Abstract:
No single acoustic parameter is believed to characterize solely voice individuality. A perceptual experiment has been performed concerning the voice individuality of human speech sounds making use of an analysis--synthesis method which is capable of independent manipulation of formant frequencies, bandwidths, and pitch frequencies. Pitch and formant frequencies are the two most important acoustic parameters for speech science and technology. This experiment has been performed to examine which parameter carries more information or is more robust, in terms of voice individuality, against acoustic distortion. Experimental results reveal that the perception of voice individuality is far more sensitive to the formant frequencies than the pitch frequency. Only a few percent shift of all formant frequencies, either toward the high-frequency region or toward low frequency, from the original values has been found to lose the voice individuality almost 100% while pitch frequency does not. Pitch frequency has been found to be far more robust than the formants. A few percent shift of pitch frequency has no damage on the individuality. To lose the individuality it is necessary to change the pitch frequency by as much as 50% or more toward either the high- or low-frequency regions. [Work supported by The Sound Technology Promotion Foundation.]