基于chmm的语音情感识别的研究-教育学；教育技术学专业论文.docxVIP

摘要 摘要 I I 摘 要 在教育教学过程中，人们往往只注重言语行为，而非言语行为常常作为人类 交流中言语行为的辅助。其实，在教学和交际过程中所利用的方法和手段除了言 语行为外，还包含有丰富的非言语行为。在教学与交流过程中恰当地使用非言语 行为，将能得到事半功倍的效果。同样语音情感作为一种包含在声音中的非言语 行为，能够在辅助知识的传授和传递教学效果的反馈方面起到重要的作用，也引 起了越来越多研究者的重视。 本文从非言语行为中的语音情感出发，对语音情感在教育中的应用点进行了 探究。对语音的产生和接收的机理进行了分析，根据分析将语音的情感特征区分 为基于发音模型的情感特征和基于语音接收模型的情感特征两类。 通过对国内外相关研究的了解并结合教育教学过程中反馈所需的情感，本文 将情感分为生气、厌烦、悲伤和高兴四类。对计算机情感识别进行了实验性的研 究。语音情感的识别也可以看成一个模式识别的问题。在特征选择方面，对两类 情感特征进行了分析与比较，选择 MFCC 作为特征，针对 MFCC 提取过程中的不 足，改用小波来提取情感语音的 MFCC。在识别方法方面，选取与人类发音过程 相吻合的 CHMM 来进行语音情感的识别。取得了较好的实验效果。 关键词：语音情感 MFCC CHMM Abstract Abstract II II Abstract In the process of education, people tend to focus on verbal communication, which is always considered to be supported by nonverbal communication. In fact, there are verbal communication and rich nonverbal communication in the process of education and communication. Using nonverbal communication properly in the process of communication and education, we will get twice the result with half the effort. As a kind of nonverbal communication included in speech, speech emotion play an important role in assisting the transmission of knowledge and transferring the feedback in teaching effectiveness. Therefore, it has attracted more and more researchers’ attentions. We research into application points of speech emotion in the education as a specific nonverbal communication, and divide speech emotion feature into two types according to the analyzing of the mechanism of sounding and listening. The types are feature based on sounding model and feature based on listening model. Emotions are classified into four categories in this paper according to some known studies and need of feedback in teaching effectiveness. Automatic recognition for these four categories of speech emotion are carried out experimentally, and automatic emotion recognition of human speech can be viewed as a pattern recognition problem. MFCC is chosen as the feature based on the characteristics of the four types of speech emotions. Because of t