Adaptive order of fractional Fourier transform for whispered speaker identification

A method widely used in speech signal analysis is based on short-time Fourier transform (STFT), but STFT only provides “average” characteristics of a signal, which can´t depict the refined structure of speech. Therefore, a new speech analysis tool called fractional Fourier transform (FRFT) is introduced into this article. The transform orders for FRFT are adaptively set according to piecewise linear fitting of instantaneous frequency (IF), which is based on AM-FM model theory of speech production. Then we present a kind of feature, namely, adaptive fractional Fourier transform cepstral coefficients (A-FRCC). The proposed speech parameters have been applied for whispered speaker identification. Experimental results show that in the test condition of mismatched channels, the new features can observe more sophisticated structure of speech and more personalized of speakers, at the same time, effectively improve the recognition rate and robustness, comparing with MFCC.