asteroid.models.dccrnet module

class asteroid.models.dccrnet.DCCRNet(*args, stft_n_filters=512, stft_kernel_size=400, stft_stride=100, **masknet_kwargs)

Bases: asteroid.models.dcunet.BaseDCUNet

DCCRNet as proposed in [1].

Parameters:

• architecture (str) – The architecture to use, must be “DCCRN-CL”.
• stft_kernel_size (int) – STFT frame length to use
• stft_stride (int, optional) – STFT hop length to use.
• sample_rate (float) – Sampling rate of the model.
• masknet_kwargs (optional) – Passed to DCCRMaskNet

References

• [1] : “DCCRN: Deep Complex Convolution Recurrent Network for Phase-Aware Speech Enhancement”, Yanxin Hu et al. https://arxiv.org/abs/2008.00264

masknet_class

alias of asteroid.masknn.recurrent.DCCRMaskNet

forward_encoder(wav)

Computes time-frequency representation of wav.

Parameters:wav (torch.Tensor) – waveform tensor in 3D shape, time last. Returns:torch.Tensor, of shape (batch, feat, seq).

apply_masks(tf_rep, est_masks)

Applies masks to time-frequency representation.

Parameters:

• tf_rep (torch.Tensor) – Time-frequency representation in (batch, feat, seq) shape.
• est_masks (torch.Tensor) – Estimated masks.

Returns:

torch.Tensor – Masked time-frequency representations.