None else np.nan for x in speech_percentage]
speech_pct_array = np.array(cleaned_speech_pct, dtype=np.float32)

speech_per_bin = {}
x_labels = []
y_values = []

for i in range(len(bins) - 1):
    mask = (duration_array >= bins[i]) & (duration_array < bins[i+1]) & ~np.isnan(speech_pct_array)
    values = speech_pct_array[mask]
    avg = float(np.mean(values)) if len(values) > 0 else None
    key = f"{bins[i]}-{bins[i+1]}"
    speech_per_bin[key] = avg
    x_labels.append(key)
    y_values.append(avg if avg is not None else 0)

with open("avg_speech_percentage_per_duration5-10_bin.json", "w") as f:
    json.dump(speech_per_bin, f, indent=2)

plt.figure(figsize=(10, 6))
plt.bar(x_labels, y_values, color='green', edgecolor='black')
plt.xticks(rotation=45)
plt.xlabel("Duration Bin (s)")
plt.ylabel("Avg Speech %")
plt.title("Average Speech Percentage per Duration Bin")
plt.grid(axis='y')
plt.tight_layout()
plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/audio_duration/AverageSpeechPercentage5-10.jpg")
plt.show()







min_duration = 10
max_duration = 15
bin_width = 0.5
bins = np.arange(min_duration, max_duration + bin_width, bin_width)

cleaned_speech_pct = [float(x) if isinstance(x, (int, float)) and x is not None else np.nan for x in speech_percentage]
speech_pct_array = np.array(cleaned_speech_pct, dtype=np.float32)

speech_per_bin = {}
x_labels = []
y_values = []

for i in range(len(bins) - 1):
    mask = (duration_array >= bins[i]) & (duration_array < bins[i+1]) & ~np.isnan(speech_pct_array)
    values = speech_pct_array[mask]
    avg = float(np.mean(values)) if len(values) > 0 else None
    key = f"{bins[i]}-{bins[i+1]}"
    speech_per_bin[key] = avg
    x_labels.append(key)
    y_values.append(avg if avg is not None else 0)

with open("avg_speech_percentage_per_duration10-15_bin.json", "w") as f:
    json.dump(speech_per_bin, f, indent=2)

plt.figure(figsize=(10, 6))
plt.bar(x_labels, y_values, color='green', edgecolor='black')
plt.xticks(rotation=45)
plt.xlabel("Duration Bin (s)")
plt.ylabel("Avg Speech %")
plt.title("Average Speech Percentage per Duration Bin")
plt.grid(axis='y')
plt.tight_layout()
plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/audio_duration/AverageSpeechPercentage10-15.jpg")
plt.show()







min_duration = 15
max_duration = 20
bin_width = 0.5
bins = np.arange(min_duration, max_duration + bin_width, bin_width)

cleaned_speech_pct = [float(x) if isinstance(x, (int, float)) and x is not None else np.nan for x in speech_percentage]
speech_pct_array = np.array(cleaned_speech_pct, dtype=np.float32)

speech_per_bin = {}
x_labels = []
y_values = []

for i in range(len(bins) - 1):
    mask = (duration_array >= bins[i]) & (duration_array < bins[i+1]) & ~np.isnan(speech_pct_array)
    values = speech_pct_array[mask]
    avg = float(np.mean(values)) if len(values) > 0 else None
    key = f"{bins[i]}-{bins[i+1]}"
    speech_per_bin[key] = avg
    x_labels.append(key)
    y_values.append(avg if avg is not None else 0)

with open("avg_speech_percentage_per_duration15-20_bin.json", "w") as f:
    json.dump(speech_per_bin, f, indent=2)

plt.figure(figsize=(10, 6))
plt.bar(x_labels, y_values, color='green', edgecolor='black')
plt.xticks(rotation=45)
plt.xlabel("Duration Bin (s)")
plt.ylabel("Avg Speech %")
plt.title("Average Speech Percentage per Duration Bin")
plt.grid(axis='y')
plt.tight_layout()
plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/audio_duration/AverageSpeechPercentage15-20.jpg")
plt.show()






min_duration = 20
max_duration = 25
bin_width = 0.5
bins = np.arange(min_duration, max_duration + bin_width, bin_width)

cleaned_speech_pct = [float(x) if isinstance(x, (int, float)) and x is not None else np.nan for x in speech_percentage]
speech_pct_array = np.array(cleaned_speech_pct, dtype=np.float32)

speech_per_bin = {}
x_labels = []
y_values = []

for i in range(len(bins) - 1):
    mask = (duration_array >= bins[i]) & (duration_array < bins[i+1]) & ~np.isnan(speech_pct_array)
    values = speech_pct_array[mask]
    avg = float(np.mean(values)) if len(values) > 0 else None
    key = f"{bins[i]}-{bins[i+1]}"
    speech_per_bin[key] = avg
    x_labels.append(key)
    y_values.append(avg if avg is not None else 0)

with open("avg_speech_percentage_per_duration20-25_bin.json", "w") as f:
    json.dump(speech_per_bin, f, indent=2)

plt.figure(figsize=(10, 6))
plt.bar(x_labels, y_values, color='green', edgecolor='black')
plt.xticks(rotation=45)
plt.xlabel("Duration Bin (s)")
plt.ylabel("Avg Speech %")
plt.title("Average Speech Percentage per Duration Bin")
plt.grid(axis='y')
plt.tight_layout()
plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/audio_duration/AverageSpeechPercentage20-25.jpg")
plt.show()






min_duration = 25
max_duration = 30
bin_width = 0.5
bins = np.arange(min_duration, max_duration + bin_width, bin_width)

cleaned_speech_pct = [float(x) if isinstance(x, (int, float)) and x is not None else np.nan for x in speech_percentage]
speech_pct_array = np.array(cleaned_speech_pct, dtype=np.float32)

speech_per_bin = {}
x_labels = []
y_values = []

for i in range(len(bins) - 1):
    mask = (duration_array >= bins[i]) & (duration_array < bins[i+1]) & ~np.isnan(speech_pct_array)
    values = speech_pct_array[mask]
    avg = float(np.mean(values)) if len(values) > 0 else None
    key = f"{bins[i]}-{bins[i+1]}"
    speech_per_bin[key] = avg
    x_labels.append(key)
    y_values.append(avg if avg is not None else 0)

with open("avg_speech_percentage_per_duration25-30_bin.json", "w") as f:
    json.dump(speech_per_bin, f, indent=2)

plt.figure(figsize=(10, 6))
plt.bar(x_labels, y_values, color='green', edgecolor='black')
plt.xticks(rotation=45)
plt.xlabel("Duration Bin (s)")
plt.ylabel("Avg Speech %")
plt.title("Average Speech Percentage per Duration Bin")
plt.grid(axis='y')
plt.tight_layout()
plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/audio_duration/AverageSpeechPercentage25-30.jpg")
plt.show()




