Alex-Songs

Results 26 comments of Alex-Songs

Add Non-Autoregressive ASR recipes for AISHELL-1

Hi Wecan-Huang0602: It's a great work. But I have a question: your code uses a known text length in the test set to construct the required positional encoding for the...

The ''decode_streaming'' in speech2text.py did not support the streamming transformer?

me,too. I hope the author will complete it.

Hello, does Transformer have not implemented beam_search_chunk_sync yet? So how does MMA test?

Can you provide some sample scripts for MMA decoding? Thanks! @hirofumi0810

Hello, does Transformer have not implemented beam_search_chunk_sync yet? So how does MMA test?

Thanks

Development plan for ESPnet2 speech enhancement

Hello, where can I find the recipe corresponding to enh_asr.py? Thank you! @LiChenda

Development plan for ESPnet2 speech enhancement

Thanks! @shincling

Development plan for ESPnet2 speech enhancement

Hello, is there any related paper introduction for "STFT consistency"? Thank you! @shincling

Development plan for ESPnet2 speech enhancement

Thanks! @Emrys365

Development plan for ESPnet2 speech enhancement

Hello, when I use tf-mask_net and ASR for joint training (2 speaker), the CTC loss function is difficult to converge. Have you encountered a similar situation? Thanks!

Development plan for ESPnet2 speech enhancement

Hello shincling, I used the default configuration of wsj0_2mix, but the training set was replaced with other simulated data sets. I found that TasNet works well, but tf-mask-net is difficult...