METHOD, SYSTEM, AND PROGRAM FOR IDENTIFYING SPEAKER

摘要: PROBLEM TO BE SOLVED: To provide a method, system, and program for speaker identification which can identify the with good real-time response high precision. SOLUTION: The likelihoods L n of respective registered speakers are calculated by using feature parameters X k generated from voice to be identified 1st code book. When it is decided at stage that differences min q between minimum likelihood x other larger than threshold η, 2nd whether θmin 3rd θmax decide acceptance or rejection. On hand, when dicided there equal smaller decision made on basis an absolute value mean vector error e COPYRIGHT: (C)2004,JPO