さ声の音響分析 デジタルサウンドスペクトログラフによる試み

Abstract

With the use of a digital sound spectrograph (Digital Sonagraph Model 7800, Kay Elemetric Corp.), clinically useful methods of objective evaluation of hoarse voice were developed. Two parameters, the extent of pitch fluctuation and the amount of noise, were measured in the digital sound spectrogram by the author's simplified methods.The extent of the pitch fluctuation of sustained phonation of /a/ (sΔF) was measured on the highest harmonic stripe of the sound spectrogram (0-2kHz range and 11.3Hz filter width). There was a significant correlation between sdF and the maximum pitch fluctuation obtained by the minicomputer system (Pearson r=0.82). The sΔF differentiated 70.4% of the pathologic voice samples from the normal ones.The amount of noise was measured from the section display with a 0-2kHz range, a 22.5Hz filter for male voices and a 37.5Hz filter for female voices. The author defined the area enclosed by the base-line and the line connecting the peaks of the harmonic stripes as the total acoustic energy, and the area enclosed by the base-line and the line connecting the troughs as the acoustic energy of the noise components. The amount of noise in a phonatory sample was expressed by the difference between the former area and the latter area. The values of the amount of noise in the pathologic samples correlated significantly with the psychophysical judgments of the degree of hoarseness (Spearman's r=0.77 in the male samples, 0.74 in the female samples). Moreover, 79.6% of the pathologic samples were differentiated from the normal ones by the two parameters. These results showed that the two parameters obtained from the present methods were reliable and clinically useful indices.