Research On Adaptive Speech Emotion Recogniton Method

In recent years,the reasons for renewed interest in speech emotion recognition are multiple,but mainly due to people have more interest about human computer interaction(HCI).Speech emotion recognition is to analyze and detect the special emotion state from given emotion speech emotion utterance and then to ascertain the subject's specific inborn emotion,achieving smarter and more natural interaction between human beings and computers.The study of speech emotion recognition has found important applied values.In this work,we firstly discuss the background and then analyze the main exiting speech emotion feature extraction algorithms,feature selection algorithms and emotion recognition algorithms.After analyzing the methods currently used by others, we firstly use some derivative parameters as features for recognition,then present a feature selection method based on Genetic Algorithm(GA).And we improve the algorithm and present an adaptive and optimal classification of speech emotion recognition.The works are described as bellows:(1) Derivative feature parameters extraction.Scholars agreed the importance that "change" impacts the emotion,so a direct description of such changes,in theory,is closer to the essence of the emotion.Mathematical description of change is derivative. Taking into account the discrete nature and the short-term stability of the voice processing,we change such discrete values into continuous function,using curve fitting methods,fitting function using cubic spline curve.From the fitting curve we obtaine derivation prameters,based on their derivative values and its derivatives we will get new characteristic parameters.(2) SVM recognition algorithm based on genetic algorithm.Genetic algorithm is a global search optimization algorithm,and the emotional voice of the process of feature selection is to search the optimal solution process,this paper,binary encoding method,the evaluation function to select,using the SVM one-on-one and one-to-many recognition rate.High recognition rate guide genetic algorithms to the evolution direction of improving recognition rate,until you find the optimal feature subset.And analyze the impact of recognition rate of the one-on-one and one-to-many SVM assess function,but also an analysis of the relationship between the population number of genetic algorithms and recognition rate is present.(3)Taking into account the characteristics of voice emotional personality characteristics,a kind of emotion has its own feature subset.According to this nature, present the design of the adaptive genetic algorithm selection method,its main idea is to use the largest recognition rate of each emotion set to distinguish between other feeling,as the assess function.This method has been an emotional choice based on the binary decision tree.In the binary tree on the basis of assessment has improved the function,adding the second genetic algorithm,so that it can be divided into different combinations of emotion,that is,multi-branch decision tree generated.(4) By combining matlab with VC++,a prototype system of speech emotion recognition based on selective features and SVM decision tree is achieved,which demonstrate the effectiveness of the algorithms mentioned above.