File size: 9,890 Bytes
c6556f2 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 | import os
import sys
import glob
import numpy as np
import pandas as pd
import matplotlib
matplotlib.use('Agg')
import matplotlib.pyplot as plt
# Ensure UTF-8 output
sys.stdout.reconfigure(encoding='utf-8')
BASE_DIR = r"c:\Users\ahmed\.gemini\antigravity\scratch\Allignments\PHONEME_TIMINGS_EACH_RECITER"
# Safe filename conversion for Windows filesystem
def sanitize_filename(name):
replacements = {
':': '_colon_',
'/': '_slash_',
'\\': '_bslash_',
'*': '_star_',
'?': '_qmark_',
'"': '_quote_',
'<': '_lt_',
'>': '_gt_',
'|': '_pipe_'
}
s = str(name)
for char, rep in replacements.items():
s = s.replace(char, rep)
return s
def get_filtered_tokens(csv_path):
df = pd.read_csv(csv_path)
df = df[df['duration'] > 0].copy()
filtered_records = []
unique_phones = sorted(df['phone'].dropna().unique())
for phone in unique_phones:
sub_df = df[df['phone'] == phone].copy()
t_raw = sub_df['duration'].values
if len(t_raw) == 0:
continue
y_log = np.log(t_raw)
q25 = np.percentile(y_log, 25)
q75 = np.percentile(y_log, 75)
iqr = q75 - q25
lower_log = q25 - 1.5 * iqr
upper_log = q75 + 1.5 * iqr
valid_mask = (y_log >= lower_log) & (y_log <= upper_log)
sub_df_filtered = sub_df[valid_mask].copy()
filtered_records.append(sub_df_filtered)
return pd.concat(filtered_records, ignore_index=True)
def main():
print("==================================================")
print("Starting Phase 2: Tempo Normalization against Husary (and Ayman Sowaid for 3 rare tokens)")
print("==================================================")
csv_files = glob.glob(os.path.join(BASE_DIR, "*_phoneme_timestamps.csv"))
reciter_data = {}
for f in csv_files:
reciter_name = os.path.basename(f).replace('_phoneme_timestamps.csv', '')
print(f"Loading & filtering Phase 1 data for: {reciter_name}")
reciter_data[reciter_name] = get_filtered_tokens(f)
# Get all unique phonemes across all datasets
all_phones = sorted(list(set().union(*[df['phone'].unique() for df in reciter_data.values()])))
print(f"\nTotal Unique Phonemes across all datasets: {len(all_phones)}")
# 3 missing tokens in Husary
rare_tokens = {'dˤdˤ', 'xx', 'ħħ'}
print(f"Special reference tokens (using Ayman Sowaid): {rare_tokens}")
# 1. Compute Reference Medians
reference_medians = {}
reference_reciter_map = {}
for p in all_phones:
if p in rare_tokens or p not in reciter_data['Husary']['phone'].values:
ref_reciter = 'Ayman_Sowaid'
else:
ref_reciter = 'Husary'
ref_df = reciter_data[ref_reciter]
sub = ref_df[ref_df['phone'] == p]
ref_med = sub['duration'].median()
reference_medians[p] = ref_med
reference_reciter_map[p] = ref_reciter
# 2. Compute Scaling Factor Matrix S(R, p)
scaling_rows = []
normalized_reciter_data = {}
for reciter, df in reciter_data.items():
df_norm = df.copy()
df_norm['scale_factor'] = 1.0
df_norm['duration_norm'] = df_norm['duration']
for p in df['phone'].unique():
ref_reciter = reference_reciter_map[p]
ref_med = reference_medians[p]
sub_mask = (df['phone'] == p)
reciter_med = df.loc[sub_mask, 'duration'].median()
if reciter == ref_reciter or reciter_med == 0 or np.isnan(reciter_med):
scale = 1.0
else:
scale = ref_med / reciter_med
df_norm.loc[sub_mask, 'scale_factor'] = scale
df_norm.loc[sub_mask, 'duration_norm'] = df.loc[sub_mask, 'duration'] * scale
scaling_rows.append({
'reciter': reciter,
'phone': p,
'reference_reciter': ref_reciter,
'reference_median_ms': round(ref_med * 1000, 2),
'original_median_ms': round(reciter_med * 1000, 2),
'scale_factor': round(scale, 4)
})
normalized_reciter_data[reciter] = df_norm
# Save normalized dataset for this reciter
reciter_dir = os.path.join(BASE_DIR, reciter)
out_csv = os.path.join(reciter_dir, f"{reciter}_phase2_normalized.csv")
df_norm.to_csv(out_csv, index=False, encoding='utf-8-sig')
print(f"Saved normalized tokens for {reciter} -> {out_csv}")
scaling_df = pd.DataFrame(scaling_rows)
scaling_csv_path = os.path.join(BASE_DIR, "phase2_scaling_factors.csv")