#########################------------------_______AVG_STD_DEV_F0--------------------------------------------------___________############################



std_dev_f0_hz = [example['audio_features']['std_dev_f0_hz'] for example in data]


min_duration = 0
max_duration = 5
bin_width = 0.5
bins = np.arange(min_duration, max_duration + bin_width, bin_width)

cleaned_std_f0 = [float(x) if isinstance(x, (int, float)) and x is not None else np.nan for x in std_dev_f0_hz]
std_f0_array = np.array(cleaned_std_f0, dtype=np.float32)

std_f0_per_bin = {}
x_labels = []
y_values = []

for i in range(len(bins) - 1):
    mask = (duration_array >= bins[i]) & (duration_array < bins[i+1]) & ~np.isnan(std_f0_array)
    values = std_f0_array[mask]
    avg = float(np.mean(values)) if len(values) > 0 else None
    key = f"{bins[i]}-{bins[i+1]}"
    std_f0_per_bin[key] = avg
    x_labels.append(key)
    y_values.append(avg if avg is not None else 0)

with open("avg_std_dev_f0_hz_per_duration0-5_bin.json", "w") as f:
    json.dump(std_f0_per_bin, f, indent=2)

plt.figure(figsize=(10, 6))
plt.bar(x_labels, y_values, color='violet', edgecolor='black')
plt.xticks(rotation=45)
plt.xlabel("Duration Bin (s)")
plt.ylabel("Avg Std Dev F0 (Hz)")
plt.title("Avg Std Dev of F0 per Duration Bin")
plt.grid(axis='y')
plt.tight_layout()
plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/audio_duration/AvgStdDev0-5.jpg")
plt.show()






min_duration = 5
max_duration = 10
bin_width = 0.5
bins = np.arange(min_duration, max_duration + bin_width, bin_width)

cleaned_std_f0 = [float(x) if isinstance(x, (int, float)) and x is not None else np.nan for x in std_dev_f0_hz]
std_f0_array = np.array(cleaned_std_f0, dtype=np.float32)

std_f0_per_bin = {}
x_labels = []
y_values = []

for i in range(len(bins) - 1):
    mask = (duration_array >= bins[i]) & (duration_array < bins[i+1]) & ~np.isnan(std_f0_array)
    values = std_f0_array[mask]
    avg = float(np.mean(values)) if len(values) > 0 else None
    key = f"{bins[i]}-{bins[i+1]}"
    std_f0_per_bin[key] = avg
    x_labels.append(key)
    y_values.append(avg if avg is not None else 0)

with open("avg_std_dev_f0_hz_per_duration5-10_bin.json", "w") as f:
    json.dump(std_f0_per_bin, f, indent=2)

plt.figure(figsize=(10, 6))
plt.bar(x_labels, y_values, color='violet', edgecolor='black')
plt.xticks(rotation=45)
plt.xlabel("Duration Bin (s)")
plt.ylabel("Avg Std Dev F0 (Hz)")
plt.title("Avg Std Dev of F0 per Duration Bin")
plt.grid(axis='y')
plt.tight_layout()
plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/audio_duration/AvgStdDev5-10.jpg")
plt.show()






min_duration = 10
max_duration = 15
bin_width = 0.5
bins = np.arange(min_duration, max_duration + bin_width, bin_width)

cleaned_std_f0 = [float(x) if isinstance(x, (int, float)) and x is not None else np.nan for x in std_dev_f0_hz]
std_f0_array = np.array(cleaned_std_f0, dtype=np.float32)

std_f0_per_bin = {}
x_labels = []
y_values = []

for i in range(len(bins) - 1):
    mask = (duration_array >= bins[i]) & (duration_array < bins[i+1]) & ~np.isnan(std_f0_array)
    values = std_f0_array[mask]
    avg = float(np.mean(values)) if len(values) > 0 else None
    key = f"{bins[i]}-{bins[i+1]}"
    std_f0_per_bin[key] = avg
    x_labels.append(key)
    y_values.append(avg if avg is not None else 0)

with open("avg_std_dev_f0_hz_per_duration10-15_bin.json", "w") as f:
    json.dump(std_f0_per_bin, f, indent=2)

plt.figure(figsize=(10, 6))
plt.bar(x_labels, y_values, color='violet', edgecolor='black')
plt.xticks(rotation=45)
plt.xlabel("Duration Bin (s)")
plt.ylabel("Avg Std Dev F0 (Hz)")
plt.title("Avg Std Dev of F0 per Duration Bin")
plt.grid(axis='y')
plt.tight_layout()
plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/audio_duration/AvgStdDev10-15.jpg")
plt.show()




min_duration = 15
max_duration = 20
bin_width = 0.5
bins = np.arange(min_duration, max_duration + bin_width, bin_width)

cleaned_std_f0 = [float(x) if isinstance(x, (int, float)) and x is not None else np.nan for x in std_dev_f0_hz]
std_f0_array = np.array(cleaned_std_f0, dtype=np.float32)

std_f0_per_bin = {}
x_labels = []
y_values = []

for i in range(len(bins) - 1):
    mask = (duration_array >= bins[i]) & (duration_array < bins[i+1]) & ~np.isnan(std_f0_array)
    values = std_f0_array[mask]
    avg = float(np.mean(values)) if len(values) > 0 else None
    key = f"{bins[i]}-{bins[i+1]}"
    std_f0_per_bin[key] = avg
    x_labels.append(key)
    y_values.append(avg if avg is not None else 0)

with open("avg_std_dev_f0_hz_per_duration15-20_bin.json", "w") as f:
    json.dump(std_f0_per_bin, f, indent=2)

plt.figure(figsize=(10, 6))
plt.bar(x_labels, y_values, color='violet', edgecolor='black')
plt.xticks(rotation=45)
plt.xlabel("Duration Bin (s)")
plt.ylabel("Avg Std Dev F0 (Hz)")
plt.title("Avg Std Dev of F0 per Duration Bin")
plt.grid(axis='y')
plt.tight_layout()
plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/audio_duration/AvgStdDev15-20.jpg")
plt.show()





min_duration = 20
max_duration = 25
bin_width = 0.5
bins = np.arange(min_duration, max_duration + bin_width, bin_width)

cleaned_std_f0 = [float(x) if isinstance(x, (int, float)) and x is not None else np.nan for x in std_dev_f0_hz]
std_f0_array = np.array(cleaned_std_f0, dtype=np.float32)

