Scaling the perceptual quality of pathological voice by the use of synthetic speech

Abstract

In our previous study on the computer analysis of pathological voice, it was shown that acoustical parameters such as the degree of fluctuation in the fundamental frequency, amplitude, and richness of harmonic components observed in sustained utterances of vowels were useful for differentiating among diseases affecting the voice [S. Hiki et al., J. Acoust. Soc. Am. 58, S111(A) (1975)]. It was also ascertained by sound spectrographic analysis that the amount of the noise component in the various frequency ranges and the characteristics of the rise and fall of utterances of vowels were of significant value in differentiating pathological voice from normal voice, and also in differentiating the voice resulting from cancer from that of polyp, nodule and recurrent nerve paralysis [S. Hiki et al., J. Acoust. Soc. Am. 63, S86(A) (1978)]. In this paper, the relationship among these acoustical parameters and the psychoacoustic factors obtained by a subjective discrimination test of pathological voices was investigated, and, based on the results, a scale for the perceptual judgment of the quality of pathological voice was made by use of synthetic speech. [Work supported by Ministry of Education, Japan.]