摘要
语音情感识别作为一个新的研究热点,因其能解决教育中情感缺失的问题,而越来越受到研究者的重视.选取符合人类听觉系统感知的M el频率倒谱系数(MFCC)与各态历经型的连续隐马尔可夫模型(CHMM)进行语音情感特征的分析,并对大量的语音信号进行情感识别实验,识别正确率达到86.7%,为教育中的情感补偿提供了切实可行的依据.
As a new research hotspot,speech emotion recognition attracts more and attentions from researchers for its solution of the emotion loss in distance education. Based on the perception of the human auditory system, the Mel Frequency Cepstral Coefficients （MFCC） and the Continuous Hidden Markov Model （CHMM） techniques are used in this paper for speech characteristic analyses and emotion identification. The recognition accuracy of 87.6% is achieved in the experiment for large numbers of speech signals and the potential application for emotion compensation is provided in education.
出处
《南京师范大学学报（工程技术版）》
CAS
2009年第2期89-92,共4页
Journal of Nanjing Normal University(Engineering and Technology Edition)
基金
教育部留学回国人员科研启动基金(2008102SBJ0154)
南京师范大学留学回国人员基金(2008102XLH0070)
南京师范大学教务处精品实验资助项目