std_f0_per_bin = {}
x_labels = []
y_values = []

for i in range(len(bins) - 1):
    mask = (duration_array >= bins[i]) & (duration_array < bins[i+1]) & ~np.isnan(std_f0_array)
    values = std_f0_array[mask]
    avg = float(np.mean(values)) if len(values) > 0 else None
    key = f"{bins[i]}-{bins[i+1]}"
    std_f0_per_bin[key] = avg
    x_labels.append(key)
    y_values.append(avg if avg is not None else 0)

with open("avg_std_dev_f0_hz_per_duration20-25_bin.json", "w") as f:
    json.dump(std_f0_per_bin, f, indent=2)

plt.figure(figsize=(10, 6))
plt.bar(x_labels, y_values, color='violet', edgecolor='black')
plt.xticks(rotation=45)
plt.xlabel("Duration Bin (s)")
plt.ylabel("Avg Std Dev F0 (Hz)")
plt.title("Avg Std Dev of F0 per Duration Bin")
plt.grid(axis='y')
plt.tight_layout()
plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/audio_duration/AvgStdDev20-25.jpg")
plt.show()




min_duration = 25
max_duration = 30
bin_width = 0.5
bins = np.arange(min_duration, max_duration + bin_width, bin_width)

cleaned_std_f0 = [float(x) if isinstance(x, (int, float)) and x is not None else np.nan for x in std_dev_f0_hz]
std_f0_array = np.array(cleaned_std_f0, dtype=np.float32)

std_f0_per_bin = {}
x_labels = []
y_values = []

for i in range(len(bins) - 1):
    mask = (duration_array >= bins[i]) & (duration_array < bins[i+1]) & ~np.isnan(std_f0_array)
    values = std_f0_array[mask]
    avg = float(np.mean(values)) if len(values) > 0 else None
    key = f"{bins[i]}-{bins[i+1]}"
    std_f0_per_bin[key] = avg
    x_labels.append(key)
    y_values.append(avg if avg is not None else 0)

with open("avg_std_dev_f0_hz_per_duration25-30_bin.json", "w") as f:
    json.dump(std_f0_per_bin, f, indent=2)

plt.figure(figsize=(10, 6))
plt.bar(x_labels, y_values, color='violet', edgecolor='black')
plt.xticks(rotation=45)
plt.xlabel("Duration Bin (s)")
plt.ylabel("Avg Std Dev F0 (Hz)")
plt.title("Avg Std Dev of F0 per Duration Bin")
plt.grid(axis='y')
plt.tight_layout()
plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/audio_duration/AvgStdDev25-30.jpg")
plt.show()




###############################3-------------------------------------_BANDWIDTH CATEGORY--------------------------------------------------------------#######

bandwidth_category = [example['audio_features']['bandwidth_category'] for example in data]

import numpy as np
import matplotlib.pyplot as plt
import json


# Define duration bin range
min_duration = 0
max_duration = 5
bin_width = 0.5
bins = np.arange(min_duration, max_duration + bin_width, bin_width)

# Convert lists to arrays
bandwidth_category_array = np.array(bandwidth_category, dtype=object)
duration_array = np.array(duration_values, dtype=np.float32)

# Create result dict
bandwidth_per_bin = {}
x_labels = []

for i in range(len(bins) - 1):
    start, end = bins[i], bins[i + 1]
    # Mask: duration in bin and bandwidth_category is valid
    mask = (
        (duration_array >= start) &
        (duration_array < end) &
        np.array([x is not None and x != '' for x in bandwidth_category_array])
    )

    # Filter categories
    categories_in_bin = bandwidth_category_array[mask]

    # Count occurrences of each category
    category_counts = {}
    for category in categories_in_bin:
        category_counts[category] = category_counts.get(category, 0) + 1

    # Save to dict
    key = f"{start}-{end}"
    bandwidth_per_bin[key] = category_counts
    x_labels.append(key)

# Save result as JSON
with open("bandwidth_category_per_duration_bin.json", "w") as f:
    json.dump(bandwidth_per_bin, f, indent=2)

# Prepare data for plotting
category_labels = list(set(cat for bin_data in bandwidth_per_bin.values() for cat in bin_data.keys()))
category_counts = {category: [] for category in category_labels}

for label in x_labels:
    bin_data = bandwidth_per_bin[label]
    for category in category_labels:
        category_counts[category].append(bin_data.get(category, 0))

# Plot
plt.figure(figsize=(10, 6))
bar_width = 0.1

for idx, category in enumerate(category_labels):
    plt.bar(
        np.arange(len(x_labels)) + idx * bar_width,
        category_counts[category],
        width=bar_width,
        label=category
    )

plt.xticks(np.arange(len(x_labels)) + bar_width * (len(category_labels) - 1) / 2, x_labels, rotation=45)
plt.xlabel("Duration Bin (s)")
plt.ylabel("Count")
plt.title("Bandwidth Category per Duration Bin")
plt.legend(title="Bandwidth Category")
plt.tight_layout()

plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/bandwidth_category/bandwidth_duration0-5_histogram.jpg")
plt.show()







import numpy as np
import matplotlib.pyplot as plt
import json


# Define duration bin range
min_duration = 5
max_duration = 10
bin_width = 0.5
bins = np.arange(min_duration, max_duration + bin_width, bin_width)

# Convert lists to arrays
bandwidth_category_array = np.array(bandwidth_category, dtype=object)
duration_array = np.array(duration_values, dtype=np.float32)

# Create result dict
bandwidth_per_bin = {}
x_labels = []

for i in range(len(bins) - 1):
    start, end = bins[i], bins[i + 1]
    # Mask: duration in bin and bandwidth_category is valid
    mask = (
        (duration_array >= start) &
        (duration_array < end) &
        np.array([x is not None and x != '' for x in bandwidth_category_array])
    )

    # Filter categories
    categories_in_bin = bandwidth_category_array[mask]

    # Count occurrences of each category
    category_counts = {}
    for category in categories_in_bin:
        category_counts[category] = category_counts.get(category, 0) + 1

    # Save to dict
    key = f"{start}-{end}"
    bandwidth_per_bin[key] = category_counts
    x_labels.append(key)

# Save result as JSON
with open("bandwidth_category_per_duration5_bin.json", "w") as f:
    json.dump(bandwidth_per_bin, f, indent=2)

# Prepare data for plotting
category_labels = list(set(cat for bin_data in bandwidth_per_bin.values() for cat in bin_data.keys()))
category_counts = {category: [] for category in category_labels}

for label in x_labels:
    bin_data = bandwidth_per_bin[label]
    for category in category_labels:
        category_counts[category].append(bin_data.get(category, 0))

