Skip to main content

log_mel_spectrogram

Function log_mel_spectrogram 

Source
pub fn log_mel_spectrogram(
    pcm: &[f32],
    n_mels: usize,
    mel_filters: &[f32],
) -> Vec<f32>
Expand description

Compute log-mel spectrogram matching Python whisper.audio.log_mel_spectrogram.

  • pcm: audio samples (f32, 16kHz mono)
  • n_mels: 80 or 128
  • mel_filters: pre-loaded filter bank, shape [n_mels, N_FFT/2 + 1], row-major

Returns flat Vec in [n_mels, n_frames] layout (row-major per mel bin).