Study on Method of Emotion Recognition of Speech based on Simulated Annealing Genetic Algorithm and Support Vector Machine

Abstract

With increasing interest in human-computer interaction, emotion recognition is becoming a hot spot. Support vector machine(SVM) is a machine learning algorithm based on statistic learning theory, finding the optimal separating hyperplane in the high dimensional feature space, having good classification and generalization ability. According to SVM method becoming very difficult for large data scale, this paper proposes a feature selection method based on simulated annealing genetic algorithm. This method let simulated annealing genetic algorithm embed into the loop body of the adaptive genetic algorithm, Using simulated annealing algorithm with a strong local search ability, and it can make searching process avoid falling into the local optimal solution, solving the problem of basic genetic algorithm having slow convergence and high time complexity. The experiments showed that this method can significantly improve speech emotion recognition performance.