# Plot
plt.figure(figsize=(10, 6))
bar_width = 0.1

for idx, category in enumerate(category_labels):
    plt.bar(
        np.arange(len(x_labels)) + idx * bar_width,
        category_counts[category],
        width=bar_width,
        label=category
    )

plt.xticks(np.arange(len(x_labels)) + bar_width * (len(category_labels) - 1) / 2, x_labels, rotation=45)
plt.xlabel("Duration Bin (s)")
plt.ylabel("Count")
plt.title("Bandwidth Category per Duration Bin")
plt.legend(title="Bandwidth Category")
plt.tight_layout()

plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/bandwidth_category/bandwidth_duration5-10_histogram.jpg")
plt.show()







import numpy as np
import matplotlib.pyplot as plt
import json


# Define duration bin range
min_duration = 5
max_duration = 10
bin_width = 0.5
bins = np.arange(min_duration, max_duration + bin_width, bin_width)

# Convert lists to arrays
bandwidth_category_array = np.array(bandwidth_category, dtype=object)
duration_array = np.array(duration_values, dtype=np.float32)

# Create result dict
bandwidth_per_bin = {}
x_labels = []

for i in range(len(bins) - 1):
    start, end = bins[i], bins[i + 1]
    # Mask: duration in bin and bandwidth_category is valid
    mask = (
        (duration_array >= start) &
        (duration_array < end) &
        np.array([x is not None and x != '' for x in bandwidth_category_array])
    )

    # Filter categories
    categories_in_bin = bandwidth_category_array[mask]

    # Count occurrences of each category
    category_counts = {}
    for category in categories_in_bin:
        category_counts[category] = category_counts.get(category, 0) + 1

    # Save to dict
    key = f"{start}-{end}"
    bandwidth_per_bin[key] = category_counts
    x_labels.append(key)

# Save result as JSON
with open("bandwidth_category_per_duration10-15_bin.json", "w") as f:
    json.dump(bandwidth_per_bin, f, indent=2)

# Prepare data for plotting
category_labels = list(set(cat for bin_data in bandwidth_per_bin.values() for cat in bin_data.keys()))
category_counts = {category: [] for category in category_labels}

for label in x_labels:
    bin_data = bandwidth_per_bin[label]
    for category in category_labels:
        category_counts[category].append(bin_data.get(category, 0))

# Plot
plt.figure(figsize=(10, 6))
bar_width = 0.1

for idx, category in enumerate(category_labels):
    plt.bar(
        np.arange(len(x_labels)) + idx * bar_width,
        category_counts[category],
        width=bar_width,
        label=category
    )

plt.xticks(np.arange(len(x_labels)) + bar_width * (len(category_labels) - 1) / 2, x_labels, rotation=45)
plt.xlabel("Duration Bin (s)")
plt.ylabel("Count")
plt.title("Bandwidth Category per Duration Bin")
plt.legend(title="Bandwidth Category")
plt.tight_layout()

plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/bandwidth_category/bandwidth_duration10-15_histogram.jpg")
plt.show()







import numpy as np
import matplotlib.pyplot as plt
import json


# Define duration bin range
min_duration = 15
max_duration = 20
bin_width = 0.5
bins = np.arange(min_duration, max_duration + bin_width, bin_width)

# Convert lists to arrays
bandwidth_category_array = np.array(bandwidth_category, dtype=object)
duration_array = np.array(duration_values, dtype=np.float32)

# Create result dict
bandwidth_per_bin = {}
x_labels = []

for i in range(len(bins) - 1):
    start, end = bins[i], bins[i + 1]
    # Mask: duration in bin and bandwidth_category is valid
    mask = (
        (duration_array >= start) &
        (duration_array < end) &
        np.array([x is not None and x != '' for x in bandwidth_category_array])
    )

    # Filter categories
    categories_in_bin = bandwidth_category_array[mask]

    # Count occurrences of each category
    category_counts = {}
    for category in categories_in_bin:
        category_counts[category] = category_counts.get(category, 0) + 1

    # Save to dict
    key = f"{start}-{end}"
    bandwidth_per_bin[key] = category_counts
    x_labels.append(key)

# Save result as JSON
with open("bandwidth_category_per_duration15-20_bin.json", "w") as f:
    json.dump(bandwidth_per_bin, f, indent=2)

# Prepare data for plotting
category_labels = list(set(cat for bin_data in bandwidth_per_bin.values() for cat in bin_data.keys()))
category_counts = {category: [] for category in category_labels}

for label in x_labels:
    bin_data = bandwidth_per_bin[label]
    for category in category_labels:
        category_counts[category].append(bin_data.get(category, 0))

# Plot
plt.figure(figsize=(10, 6))
bar_width = 0.1

for idx, category in enumerate(category_labels):
    plt.bar(
        np.arange(len(x_labels)) + idx * bar_width,
        category_counts[category],
        width=bar_width,
        label=category
    )

plt.xticks(np.arange(len(x_labels)) + bar_width * (len(category_labels) - 1) / 2, x_labels, rotation=45)
plt.xlabel("Duration Bin (s)")
plt.ylabel("Count")
plt.title("Bandwidth Category per Duration Bin")
plt.legend(title="Bandwidth Category")
plt.tight_layout()
plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/bandwidth_category/bandwidth_duration15-20_histogram.jpg")
plt.show()






import numpy as np
import matplotlib.pyplot as plt
import json


# Define duration bin range
min_duration = 20
max_duration = 25
bin_width = 0.5
bins = np.arange(min_duration, max_duration + bin_width, bin_width)

# Convert lists to arrays
bandwidth_category_array = np.array(bandwidth_category, dtype=object)
duration_array = np.array(duration_values, dtype=np.float32)

# Create result dict
bandwidth_per_bin = {}
x_labels = []

for i in range(len(bins) - 1):
    start, end = bins[i], bins[i + 1]
    # Mask: duration in bin and bandwidth_category is valid
    mask = (
        (duration_array >= start) &
        (duration_array < end) &
        np.array([x is not None and x != '' for x in bandwidth_category_array])
    )

    # Filter categories
    categories_in_bin = bandwidth_category_array[mask]

    # Count occurrences of each category
    category_counts = {}
    for category in categories_in_bin:
        category_counts[category] = category_counts.get(category, 0) + 1

    # Save to dict
    key = f"{start}-{end}"
    bandwidth_per_bin[key] = category_counts
    x_labels.append(key)

# Save result as JSON
with open("bandwidth_category_per_duration20-25_bin.json", "w") as f:
    json.dump(bandwidth_per_bin, f, indent=2)

