Search

This paper describes a computer model that attempts to separate two simultaneous talkers. Using such a model, a speech recognition system can improve its ability to recognize separate utterances made simultaneously by two different talkers. In order to achieve this goal, the computer model’s algorithms were designed to functionally simulate the major steps that are performed by the human auditory system when processing multiple sound sources monaurally. It is based on my belief that the human auditory system computes similar representations as the computer model, even if the physiological processes and representations that the auditory system uses differ slightly from those presented here.