| import librosa |
| import numpy as np |
| from tensorflow.keras.models import load_model |
|
|
| |
| model = load_model('/kaggle/working/Dubai_Audio.h5') |
| speaker_to_int = { |
| 'Brene Brown': 0, |
| 'Eckhart Tolle': 1, |
| 'Eric Thomas': 2, |
| 'Gary Vee': 3, |
| 'Jay Shetty': 4, |
| 'Les Brown': 5, |
| 'Mel Robbins': 6, |
| 'Nick Vujicic': 7, |
| 'Oprah Winfrey': 8, |
| 'Rabin Sharma': 9, |
| 'Simon Sinek': 10 |
| } |
|
|
| def preprocess_audio(file_path, n_mfcc=40, max_pad_len=216): |
| audio, sample_rate = librosa.load(file_path, sr=None) |
| mfcc = librosa.feature.mfcc(y=audio, sr=sample_rate, n_mfcc=n_mfcc) |
| if mfcc.shape[1] < max_pad_len: |
| pad_width = max_pad_len - mfcc.shape[1] |
| mfcc = np.pad(mfcc, pad_width=((0,0), (0,pad_width)), mode='constant') |
| else: |
| mfcc = mfcc[:, :max_pad_len] |
| mfcc = mfcc[np.newaxis, ..., np.newaxis] |
| return mfcc |
|
|
| def predict_audio_class(file_path, model, int_to_speaker): |
| processed_audio = preprocess_audio(file_path) |
| predictions = model.predict(processed_audio) |
| predicted_index = np.argmax(predictions, axis=1)[0] |
| print(predicted_index) |
| predicted_class = int_to_speaker[predicted_index] |
| return predicted_class |
|
|
| |
| int_to_speaker = {v: k for k, v in speaker_to_int.items()} |
|
|
| |
| uploaded_audio_path = '/kaggle/input/audio-classifier-dataset/augmented-audio/Mel Robbins/113Mel Robbins113.wav' |
|
|
| predicted_class = predict_audio_class(uploaded_audio_path, model, int_to_speaker) |
| print(f'Predicted class: {predicted_class}') |
|
|