# Prepare data for plotting
category_labels = list(set(cat for bin_data in bandwidth_per_bin.values() for cat in bin_data.keys()))
category_counts = {category: [] for category in category_labels}

for label in x_labels:
    bin_data = bandwidth_per_bin[label]
    for category in category_labels:
        category_counts[category].append(bin_data.get(category, 0))

# Plot
plt.figure(figsize=(10, 6))
bar_width = 0.1

for idx, category in enumerate(category_labels):
    plt.bar(
        np.arange(len(x_labels)) + idx * bar_width,
        category_counts[category],
        width=bar_width,
        label=category
    )

plt.xticks(np.arange(len(x_labels)) + bar_width * (len(category_labels) - 1) / 2, x_labels, rotation=45)
plt.xlabel("Duration Bin (s)")
plt.ylabel("Count")
plt.title("Bandwidth Category per Duration Bin")
plt.legend(title="Bandwidth Category")
plt.tight_layout()
plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/bandwidth_category/bandwidth_duration20-25_histogram.jpg")
plt.show()








import numpy as np
import matplotlib.pyplot as plt
import json


# Define duration bin range
min_duration = 20
max_duration = 25
bin_width = 0.5
bins = np.arange(min_duration, max_duration + bin_width, bin_width)

# Convert lists to arrays
bandwidth_category_array = np.array(bandwidth_category, dtype=object)
duration_array = np.array(duration_values, dtype=np.float32)

# Create result dict
bandwidth_per_bin = {}
x_labels = []

for i in range(len(bins) - 1):
    start, end = bins[i], bins[i + 1]
    # Mask: duration in bin and bandwidth_category is valid
    mask = (
        (duration_array >= start) &
        (duration_array < end) &
        np.array([x is not None and x != '' for x in bandwidth_category_array])
    )

    # Filter categories
    categories_in_bin = bandwidth_category_array[mask]

    # Count occurrences of each category
    category_counts = {}
    for category in categories_in_bin:
        category_counts[category] = category_counts.get(category, 0) + 1

    # Save to dict
    key = f"{start}-{end}"
    bandwidth_per_bin[key] = category_counts
    x_labels.append(key)

# Save result as JSON
with open("bandwidth_category_per_duration20-25_bin.json", "w") as f:
    json.dump(bandwidth_per_bin, f, indent=2)

# Prepare data for plotting
category_labels = list(set(cat for bin_data in bandwidth_per_bin.values() for cat in bin_data.keys()))
category_counts = {category: [] for category in category_labels}

for label in x_labels:
    bin_data = bandwidth_per_bin[label]
    for category in category_labels:
        category_counts[category].append(bin_data.get(category, 0))

# Plot
plt.figure(figsize=(10, 6))
bar_width = 0.1

for idx, category in enumerate(category_labels):
    plt.bar(
        np.arange(len(x_labels)) + idx * bar_width,
        category_counts[category],
        width=bar_width,
        label=category
    )

plt.xticks(np.arange(len(x_labels)) + bar_width * (len(category_labels) - 1) / 2, x_labels, rotation=45)
plt.xlabel("Duration Bin (s)")
plt.ylabel("Count")
plt.title("Bandwidth Category per Duration Bin")
plt.legend(title="Bandwidth Category")
plt.tight_layout()
plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/bandwidth_category/bandwidth_duration20-25_histogram.jpg")
plt.show()






import numpy as np
import matplotlib.pyplot as plt
import json


# Define duration bin range
min_duration = 25
max_duration = 30
bin_width = 0.5
bins = np.arange(min_duration, max_duration + bin_width, bin_width)

# Convert lists to arrays
bandwidth_category_array = np.array(bandwidth_category, dtype=object)
duration_array = np.array(duration_values, dtype=np.float32)

# Create result dict
bandwidth_per_bin = {}
x_labels = []

for i in range(len(bins) - 1):
    start, end = bins[i], bins[i + 1]
    # Mask: duration in bin and bandwidth_category is valid
    mask = (
        (duration_array >= start) &
        (duration_array < end) &
        np.array([x is not None and x != '' for x in bandwidth_category_array])
    )

    # Filter categories
    categories_in_bin = bandwidth_category_array[mask]

    # Count occurrences of each category
    category_counts = {}
    for category in categories_in_bin:
        category_counts[category] = category_counts.get(category, 0) + 1

    # Save to dict
    key = f"{start}-{end}"
    bandwidth_per_bin[key] = category_counts
    x_labels.append(key)

# Save result as JSON
with open("bandwidth_category_per_duration25-30_bin.json", "w") as f:
    json.dump(bandwidth_per_bin, f, indent=2)

# Prepare data for plotting
category_labels = list(set(cat for bin_data in bandwidth_per_bin.values() for cat in bin_data.keys()))
category_counts = {category: [] for category in category_labels}

for label in x_labels:
    bin_data = bandwidth_per_bin[label]
    for category in category_labels:
        category_counts[category].append(bin_data.get(category, 0))

# Plot
plt.figure(figsize=(10, 6))
bar_width = 0.1

for idx, category in enumerate(category_labels):
    plt.bar(
        np.arange(len(x_labels)) + idx * bar_width,
        category_counts[category],
        width=bar_width,
        label=category
    )

plt.xticks(np.arange(len(x_labels)) + bar_width * (len(category_labels) - 1) / 2, x_labels, rotation=45)
plt.xlabel("Duration Bin (s)")
plt.ylabel("Count")
plt.title("Bandwidth Category per Duration Bin")
plt.legend(title="Bandwidth Category")
plt.tight_layout()
# plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/bandwidth_category/audio_duration_histogram0-5.jpg")
plt.show()





##########--------------------------CLIPPING PERCENTAGE----------------------------------________############################

import numpy as np
import matplotlib.pyplot as plt
import json

# Extract data
clipping_percentage = [example['audio_features']['clipping_percentage'] for example in data]
duration_values = [example['audio_features']['duration'] for example in data]

# Define duration bin range
min_duration = 0
max_duration = 5
bin_width = 0.5
bins = np.arange(min_duration, max_duration + bin_width, bin_width)

# Convert to numpy arrays
clipping_array = np.array([
    float(x) if isinstance(x, (float, int)) and x is not None else np.nan
    for x in clipping_percentage
], dtype=np.float32)

duration_array = np.array([
    float(x) if isinstance(x, (float, int)) and x is not None else np.nan
    for x in duration_values
], dtype=np.float32)

