compute_mel_spectrogram

Compute mel spectrogram (S3Gen compatible)

Description

Compute mel spectrogram (S3Gen compatible)

Usage

 1compute_mel_spectrogram(
 2  y,
 3  n_fft = 1920,
 4  n_mels = 80,
 5  sr = 24000,
 6  hop_size = 480,
 7  win_size = 1920,
 8  fmin = 0,
 9  fmax = 8000,
10  center = FALSE
11)

Arguments

  • y: Audio samples as torch tensor or numeric vector
  • n_fft: FFT size (default 1920 for 24kHz)
  • n_mels: Number of mel bins (default 80)
  • sr: Sample rate (default 24000)
  • hop_size: Hop size (default 480)
  • win_size: Window size (default 1920)
  • fmin: Minimum frequency (default 0)
  • fmax: Maximum frequency (default 8000)
  • center: Whether to center frames (default FALSE)

Value

Mel spectrogram tensor (batch, n_mels, time)