asr

asr

1. Please prepare the wav.scp, text, utt2dur for training and development sets under data/aishell_v4/train/ and data/aishell_v4/dev/ respectively following the kaldi format.

2. Please prepare the wav_nospk_nofe.scp, wav_spk_nofe.scp, wav_nospk_fe.scp and wav_spk_fe.scp for testing data of Speaker Independent and Speaker Dependent tasks under data/aishell_v4/test/ respectively following the kaldi format.

3. Train the asr model

./run.sh

4. Generate the asr results

./decode.sh data/test/wav_nospk_nofe.scp wav_nospk_nofe
./decode.sh data/test/wav_spk_nofe.scp wav_spk_nofe
./decode.sh data/test/wav_nospk_fe.scp wav_nospk_fe
./decode.sh data/test/wav_spk_fe.scp wav_spk_fe

You can also use our pretrained model to generate the asr results directly. Please download the pretrain the model from download path and save at exp/aishell_v4/asr/.