Function mel_spectrogram

cake-core/src/models/luxtts/mel.rs:13–49

(
    samples: &[f32],
    n_fft: usize,
    hop_length: usize,
    n_mels: usize,
    sample_rate: usize,
    device: &Device,
    dtype: DType,
)

/// Returns tensor of shape `[1, n_mels, time_frames]`.
#[allow(clippy::too_many_arguments)]
pub fn mel_spectrogram(
    samples: &[f32],
    n_fft: usize,
    hop_length: usize,
    n_mels: usize,
    sample_rate: usize,
    device: &Device,
    dtype: DType,
) -> Result<Tensor> {
    let stft = compute_stft(samples, n_fft, hop_length);
    let n_freq = n_fft / 2 + 1;
    let n_frames = stft.len() / n_freq;

    // Build mel filterbank
    let mel_filters = mel_filterbank(n_mels, n_freq, sample_rate, n_fft);

    // Apply mel filterbank: [n_mels, n_freq] @ [n_freq, n_frames] -> [n_mels, n_frames]
    let mut mel = vec![0.0f32; n_mels * n_frames];
    for m in 0..n_mels {
        for t in 0..n_frames {
            let mut sum = 0.0f32;
            for f in 0..n_freq {
                sum += mel_filters[m * n_freq + f] * stft[t * n_freq + f];
            }
            mel[m * n_frames + t] = sum;
        }
    }

    // Log mel (with floor to avoid log(0))
    for v in &mut mel {
        *v = (*v).max(1e-10).ln();
    }

    // Convert to tensor [1, n_mels, n_frames]
    let mel_tensor = Tensor::from_vec(mel, (1, n_mels, n_frames), device)?.to_dtype(dtype)?;
    Ok(mel_tensor)
}

/// Compute STFT magnitude squared, returned as flat Vec [n_frames * n_freq].
fn compute_stft(samples: &[f32], n_fft: usize, hop_length: usize) -> Vec<f32> {