| Lecturer Profile |
Professor Chin-Hui Lee used to work at The Bell Labs. Currently, he works as a full-time professor at Georgia Institute of Technology. Till now, he has published over 400 papers and owned 30 patents. Prof. Lee used to work as an associate editor for IEEE Transaction on Signal Processing and IEEE Transactions on Speech and Audio Processing. Now, he is a fellow of IEEE and ISCA. |
| Lecture Abstract |
We cast the classical speech deriverberation problem into a regression setting by mapping log power spectral features of reverberant speech to time-delayed features of anechoic speech. Depending on the reverberation time of the acoustic environment we found that different signal processing parameters are needed to deliver a good quality for deriver berated speech. Furthermore, reverberation-time-aware DNN training and decoding procedures can be designed to optimize the deriverberation performance across a wide range of reverberation times. In addition, a single DNN can also be trained to perform simultaneous beam forming and deriverberation for microphone array speech. Furthermore, as a side benefit, using DNN-based speech deriverberation as a pre-processor in the REVERB Challenge automatic speech recognition (ASR) task, we get the lowest word error rate without retraining the deriverberation front-end and the ASR back-end. It is expected the ASR accuracy and robustness could still be improved with joint training of an integrated deriverberation-ASR system. |
