SPEAKER IDENTIFICATION OR VERIFICATION USING TEMPORAL MEASUREMENTS

摘要

A method for comparing utterances made by the speakers is disclosed. The method comprises obtaining a first sample of speech from the first speaker; providing a set of temporal interval measurements made from the first speech sample; and calculating the similarity of the temporal interval measurements from the first speech sample to that of a plurarity of instances of like measurements obtained from the speech of a set of speakers. The method is characterized in that at least some of the temporal intervals are expressed as proportions of a larger time interval.