# Compute average clipping_percentage per duration bin
clipping_per_bin = {}
x_labels = []

for i in range(len(bins) - 1):
    start, end = bins[i], bins[i + 1]
    mask = (duration_array >= start) & (duration_array < end)
    values_in_bin = clipping_array[mask]
    values_in_bin = values_in_bin[~np.isnan(values_in_bin)]

    avg_clip = float(np.mean(values_in_bin)) if len(values_in_bin) > 0 else None

    key = f"{start}-{end}"
    clipping_per_bin[key] = avg_clip
    x_labels.append(key)

# Save JSON
with open("clipping_percentage_per_duration0-5_bin.json", "w") as f:
    json.dump(clipping_per_bin, f, indent=2)

# Plot
heights = [clipping_per_bin[label] if clipping_per_bin[label] is not None else 0 for label in x_labels]

plt.figure(figsize=(10, 6))
plt.bar(x_labels, heights, color='teal', edgecolor='black')
plt.xlabel("Duration Bin (s)")
plt.ylabel("Average Clipping Percentage")
plt.title("Average Clipping Percentage per Duration Bin")
plt.xticks(rotation=45)
plt.tight_layout()

plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/Avg_clipping_percentage/clipping_Percent_histogram0-5.jpg")
plt.show()







import numpy as np
import matplotlib.pyplot as plt
import json

# Define duration bin range
min_duration = 5
max_duration = 10
bin_width = 0.5
bins = np.arange(min_duration, max_duration + bin_width, bin_width)

# Convert to numpy arrays
clipping_array = np.array([
    float(x) if isinstance(x, (float, int)) and x is not None else np.nan
    for x in clipping_percentage
], dtype=np.float32)

duration_array = np.array([
    float(x) if isinstance(x, (float, int)) and x is not None else np.nan
    for x in duration_values
], dtype=np.float32)

# Compute average clipping_percentage per duration bin
clipping_per_bin = {}
x_labels = []

for i in range(len(bins) - 1):
    start, end = bins[i], bins[i + 1]
    mask = (duration_array >= start) & (duration_array < end)
    values_in_bin = clipping_array[mask]
    values_in_bin = values_in_bin[~np.isnan(values_in_bin)]

    avg_clip = float(np.mean(values_in_bin)) if len(values_in_bin) > 0 else None

    key = f"{start}-{end}"
    clipping_per_bin[key] = avg_clip
    x_labels.append(key)

# Save JSON
with open("clipping_percentage_per_duration5-10_bin.json", "w") as f:
    json.dump(clipping_per_bin, f, indent=2)

# Plot
heights = [clipping_per_bin[label] if clipping_per_bin[label] is not None else 0 for label in x_labels]

plt.figure(figsize=(10, 6))
plt.bar(x_labels, heights, color='teal', edgecolor='black')
plt.xlabel("Duration Bin (s)")
plt.ylabel("Average Clipping Percentage")
plt.title("Average Clipping Percentage per Duration Bin")
plt.xticks(rotation=45)
plt.tight_layout()

plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/Avg_clipping_percentage/clipping_Percent_histogram5-10.jpg")
plt.show()








import numpy as np
import matplotlib.pyplot as plt
import json

# Define duration bin range
min_duration = 10
max_duration = 15
bin_width = 0.5
bins = np.arange(min_duration, max_duration + bin_width, bin_width)

# Convert to numpy arrays
clipping_array = np.array([
    float(x) if isinstance(x, (float, int)) and x is not None else np.nan
    for x in clipping_percentage
], dtype=np.float32)

duration_array = np.array([
    float(x) if isinstance(x, (float, int)) and x is not None else np.nan
    for x in duration_values
], dtype=np.float32)

# Compute average clipping_percentage per duration bin
clipping_per_bin = {}
x_labels = []

for i in range(len(bins) - 1):
    start, end = bins[i], bins[i + 1]
    mask = (duration_array >= start) & (duration_array < end)
    values_in_bin = clipping_array[mask]
    values_in_bin = values_in_bin[~np.isnan(values_in_bin)]

    avg_clip = float(np.mean(values_in_bin)) if len(values_in_bin) > 0 else None

    key = f"{start}-{end}"
    clipping_per_bin[key] = avg_clip
    x_labels.append(key)

# Save JSON
with open("clipping_percentage_per_duration10-15_bin.json", "w") as f:
    json.dump(clipping_per_bin, f, indent=2)

# Plot
heights = [clipping_per_bin[label] if clipping_per_bin[label] is not None else 0 for label in x_labels]

plt.figure(figsize=(10, 6))
plt.bar(x_labels, heights, color='teal', edgecolor='black')
plt.xlabel("Duration Bin (s)")
plt.ylabel("Average Clipping Percentage")
plt.title("Average Clipping Percentage per Duration Bin")
plt.xticks(rotation=45)
plt.tight_layout()

plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/Avg_clipping_percentage/clipping_Percent_histogram10-15.jpg")
plt.show()








import numpy as np
import matplotlib.pyplot as plt
import json

# Define duration bin range
min_duration = 15
max_duration = 20
bin_width = 0.5
bins = np.arange(min_duration, max_duration + bin_width, bin_width)

# Convert to numpy arrays
clipping_array = np.array([
    float(x) if isinstance(x, (float, int)) and x is not None else np.nan
    for x in clipping_percentage
], dtype=np.float32)

duration_array = np.array([
    float(x) if isinstance(x, (float, int)) and x is not None else np.nan
    for x in duration_values
], dtype=np.float32)

# Compute average clipping_percentage per duration bin
clipping_per_bin = {}
x_labels = []

for i in range(len(bins) - 1):
    start, end = bins[i], bins[i + 1]
    mask = (duration_array >= start) & (duration_array < end)
    values_in_bin = clipping_array[mask]
    values_in_bin = values_in_bin[~np.isnan(values_in_bin)]

    avg_clip = float(np.mean(values_in_bin)) if len(values_in_bin) > 0 else None

    key = f"{start}-{end}"
    clipping_per_bin[key] = avg_clip
    x_labels.append(key)

# Save JSON
with open("clipping_percentage_per_duration15-20_bin.json", "w") as f:
    json.dump(clipping_per_bin, f, indent=2)

# Plot
heights = [clipping_per_bin[label] if clipping_per_bin[label] is not None else 0 for label in x_labels]

plt.figure(figsize=(10, 6))
plt.bar(x_labels, heights, color='teal', edgecolor='black')
plt.xlabel("Duration Bin (s)")
plt.ylabel("Average Clipping Percentage")
plt.title("Average Clipping Percentage per Duration Bin")
plt.xticks(rotation=45)
plt.tight_layout()

plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/Avg_clipping_percentage/clipping_Percent_histogram15-20.jpg")
plt.show()







import numpy as np
import matplotlib.pyplot as plt
import json

# Define duration bin range
min_duration = 20
max_duration = 25
bin_width = 0.5
bins = np.arange(min_duration, max_duration + bin_width, bin_width)

