音による感情理解

概要

論文の詳細を見る
The conventional sound recognition methods pursue the understanding of meaning of a sound by analytical approach. But a human can recognize the emotional feelings of a speaker no matter what language he or she may use. There seems to be two kinds of such expressions. One is truly emotional at heart and the other is to convey how he or she feels as messages. This work is aimed at understanding such messages of the emotional feelings of a speaker. Therefore, we paid more attention to the variation in time series. Our preliminary work demonstrated that anger can be extracted from human sound and characterized the sound time series by noting the pitch frequency and amplitude variation using cepstrum method.
1996-06-25