This code implements a basic MLP for speech recognition. The MLP is trained with pytorch, while feature extraction, alignments, and decoding are performed with Kaldi. The current implementation supports dropout and batch normalization. An example for phoneme recognition using the standard TIMIT dataset is provided.
Hello Mirco,
When I am trying to run the create_chunks.sh, it gives me such of this error
(base) [mnabih@piccadilly pytorch_MLP_for_ASR-master]$ ./create_chunks.sh /home/kaldi/egs/timit/s5/data/train splits_fea 5 train 1
Computing sentence lenghts...
./create_chunks.sh: line 60: /home/kaldi/egs/timit/s5/data/train/file_len.scp: No such file or directory
cat: /home/kaldi/egs/timit/s5/data/train/wav.scp: No such file or directory
./create_chunks.sh: line 42: /home/kaldi/egs/timit/s5/data/train/feats_ord.scp: No such file or directory
paste: /home/kaldi/egs/timit/s5/data/train/file_len.scp: No such file or directory
./create_chunks.sh: line 44: /home/kaldi/egs/timit/s5/data/train/feats_ord.scp: No such file or directory
split: 0: invalid number of lines
Try 'split --help' for more information.
features in splits_fea have been ordered and split into 5 chunks.
./create_chunks.sh: line 64: compute-cmvn-stats: command not found
./create_chunks.sh: line 65: compute-cmvn-stats: command not found
done cmnvs.
(base) [mnabih@piccadilly pytorch_MLP_for_ASR-master]$