scaling_df.to_csv(scaling_csv_path, index=False, encoding='utf-8-sig')
print(f"\nScaling factor matrix saved to: {scaling_csv_path}")
# 3. Compute Phase 2 Post-Normalization Summary Statistics
summary_rows = []
for reciter, df in normalized_reciter_data.items():
for p in sorted(df['phone'].unique()):
sub = df[df['phone'] == p]
t_raw = sub['duration'].values * 1000
t_norm = sub['duration_norm'].values * 1000
summary_rows.append({
'reciter': reciter,
'phone': p,
'reference_used': reference_reciter_map[p],
'scale_factor': round(sub['scale_factor'].iloc[0], 4),
'count': len(sub),
'raw_median_ms': round(np.median(t_raw), 1),
'norm_median_ms': round(np.median(t_norm), 1),
'raw_mean_ms': round(np.mean(t_raw), 1),
'norm_mean_ms': round(np.mean(t_norm), 1),
'norm_min_ms': round(np.min(t_norm), 1),
'norm_max_ms': round(np.max(t_norm), 1),
'norm_std_ms': round(np.std(t_norm), 1),
'norm_min_sec': round(np.min(t_norm) / 1000.0, 4),
'norm_max_sec': round(np.max(t_norm) / 1000.0, 4),
'norm_mean_sec': round(np.mean(t_norm) / 1000.0, 4),
'norm_median_sec': round(np.median(t_norm) / 1000.0, 4)
})
summary_df = pd.DataFrame(summary_rows)
summary_csv_path = os.path.join(BASE_DIR, "all_reciters_phase2_summary.csv")
summary_df.to_csv(summary_csv_path, index=False, encoding='utf-8-sig')
print(f"Master Phase 2 summary saved to: {summary_csv_path}")
# 4. Generate Comparative Multi-Reciter Plots (Before vs. After Normalization)
plots_dir = os.path.join(BASE_DIR, "Phase2_Comparative_Plots")
os.makedirs(plots_dir, exist_ok=True)
color_palette = {
'Husary': '#d90429', # Bold Red (Primary Ref)
'Ayman_Sowaid': '#7209b7', # Purple (Secondary Ref)
'Abdullah_Basfar': '#2b5c8f', # Deep Blue
'Abu_Bakr_Ash-Shaatree': '#0077b6',# Blue
'Alafasy': '#0096c7', # Cyan-Blue
'Ghamadi': '#48cae4', # Light Blue
'Maher_AlMuaiqly': '#f77f00' # Orange
}
print(f"\nGenerating 70 Phase 2 Comparative Visualizations in {plots_dir}...")
for phone in all_phones:
phone_str = str(phone)
safe_phone = sanitize_filename(phone_str)
ref_reciter = reference_reciter_map[phone]
ref_med_ms = reference_medians[phone] * 1000.0
fig, (ax1, ax2) = plt.subplots(2, 1, figsize=(10, 8), dpi=100)
# Upper Subplot: Raw Filtered Durations (Before Tempo Normalization)
for reciter, df in reciter_data.items():
sub = df[df['phone'] == phone]
if len(sub) > 0:
t_ms = sub['duration'].values * 1000.0
ax1.hist(t_ms, bins=30, density=True, alpha=0.35,
color=color_palette.get(reciter, 'gray'),
label=f"{reciter} (Med: {np.median(t_ms):.0f} ms)")
ax1.set_title(f"Phase 2: Before Tempo Normalization | Phoneme [{phone_str}]\n(Different Imams Reciting at Varying Speeds)", fontsize=11, fontweight='bold')
ax1.set_xlabel("Raw Filtered Duration (ms)")
ax1.set_ylabel("Probability Density (PDF)")
ax1.grid(axis='y', alpha=0.3)
ax1.legend(loc='upper right', fontsize=8, framealpha=0.85)
# Lower Subplot: Normalized Durations (Aligned to Reference Timeline)
for reciter, df in normalized_reciter_data.items():
sub = df[df['phone'] == phone]
if len(sub) > 0:
t_norm_ms = sub['duration_norm'].values * 1000.0
lw = 2.0 if reciter == ref_reciter else 1.0
ax2.hist(t_norm_ms, bins=30, density=True, alpha=0.35,
color=color_palette.get(reciter, 'gray'),
label=f"{reciter} (Norm Med: {np.median(t_norm_ms):.0f} ms)")
ax2.axvline(ref_med_ms, color='black', linestyle='--', linewidth=1.8, label=f"Ref Baseline [{ref_reciter}]: {ref_med_ms:.0f} ms")
ax2.set_title(f"Phase 2: After Median-Ratio Normalization to [{ref_reciter}] | Phoneme [{phone_str}]\n(All 7 Imams Perfectly Tempo-Aligned)", fontsize=11, fontweight='bold')
ax2.set_xlabel("Normalized Duration (ms)")
ax2.set_ylabel("Probability Density (PDF)")
ax2.grid(axis='y', alpha=0.3)
ax2.legend(loc='upper right', fontsize=8, framealpha=0.85)
fig.tight_layout()
plot_path = os.path.join(plots_dir, f"{safe_phone}_normalization_comparison.png")
fig.savefig(plot_path)
plt.close(fig)
print(f"All 70 comparative plots generated successfully!")
print("Phase 2 complete.")
if __name__ == '__main__':
main()
|