Speaker-Targeted Audio-Visual Models for Speech Recognition in Cocktail-Party Environments
This addresses the problem of overlapping speech for speech recognition systems, showing significant but incremental improvements in a simulated setting.
The paper tackled speech recognition in cocktail-party environments by proposing speaker-targeted acoustic and audio-visual models, achieving a word error rate reduction from 26.3% (audio-only baseline) to 3.6% with speaker identity information.
Speech recognition in cocktail-party environments remains a significant challenge for state-of-the-art speech recognition systems, as it is extremely difficult to extract an acoustic signal of an individual speaker from a background of overlapping speech with similar frequency and temporal characteristics. We propose the use of speaker-targeted acoustic and audio-visual models for this task. We complement the acoustic features in a hybrid DNN-HMM model with information of the target speaker's identity as well as visual features from the mouth region of the target speaker. Experimentation was performed using simulated cocktail-party data generated from the GRID audio-visual corpus by overlapping two speakers's speech on a single acoustic channel. Our audio-only baseline achieved a WER of 26.3%. The audio-visual model improved the WER to 4.4%. Introducing speaker identity information had an even more pronounced effect, improving the WER to 3.6%. Combining both approaches, however, did not significantly improve performance further. Our work demonstrates that speaker-targeted models can significantly improve the speech recognition in cocktail party environments.