Breaking News: Grepper is joining You.com. Read the official announcement!

audio feature extraction python

Innocent Iguana answered on February 5, 2024 Popularity 5/10 Helpfulness 7/10

answer audio feature extraction python

audio feature extraction python

Comment

import librosa as lr
from glob import glob

# Function to extract audio features from a single file
def extract_features(file_path):
    audio, sfreq = lr.load(file_path) # load an audio file into amplitude data array and number of sample per second
    time = np.arange(0, len(audio)) / sfreq # Create time array
    # Plot : Auditory envelope (smoothing signal) for visualization
    audio_rectified = audio.apply(np.abs)
    audio_envelope = audio_rectified.rolling(50).mean() 
    fig, ax = plt.subplots()
    ax.plot(time, audio) # time on X axis and amplitude on Y-axis
    ax.set(xlabel='Time (s)', ylabel='Sound Amplitude')
    plt.show()
    #### FEATURES ####
    mfccs = np.mean(lr.feature.mfcc(y=audio, sr=sfreq, n_mfcc=13), axis=1)
    chroma = np.mean(lr.feature.chroma_stft(y=audio, sr=sfreq), axis=1)
    spectral_contrast = np.mean(lr.feature.spectral_contrast(y=audio, sr=sfreq), axis=1)
    tonnetz = np.mean(lr.feature.tonnetz(y=audio, sr=sfreq), axis=1)
    tempogram = np.mean(lr.feature.tempogram(y=audio, sr=sfreq), axis=1)
    # Compute spectrogram (STFT) : THIS IS A TIME SERIES OF OSCILLATION
    spectrogram = lr.stft(y=audio)
    spec_db = np.mean(lr.amplitude_to_db(np.abs(spectrogram)), axis=1) # Convert to decibel for easier representation
    spectral_centroid = np.mean(lr.feature.spectral_centroid(y=audio, sr=sfreq))
    spectral_bandwidth = np.mean(lr.feature.spectral_bandwidth(y=audio, sr=sfreq))
    features = np.concatenate([mfccs, chroma, spectral_contrast, tonnetz, tempogram, spec_db,
                               [spectral_centroid, spectral_bandwidth]])
    return features

# Directory containing audio files
audio_files = glob('dataset/*.wav')

# Create feature matrix X
X = []
for file_path in audio_files:
    features = extract_features(file_path)
    X.append(features)

X = np.array(X)

Popularity 5/10 Helpfulness 7/10 Language python

Source: Grepper

Tags: audio feature-extraction python

Link to this answer
Share Copy Link

Contributed on Feb 05 2024

Innocent Iguana

0 Answers Avg Quality 2/10

audio feature extraction python

Contents

More Related Answers

audio feature extraction python

Grepper

Documentation

Social

Legal

Contact

Oops, You will need to install Grepper and log-in to perform this action.