| As the main way of human daily life communication,speech signal carries the speaker's emotional information.The real sense of artificial intelligence requires machine to understand human intentions from the emotional level,so the speech emotion recognition has broad application prospect in the field of artificial intelligence in the future.At present,looking for a speech emotion feature with high degree of discrimination and constructing an efficient recognition model are the hot issues in the field of speech emotion recognition.And their quality affects the recognition effect of the whole system directly.Aiming at the situation of multiple emotion recognition,a speech emotion recognition method based on feature selection and decision tree SVM and speech emotion recognition method based on DNN-decision tree SVM are proposed from two aspects of feature parameters and models in this paper.The main research works of this paper are as follows:(1)Through the study of the background of speech emotion recognition and the current research situation,the basic framework of speech emotion recognition and the commonly used recognition methods are mastered.The paper introduces the commonly used corpus,speech preprocessing technology and commonly used feature parameters in the field of speech emotion recognition,and analyzes the characteristics of all kinds of parameters.In addition,the paper also introduces the method of calculating the statistical variables of the characteristic parameters and the normalization,which laid a solid foundation for the follow-up research work.(2)In order to solve the problem of overall recognition rate reduce due to the increase of emotional confusion for multiple speech emotion recognition.The speech emotion recognition algorithm of the decision tree SVM model with Fisher feature selection is proposed in this paper.In this algorithm,the decision tree SVM framework is firstly established by calculating the confusion degree of emotion,and then the features with higher distinguish ability are selected for each SVM of the decision tree according to Fisher criterion from the optional feature parameters.Finally speech emotion recognition is realized based on this model.The experimental results show that the average emotion recognition rate based on the proposed method is higher than traditional SVM classification method.(3)To deal with the problem that the traditional features can not mine the deeper emotion information of speech signals,this paper designs a depth neural network to extract the bottleneck feature of the speaker's speech,and establish the speech emotion recognition system based on DNN-SVM.Next,because the contribution of the same feature parameters to different emotion categories is different,different DNN networks are trained for different sentiment classifications to extract the bottleneck features which are used to train each SVM in the decision tree,and a speech emotion recognition system based on DNN-decision tree SVM is proposed in this paper.The experimental results show that the speech emotion recognition system based on DNN-decision tree SVM is more effective than that traditional SVM classification method under the same input parameters. |
