Caution

You're reading an old version of this documentation. If you want up-to-date information, please have a look at 0.9.1.

Feature extraction

Spectral features

chroma_stft([y, sr, S, norm, n_fft, ...])

Compute a chromagram from a waveform or power spectrogram.

chroma_cqt([y, sr, C, hop_length, fmin, ...])

Constant-Q chromagram

chroma_cens([y, sr, C, hop_length, fmin, ...])

Computes the chroma variant "Chroma Energy Normalized" (CENS)

melspectrogram([y, sr, S, n_fft, ...])

Compute a mel-scaled spectrogram.

mfcc([y, sr, S, n_mfcc, dct_type, norm, lifter])

Mel-frequency cepstral coefficients (MFCCs)

rms([y, S, frame_length, hop_length, ...])

Compute root-mean-square (RMS) value for each frame, either from the audio samples y or from a spectrogram S.

spectral_centroid([y, sr, S, n_fft, ...])

Compute the spectral centroid.

spectral_bandwidth([y, sr, S, n_fft, ...])

Compute p'th-order spectral bandwidth.

spectral_contrast([y, sr, S, n_fft, ...])

Compute spectral contrast

spectral_flatness([y, S, n_fft, hop_length, ...])

Compute spectral flatness

spectral_rolloff([y, sr, S, n_fft, ...])

Compute roll-off frequency.

poly_features([y, sr, S, n_fft, hop_length, ...])

Get coefficients of fitting an nth-order polynomial to the columns of a spectrogram.

tonnetz([y, sr, chroma])

Computes the tonal centroid features (tonnetz)

zero_crossing_rate(y[, frame_length, ...])

Compute the zero-crossing rate of an audio time series.

Rhythm features

tempogram([y, sr, onset_envelope, ...])

Compute the tempogram: local autocorrelation of the onset strength envelope.

fourier_tempogram([y, sr, onset_envelope, ...])

Compute the Fourier tempogram: the short-time Fourier transform of the onset strength envelope.

Feature manipulation

delta(data[, width, order, axis, mode])

Compute delta features: local estimate of the derivative of the input data along the selected axis.

stack_memory(data[, n_steps, delay])

Short-term history embedding: vertically concatenate a data vector or matrix with delayed copies of itself.

Feature inversion

inverse.mel_to_stft(M[, sr, n_fft, power])

Approximate STFT magnitude from a Mel power spectrogram.

inverse.mel_to_audio(M[, sr, n_fft, ...])

Invert a mel power spectrogram to audio using Griffin-Lim.

inverse.mfcc_to_mel(mfcc[, n_mels, ...])

Invert Mel-frequency cepstral coefficients to approximate a Mel power spectrogram.

inverse.mfcc_to_audio(mfcc[, n_mels, ...])

Convert Mel-frequency cepstral coefficients to a time-domain audio signal