Acoustics Feature Based Chineses Speech Mood Recognition System

Speech is the most basic, efficient and important way of communicating among human beings. With the continuous development of computer science and technology, making full use of speech signals has become a research field of significant importance and value. A considerable number research and study in speech recognition, speaker recognition and speech synthesis has taken place for decades all over the world. However, in comparison, few study results or theory in the field of speech mood recognition has been put forward.Mood information is a very important recourse. Be different from speech recognition which aims at recognizing speech content, speech mood recognition emphasizes in the way people speak, and excavate the tone and attitude hidden in literal meanings. In fact, same words can convey very different meanings by different intonation, amplitude or stress. Traditional speech recognition, however, discards such information as personal difference, and therefore lose very valuable information.This paper mainly introduces and studies the design and implementation of acoustic feature based speech mood recognition system. Its main contents involve:(?) The background and previous works of speech mood recognition.(?) Brief introduction of human vocal system and Chinese phonetics.(?) The extraction algorithms as well as selection of Acoustic features, feature dimension deductions.(?) Discuss and compare of mood classifiers, support vector machines. (?) Design, implementation and test of mood recognition system on call center.This speech mood recognition system designed and implemented in this paper, which has a relatively high performance and accuracy, provide some reference value for the future works.