Challenge Problem for Computational Auditory Scene Analysis: Understanding Three Simultaneous Speeches

Understanding three simultaneous speeches is proposed as a challenge problem to foster arti cial intelligence, speech and sound understanding or recognition, and computational auditory scene analysis research. Automatic speech recognition under noisy environments is attacked by speech enhancement techniques such as noise reduction and speaker adaptation. However, the signal-to-noise ratio of speech in two simultaneous speeches is too poor to apply these techniques. Therefore, novel techniques need to be developed. One candidate is to use speech stream segregation as a front-end of automatic speech recognition systems. Preliminary experiments on understanding two simultaneous speeches with or without an additional interfering sound show that the proposed challenge problem will be feasible with speech stream segregation. The detailed plan of the research on and benchmark sounds for the proposed challenge problem is also presented.