Contributions of pitch, formant frequency and bandwidth to the perception of voice-personality

To investigate the contributions of the resonance characteristics of vocal tract and the pitch frequency to the voice-personality, a pitch synchronous analysis/synthesis system has been developed which is capable of independent manipulation of formant frequencies, bandwidths, and pitch frequencies. This paper gives the results of perceptual experiments on voice personality for spectrum and pitch modified speech using this system. Experimental results show that the perception of voice-personality is significantly sensitive to the formant frequency shift and it is almost completely lost for the uniform shift of all formant frequencies larger than 5 percent. The manipulation of pitch frequency and formant bandwidths, on the other hand, is less sensitive to the perception of voice-personality. This study is important as a basic research for the speaker normalization on one hand, and for adding personal information to the non-personal synthetic speech on the other.