Emotional corpus based on MFCC and the coefficient of correlation

Emotive expression plays a key role in the staff of call center of the business. The customer service of call center through telephone service to address customer questions on company products or services. In the communication process, if staffers of call center can understand customer signs of emotional changes, that can prevent disputes in advance. In this research, we discuss on this article trying to use correlation coefficient, emotion of the time before and after changes in the characteristics of individual statements. We focus on analyzing emotive utterances for anger and non-anger in corpus D80. We use MFCC to extract the coefficients in D80. Then, we used coefficient of correlation to analyze MFCCs to distinguish information of emotional utterances in D80. When the std of a coefficient of correlation is greater than 0.07, the angry utterances have 83% in D80. There is 17% margin of error in our analysis. When the standard deviation of the coefficient of correlation is greater than 0.08, the angry utterances have 71% in D80. However, happiness utterances decrease from 55% to 29%. So when the coefficient of correlation is set to 0.07, it's useful to detect the utterances for anger. We recommend this method (the coefficient of correlation) to be an emotive analysis parameters, for judge angry with non-angry emotion changes. Thus, it can be implemented as a strategy to improve the quality of service in automated call center applications.