Shift Window Based Framework for Emotional Change Detection of Speech

Compared with single emotion recognition, the emotional change detection of speech has greater practical values. However, there are only few researches on this subject at present. This paper presented a basic framework for emotional change detection. Firstly, the smoothing strategy was adopted and the effect of smoothing width was also investigated to improve the correct detection rate. Then four different smoothing functions were compared and that utilizing the mean value smoothing was considered as the elementary detection method. Finally, how window width and shift step affect the correct detection rate was analyzed and time precision of elementary method was evaluated. Experimental results demonstrated that the proposed framework is feasible and effective.