Research And Implementation Of Emotion Monitoring System Based On Speech Emotion Recognition

With the rapid development of computer technology and artificial intelligence,the convenience brought by artificial intelligence products is seen everywhere in people's daily life.Meanwhile,people's demand for artificial intelligence products is also increasing day by day.Language is the most common way of communication in people's daily life.The acoustical manifestation of language is speech,and the speech signal contains a lot of emotional information.Therefore,analyzing emotion information in speech signal and applying it to artificial intelligence products is a hot research topic in speech emotion recognition.At first,the speech signals to be recognized are preprocessed,and then the effective emotional parameters were extracted.Five parameters of pitch frequency,short-time energy,the formant,linear predictive cepstrum coefficient(LPCC)and Mel prediction cepstrum coefficient(MFCC)from emotional speech as the input of the identification phase.In this thesis,a speech emotion recognition method based on HMM / RBF hybrid model is proposed.By using the powerful dynamic timing modeling ability of HMM model and the strong classification decision ability of RBF model,the speech emotion recognition rate can be further improved.In addition,in the learning process of RBF network,the concept of "dynamic optimal learning rate" is introduced.The learning rate of the traditional RBF neural network is pre-set with a fixed value and cannot be changed.Setting too large or too small will lead to network instability and affect the learning effect.The method proposed in this thesis to set the dynamic optimal learning rate can not only improve the convergence speed of the network,but also improve the operation efficiency.The MATLAB simulation results show that the recognition rate of HMM / RBF hybrid model is better than that of HMM and RBF alone,and it has faster convergence speed.Finally,this thesis designs and develops an emotion monitoring system based on speech emotion recognition,and applies the established emotion model to monitor the emotional state of the user and help the user adjust the emotion.