Documents
Poster
Spectral feature mapping with mimic loss for robust speech recognition
![](https://sigport.org/sites/default/files/styles/home/public/ICASSP_2018_docs.jpg?itok=gxiDHw0M)
- Citation Author(s):
- Submitted by:
- Deblin Bagchi
- Last updated:
- 16 April 2018 - 3:17am
- Document Type:
- Poster
- Document Year:
- 2018
- Event:
- Presenters:
- Deblin Bagchi
- Paper Code:
- 3472
- Categories:
- Log in to post comments
For the task of speech enhancement, local learning objectives are agnostic to phonetic structures helpful for speech recognition. We propose to add a global criterion to ensure de-noised speech is useful for downstream tasks like ASR. We first train a spectral classifier on clean speech to predict senone labels. Then, the spectral classifier is joined with our speech enhancer as a noisy speech recognizer. This model is taught to imitate the output of the spectral classifier alone on clean speech. This \textit{mimic loss} is combined with the traditional local criterion to train the speech enhancer to produce de-noised speech. Feeding the de-noised speech to an off-the-shelf Kaldi training recipe for the CHiME-2 corpus shows significant improvements in WER.