Research And Implementation For Technique Of Speech Emotion Recognition

With the rapid development of computer network communication and multi-media technology, the technology of new Human Machine Interaction(HCI) has become a very active study subject in the computer science field at present. The study on the speech emotion recognition has found important realistic values in such aspects as enhancing the intelligence and humanity of computer, developing new human-machine environments, promoting the study of psychology.In this paper, we firstly introduce the study background and other related key technologies of speech emotion recognition based on audio information, emphasizing on the knowledge of dealing with speech, analyzing and extracting speech emotion features, recognition methods. The study actuality and its trend in this field in the world at present are also emphasized.Secondly we introduce the proposed methods of analyzing, designing and recognizing the speech emotion. Then we discuss in details the process of analyses and main design ideas, which include mainly: the construction of emotion-speech templates database, the preprocess of speech signals, speech emotion features extraction based on hamming filter and wavelet transformation, speech emotion recognition based on templates matching, combining weighted Euclidean distance. During extracting the parameters of base frequency of speech, the algorithm of filtering noises and reconstructing speech signals based on Mallat algorithm is presented and used, combining the algorithm of detecting signal singular point based on wavelet transformation. During recognition of speech emotion base on audio frequency, we analyze and summarize according to examinations that the different extracted speech emotion features have different contribute in degree to every speech emotion status. Therefore, we present the contributes analyzing algorithm to give different weights to different extracted speech emotion features and then construct the templates. Then we can use the templates matching methods based on weighted Euclidean distance to achieve speech emotion recognition. The real-time command of the system can be met with, ensuring the high recognition rate meanwhile.In the end of this paper, we adopt the object oriented design methods to design the prototype system of speech emotion recognition. Then according to the experiments, the validity of above methods is proved.