System for recognizing utterances alternately spoken by plural speakers with an improved recognition accuracy

摘要

A system for recognizing a speech spoken by one of plurality of speakers. The system is provided with a parameter table. Each record of the parameter table containing acoustic features and at least one parameter associated with each of prospective speakers. A set of speech signals is obtained from the speech. If the speaker is included in the prospective speakers, the speech is recognized by using the parameter(s) associated with the speaker. If the speaker is not included in the prospective speakers, a record for the speaker is added to the parameter table by using acoustic features and at least one parameters obtained through an analysis of the set of speech signals. The parameter associated with the speaker is updated by using at least one parameters obtained through an analysis of the set of speech signals.