Open cocoakeith opened 6 years ago
In the git: https://github.com/tramphero/kaldi , the example of librispeech
Thanks for quick reply. The dimension of FBK features is 72 in paper, however, in fbank.conf the FBK features is 80. So, which one is better?
Another question, have you compare the TDNN-F[](Semi-Orthogonal Low-Rank Matrix Factorization for Deep Neural Networks) model with DFSMN? The low-rank weight matrix factorization and skip connection are used in both of these two models. These two models are very similar !
When to extract the fbank feature, the fbank.cfg is not in conf dirs, so how can I get it?