Speech Emotion Recognition Based on Learning Automata in

This paper explores how fuzzy features’ number and reasoning rules can influence the rate of emotional speech recognition. The speech emotion signal is one of the most effective and neutral methods in individuals’ relationships that facilitate communication between man and machine. This paper introduces a novel method based on mind inference and recognition of speech emotion recognition. The foundation of the proposed method is the inference of rules in Fuzzy Petri-net (FPN) and the learning automata. FPN is a new method of classification which is introduced for the first time on emotion speech recognition. This method helps to analyze different rules in a dynamic environment like human’s mind. The input of FPN is computed by learning automata. Therefore learning automata has been used to adjust the membership functions for each feature vector in the dynamic environment. The proposed algorithm is divided into different parts: preprocessing; feature extraction; learning automata; fuzzification; inference engine and defuzzification. The proposed model has been compared with different models of classification. Experimental results show that the proposed algorithm outperforms other models.