librosax.feature.mfcc

mfcc(*, y: Array | None = None, sr: float = 22050, S: Array | None = None, n_mfcc: int = 20, dct_type: int = 2, norm: str | None = 'ortho', lifter: int = 0, n_fft: int = 2048, hop_length: int = 512, win_length: int | None = None, window: str = 'hann', center: bool = True, pad_mode: str = 'constant', power: float = 2.0, n_mels: int = 128, fmin: float = 0.0, fmax: float | None = None, htk: bool = False, melspectrogram_params: dict | None = None) → Array[source]

Compute Mel-frequency cepstral coefficients (MFCCs).

MFCCs are computed from the log-power mel spectrogram.

Note

For JAX JIT compilation, all arguments except y and S should be marked as static. This includes all the melspectrogram parameters and MFCC-specific parameters: sr, n_mfcc, dct_type, norm, lifter, plus all other kwargs.

Parameters:

  • y – Audio time series. Multichannel is supported.

  • sr – Audio sampling rate

  • S – (optional) log-power mel spectrogram

  • n_mfcc – Number of MFCCs to return (default: 20)

  • dct_type – Discrete cosine transform (DCT) type (default: 2)

  • norm – If “ortho”, use orthonormal DCT basis. Default: “ortho”

  • lifter – If lifter>0, apply liftering (cepstral filtering) to the MFCCs. If lifter=0, no liftering is applied.

  • n_fft – FFT window size (used if y is provided)

  • hop_length – Hop length for STFT (used if y is provided)

  • win_length – Window length (used if y is provided)

  • window – Window function (used if y is provided)

  • center – If True, pad the signal (used if y is provided)

  • pad_mode – Padding mode (used if y is provided)

  • power – Exponent for the magnitude melspectrogram (used if y is provided)

  • n_mels – Number of mel bands (used if y is provided)

  • fmin – Lowest frequency in Hz (used if y is provided)

  • fmax – Highest frequency in Hz (used if y is provided)

  • htk – Use HTK formula for mel scale (used if y is provided)

  • melspectrogram_params – Additional keyword arguments for melspectrogram (used if y is provided)

Returns:

MFCC sequence [shape=(…, n_mfcc, t)]

Return type:

jnp.ndarray