Robust automatic speech recognition with missing and unreliable acoustic data
Speech Communication
Missing Data Techniques for Robust Speech Recognition
ICASSP '97 Proceedings of the 1997 IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP '97)-Volume 2 - Volume 2
Hi-index | 0.00 |
This paper summarizes our work carried out over the last four years on multi-band and multi-stream processing as means to achieve robust automatic speech recognition (ASR). Main focused is laid on the "full combination" approach which integrates over all possible positions of reliable data, instead of combining the nominal subbands or streams only. The main combination rules are summarized and the final results from a wide range of experiments are presented.