|
|
|
sample_rate: 16000 |
|
win_length: 32 |
|
hop_length: 16 |
|
n_fft: 512 |
|
window_fn: !name:torch.hamming_window |
|
|
|
compute_stft: !new:speechbrain.processing.features.STFT |
|
sample_rate: !ref <sample_rate> |
|
n_fft: !ref <n_fft> |
|
win_length: !ref <win_length> |
|
hop_length: !ref <hop_length> |
|
window_fn: !ref <window_fn> |
|
|
|
compute_istft: !new:speechbrain.processing.features.ISTFT |
|
sample_rate: !ref <sample_rate> |
|
n_fft: !ref <n_fft> |
|
win_length: !ref <win_length> |
|
hop_length: !ref <hop_length> |
|
window_fn: !ref <window_fn> |
|
|
|
spectral_magnitude: !name:speechbrain.processing.features.spectral_magnitude |
|
power: 0.5 |
|
|
|
resynth: !name:speechbrain.processing.signal_processing.resynthesize |
|
stft: !ref <compute_stft> |
|
istft: !ref <compute_istft> |
|
|
|
enhance_model: !new:speechbrain.lobes.models.MetricGAN.EnhancementGenerator |
|
input_size: !ref <n_fft> // 2 + 1 |
|
hidden_size: 200 |
|
num_layers: 2 |
|
dropout: 0 |
|
|
|
modules: |
|
enhance_model: !ref <enhance_model> |
|
|
|
pretrainer: !new:speechbrain.utils.parameter_transfer.Pretrainer |
|
loadables: |
|
enhance_model: !ref <enhance_model> |
|
|