# Convert to numpy arrays
clipping_array = np.array([
    float(x) if isinstance(x, (float, int)) and x is not None else np.nan
    for x in clipping_percentage
], dtype=np.float32)

duration_array = np.array([
    float(x) if isinstance(x, (float, int)) and x is not None else np.nan
    for x in duration_values
], dtype=np.float32)

# Compute average clipping_percentage per duration bin
clipping_per_bin = {}
x_labels = []

for i in range(len(bins) - 1):
    start, end = bins[i], bins[i + 1]
    mask = (duration_array >= start) & (duration_array < end)
    values_in_bin = clipping_array[mask]
    values_in_bin = values_in_bin[~np.isnan(values_in_bin)]

    avg_clip = float(np.mean(values_in_bin)) if len(values_in_bin) > 0 else None

    key = f"{start}-{end}"
    clipping_per_bin[key] = avg_clip
    x_labels.append(key)

# Save JSON
with open("clipping_percentage_per_duration20-25_bin.json", "w") as f:
    json.dump(clipping_per_bin, f, indent=2)

# Plot
heights = [clipping_per_bin[label] if clipping_per_bin[label] is not None else 0 for label in x_labels]

plt.figure(figsize=(10, 6))
plt.bar(x_labels, heights, color='teal', edgecolor='black')
plt.xlabel("Duration Bin (s)")
plt.ylabel("Average Clipping Percentage")
plt.title("Average Clipping Percentage per Duration Bin")
plt.xticks(rotation=45)
plt.tight_layout()

plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/Avg_clipping_percentage/clipping_Percent_histogram20-25.jpg")
plt.show()







import numpy as np
import matplotlib.pyplot as plt
import json

# Define duration bin range
min_duration = 25
max_duration = 30
bin_width = 0.5
bins = np.arange(min_duration, max_duration + bin_width, bin_width)

# Convert to numpy arrays
clipping_array = np.array([
    float(x) if isinstance(x, (float, int)) and x is not None else np.nan
    for x in clipping_percentage
], dtype=np.float32)

duration_array = np.array([
    float(x) if isinstance(x, (float, int)) and x is not None else np.nan
    for x in duration_values
], dtype=np.float32)

# Compute average clipping_percentage per duration bin
clipping_per_bin = {}
x_labels = []

for i in range(len(bins) - 1):
    start, end = bins[i], bins[i + 1]
    mask = (duration_array >= start) & (duration_array < end)
    values_in_bin = clipping_array[mask]
    values_in_bin = values_in_bin[~np.isnan(values_in_bin)]

    avg_clip = float(np.mean(values_in_bin)) if len(values_in_bin) > 0 else None

    key = f"{start}-{end}"
    clipping_per_bin[key] = avg_clip
    x_labels.append(key)

# Save JSON
with open("clipping_percentage_per_duration25-30bin.json", "w") as f:
    json.dump(clipping_per_bin, f, indent=2)

# Plot
heights = [clipping_per_bin[label] if clipping_per_bin[label] is not None else 0 for label in x_labels]

plt.figure(figsize=(10, 6))
plt.bar(x_labels, heights, color='teal', edgecolor='black')
plt.xlabel("Duration Bin (s)")
plt.ylabel("Average Clipping Percentage")
plt.title("Average Clipping Percentage per Duration Bin")
plt.xticks(rotation=45)
plt.tight_layout()

plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/Avg_clipping_percentage/clipping_Percent_histogram25-30.jpg")
plt.show()





########################_-----------------------------------__SPEAKER COUNT--------------------------------___############################33

import numpy as np
import matplotlib.pyplot as plt
import json

# Work with the Dataset object format
# Your data is already loaded in a variable called 'data'

# Extract duration and speaker count from the dataset
durations = []
speaker_counts = []

# Process each item in the dataset
# For HuggingFace/datasets library
try:
    # This should work for most Dataset objects
    audio_features = data['audio_features']
    
    # If audio_features is directly accessible
    if isinstance(audio_features, list):
        for features in audio_features:
            if features is not None:
                duration = features.get('duration')
                speaker_count = features.get('speaker_count')
                if duration is not None and speaker_count is not None:
                    durations.append(duration)
                    speaker_counts.append(speaker_count)
    else:
        # If we need to iterate through the dataset
        for i in range(len(data)):
            features = data[i]['audio_features']
            if features is not None:
                duration = features.get('duration')
                speaker_count = features.get('speaker_count')
                if duration is not None and speaker_count is not None:
                    durations.append(duration)
                    speaker_counts.append(speaker_count)
except Exception as e:
    print(f"Error accessing dataset: {e}")
    
    # Alternative approach if the above doesn't work
    print("Trying alternative approach...")
    
    # If data is a list of dictionaries
    if isinstance(data, list):
        for item in data:
            if isinstance(item, dict) and 'duration' in item and 'speaker_count' in item:
                duration = item.get('duration')
                speaker_count = item.get('speaker_count')
                if duration is not None and speaker_count is not None:
                    durations.append(duration)
                    speaker_counts.append(speaker_count)

# Convert to numpy arrays
duration_array = np.array(durations)
speaker_count_array = np.array(speaker_counts)

# Print data summary
print(f"Duration array shape: {duration_array.shape}")
print(f"Speaker count array shape: {speaker_count_array.shape}")
if len(durations) > 0:
    print(f"Duration range: {np.min(duration_array):.2f} to {np.max(duration_array):.2f} seconds")
    print(f"Speaker count range: {np.min(speaker_count_array)} to {np.max(speaker_count_array)} speakers")
else:
    print("No valid data extracted!")

# Based on your sample, the durations are quite short (around 2-3 seconds)
# Adjust min_duration and max_duration to match your actual data
min_duration = 0
max_duration = 5  # Increased range to capture more data
bin_width = 0.5
bins = np.arange(min_duration, max_duration + bin_width, bin_width)

# Count speakers per bin
speaker_per_bin = {}
x_labels = []
y_values = []

for i in range(len(bins) - 1):
    mask = (duration_array >= bins[i]) & (duration_array < bins[i+1])
    # Count the number of samples in this bin
    count = np.sum(mask)
    key = f"{bins[i]}-{bins[i+1]}"
    speaker_per_bin[key] = int(count)
    x_labels.append(key)
    y_values.append(count)
    
    # Print counts for debugging
    print(f"Bin {key}: {count} audio files")

# Save to JSON
with open("speaker_count_per_duration_bin0-5.json", "w") as f:
    json.dump(speaker_per_bin, f, indent=2)

# Plot the results
if sum(y_values) > 0:
    plt.figure(figsize=(10, 6))
    plt.bar(x_labels, y_values, color='orange', edgecolor='black')
    plt.xticks(rotation=45)
    plt.xlabel("Duration Bin (s)")
    plt.ylabel("Count")
    plt.title("Number of Audio Files per Duration Bin")
    plt.grid(axis='y')
    plt.tight_layout()
    plt.savefig("/home/sidd/Desktop/work/IME DATASET REVIEW/audio_duration/SpeakerCountPerDurationBin0-5.jpg")
    plt.show()
else:
    print("No data found in the specified duration range! Try adjusting your min_duration and max_duration.")
    
