System for judgment of depressive tendency from speech analysis

The objective of this investigation was to construct a system for judgment of depressive tendency. A sentence reading experiment was performed with 15 subjects with a highly depressive tendency and 15 subjects with a mildly depressive tendency selected using the BDI, a measure of depressive tendency. The spoken sentences were fed to a computer, and the amplitude, fundamental frequency, and utterance speed were calculated. Statistical tests were applied to each physical feature parameter of the speech, and the relation to the degree of depressive tendency of the subject in the sentence reading was determined. The objective was to identify index words that well express the depressive tendency. For three index word selection strategies, the important physical parameters of speech were extracted by logistic multiple regression analysis. The experiment and analysis indicated that the statistical index word selection strategy was more effective in judgment than the other two index word selection strategies.