EMOTION IN SPEECH: RECOGNITION AND APPLICATION TO CALL CENTERS 论文

摘要

The paper describes two experimental studies on vocal emotion expression and recognition. The first study deals with a corpus of 700 short ut terances expressing five emotions: happiness, anger, sadness, fear, and normal (unemotional) state, which were portrayed by thirty non-professional actors. After evaluation a part of this corpus was used for extracting features and training backpropagation neural network models. Some statistics of the pitch, the first and second formants, energy and the speaking rate were selected as relevant features using feature selection techniques. Several neural network recognizers and ensembles of recognizers were created. The recognizers have demonstrated the following accuracy: normal state - 60-75%, happiness -- 6070 %, anger -- 70-80%, sadness -- 70-85%, and fear -- 35-55%. The total average accuracy is about 70%. The second study uses a corpus of 56 telephone messages of varying length (from 15 to 90 seconds) expressing mostly normal and ...