    # If no data in specified range, show the actual duration distribution
    print("\nAnalyzing your actual data distribution...")
    actual_min = np.min(duration_array)
    actual_max = np.max(duration_array)
    print(f"Your data ranges from {actual_min:.2f} to {actual_max:.2f} seconds")
    
    # Suggest a better range based on actual data
    hist, bin_edges = np.histogram(duration_array, bins=10)
    print("Top duration ranges in your data:")
    for i in range(len(hist)):
        if hist[i] > 0:
            print(f"{bin_edges[i]:.2f}-{bin_edges[i+1]:.2f} seconds: {hist[i]} samples")










import numpy as np
import matplotlib.pyplot as plt
import json
import os

# Create output directory if it doesn't exist
output_dir = "/home/sidd/Desktop/work/IME DATASET REVIEW/audio_duration"
os.makedirs(output_dir, exist_ok=True)

# Define duration ranges to analyze
duration_ranges = [
    (5, 10),
    (10, 15),
    (15, 20),
    (20, 25),
    (25, 30)
]

# Extract duration and speaker count from the dataset
durations = []
speaker_counts = []

# Process each item in the dataset
try:
    # For HuggingFace datasets
    audio_features = data['audio_features']
    
    if isinstance(audio_features, list):
        for features in audio_features:
            if features is not None:
                duration = features.get('duration')
                speaker_count = features.get('speaker_count')
                if duration is not None and speaker_count is not None:
                    durations.append(duration)
                    speaker_counts.append(speaker_count)
    else:
        # If we need to iterate through the dataset
        for i in range(len(data)):
            features = data[i]['audio_features']
            if features is not None:
                duration = features.get('duration')
                speaker_count = features.get('speaker_count')
                if duration is not None and speaker_count is not None:
                    durations.append(duration)
                    speaker_counts.append(speaker_count)
except Exception as e:
    print(f"Error accessing dataset: {e}")
    
    # Alternative approach if the above doesn't work
    print("Trying alternative approach...")
    
    # If data is a list of dictionaries
    if isinstance(data, list):
        for item in data:
            if isinstance(item, dict) and 'duration' in item and 'speaker_count' in item:
                duration = item.get('duration')
                speaker_count = item.get('speaker_count')
                if duration is not None and speaker_count is not None:
                    durations.append(duration)
                    speaker_counts.append(speaker_count)

# Convert to numpy arrays
duration_array = np.array(durations)
speaker_count_array = np.array(speaker_counts)

# Print overall data summary
print(f"Total data points extracted: {len(durations)}")
if len(durations) > 0:
    print(f"Overall duration range: {np.min(duration_array):.2f} to {np.max(duration_array):.2f} seconds")
    print(f"Overall speaker count range: {np.min(speaker_count_array)} to {np.max(speaker_count_array)} speakers")
else:
    print("No valid data extracted!")
    exit()

# Process each duration range
for min_duration, max_duration in duration_ranges:
    print(f"\nProcessing duration range {min_duration}-{max_duration} seconds...")
    
    # Filter data for this range
    range_mask = (duration_array >= min_duration) & (duration_array < max_duration)
    range_count = np.sum(range_mask)
    
    print(f"Found {range_count} audio files in this range")
    
    # If no data in this range, skip to the next
    if range_count == 0:
        print(f"No data in the {min_duration}-{max_duration} second range, skipping...")
        continue
    
    # Define bins for this range
    bin_width = 0.5
    bins = np.arange(min_duration, max_duration + bin_width, bin_width)
    
    # Count speakers per bin
    speaker_per_bin = {}
    x_labels = []
    y_values = []
    
    for i in range(len(bins) - 1):
        mask = (duration_array >= bins[i]) & (duration_array < bins[i+1])
        # Count the number of samples in this bin
        count = np.sum(mask)
        key = f"{bins[i]}-{bins[i+1]}"
        speaker_per_bin[key] = int(count)
        x_labels.append(key)
        y_values.append(count)
        
        # Print counts for debugging
        print(f"Bin {key}: {count} audio files")
    
    # Save to JSON for this range
    json_filename = f"speaker_count_per_duration_bin{min_duration}-{max_duration}.json"
    json_path = os.path.join(output_dir, json_filename)
    with open(json_path, "w") as f:
        json.dump(speaker_per_bin, f, indent=2)
    print(f"Saved results to {json_path}")
    
    # Plot the results for this range
    plt.figure(figsize=(12, 6))
    plt.bar(x_labels, y_values, color='orange', edgecolor='black')
    plt.xticks(rotation=90)
    plt.xlabel("Duration Bin (s)")
    plt.ylabel("Count")
    plt.title(f"Number of Audio Files per Duration Bin ({min_duration}-{max_duration}s)")
    plt.grid(axis='y')
    plt.tight_layout()
    
    # Save the plot for this range
    plot_filename = f"SpeakerCountPerDurationBin{min_duration}-{max_duration}.jpg"
    plot_path = os.path.join(output_dir, plot_filename)
    plt.savefig(plot_path)
    print(f"Saved plot to {plot_path}")
    plt.close()

print("\nAll duration ranges processed!")

# Create a combined plot showing summary for all ranges
range_counts = []
range_labels = []

for min_duration, max_duration in duration_ranges:
    range_mask = (duration_array >= min_duration) & (duration_array < max_duration)
    range_count = np.sum(range_mask)
    range_counts.append(range_count)
    range_labels.append(f"{min_duration}-{max_duration}s")

# Plot the summary
plt.figure(figsize=(10, 6))
plt.bar(range_labels, range_counts, color='blue', edgecolor='black')
plt.xlabel("Duration Range (seconds)")
plt.ylabel("Number of Audio Files")
plt.title("Distribution of Audio Files Across Duration Ranges")
plt.grid(axis='y')
plt.tight_layout()

# Save the summary plot
summary_path = os.path.join(output_dir, "DurationRangeSummary.jpg")
plt.savefig(summary_path)
print(f"Saved summary plot to {summary_path}")
plt.close()

print("All processing completed successfully!")