Research Of Fuzzy Case Based Reasoning Emotion Recognition Based On Speech

One intelligence computer should notices and understands the users’emotion; the intelligence computer should also change the human-machine interface environment if the users’emotions change. How to make the computer exactly recognize the emotion of human has significant meanings not only in theory but also in practical. The research of human emotion recognition mainly contain:Facial Expression Recognition, speech emotion recognition, physiological signals emotion recognition.As one part of human affective computing, Speech emotion recognition has attracted lot of researchers’interest in this field, suppose computer has human speech information, after detail with the speech that contain human emotion information, the computer should automatically recognize speaker’s emotion that is important part of intelligence human-machine interaction.The research of speech emotion recognition generally considered several parts:speech signal Emotional characteristics analysis, extraction speech signal Emotional characteristics method, build speech emotion recognition model, speech emotion recognition algorithm. The main researches of this paper are as follows:(1) In extraction speech signal Emotional characteristics aspect, how to extract emotional characteristics play irreplaceable role in speech emotion recognition. We find that the Time-domain characteristics can fully express the difference oh human emotion, so we propose use the Time-domain characteristics as the main speech signal Emotional characteristics. The Time-domain Characteristic parameters include:Time structure, an amplitude structure, the formant structure.(2) In speech emotion recognition model aspect, we propose use fuzzy rules to replace the speech emotion recognition model for fuzzy rules can do a good performance in a system when the system model is not clear. We propose use modified WANG algorithm to extract fuzzy rule based on speech Time-domain Characteristic and then use the extracted fuzzy rule to classify speech emotion.(3) In speech emotion recognition algorithm aspect, after study of Case Based Reasoning(CBR) theory and WANG algorithm, we combine two of them together, that is Fuzzy Case Based Reasoning. This method not only keeps the advantage of fuzzy rules but also uses the great performance of Case Based Reasoning.(4) In this paper we use Berlin speech emotion data base as the training and test data base. the experiment result can show the method we propose achieve good performances.