摘要
Based on analyzing contribution of short-time spectrum in different frequency subbands to speaker recognition and using of polynomial curve matching techniques, a non-linear frequency transform and feature detection algorithm are proposed to highlight the speaker＇s individuality in short-time spectrum of speech. The experimental results show that the performance of speaker recognition system is improved effectively, the average error rate of recognition relatively falls about 70.5%, 60.8% and 70.5% in comparison with classical frequency transform of Mel, Bark and ERB （Equivalent Rectangular Bandwidth） respectively.
Based on analyzing contribution of short-time spectrum in different frequency subbands to speaker recognition and using of polynomial curve matching techniques, a non-linear frequency transform and feature detection algorithm are proposed to highlight the speaker＇s individuality in short-time spectrum of speech. The experimental results show that the performance of speaker recognition system is improved effectively, the average error rate of recognition relatively falls about 70.5%, 60.8% and 70.5% in comparison with classical frequency transform of Mel, Bark and ERB （Equivalent Rectangular Bandwidth） respectively.
