Designation And Implementation Of Speech Emotion Recognition System Based On Android

Speech emotion recognition technology is a hot issue in the field of emotional computing and speech signal.As an important part of human-computer interaction,Speech emotion recognition has been widely used in disease diagnosis,criminal investigation,distance education and so on.As an important part of human-machine interaction,speech emotion recognition technology has become a problem because of the uncertainty of emotion itself and the fuzziness of characterization of emotion.In order to solve the problem that the speech emotion recognition rate is low and can not achieve human-computer interaction,the main contents of this paper is as follows:1.Introduction of the nonlinear characteristics of Teager energy operator,and Teager energy operator and Mel frequency cepstral coefficients(MFCC)are combined to extract NFD_Mel(Nonlinear Frequency Domain Mel).the experimental results show that the NFD_Mel combined with traditional features can effectively improve the recognition rate,the recognition rate in the German Berlin emotional database reached 80.02%,compared with the traditional method of recognition is increased by 3.24%.2.We put forward the innovative a independent speech emotion recognition method by cepstrum signal separation: the glottis and channel signal contains much emotional information,due to differences in individual channels,channel information usually includes personal characteristics,There is a disturbance to our work.The effect of emotional recognition based on independent speech emotion recognition is not as good as that of a particular person.Is that in order to overcome the shortcomings,we put forward the innovative a signal separation method based on cepstrum independent speech emotion recognition,the method of using cepstrum separation signal to retain full glottis information and discard the channel information,and find the best separation point,we finally extract the features by the reconstructed signa.In this paper We proposed a new feature that the CSS-MFCC(Cepstrum separate signal Mel Frequency Cepstral Coefficients)method to cepstrum separation signal and human auditory.and theexperimental results show the features combine with traditional feature can effectively improve the recognition rate in the German Berlin emotional database.The recognition rate is up to84.29%.3.The ultimate goal is to achieve human-computer interaction,so we built a speech emotion interaction framework based on Android system,the speech emotion recognition system could calculating and recognize speech emotion based on Android system.