A speech recognition system based on structure equivalent fuzzy neural network trained by firefly algorithm

Abstract

Speech recognition technology is a technology that allows a computer to recognize the speech and the words that express through the microphone or speaks by phone. A Fuzzy Neural Network (FNN) is a learning approach that finds the parameters of a fuzzy system by exploiting approximation techniques from neural network. But FNN has some difficulty about how to automatically generate and adapt the membership function and fuzzy rules. To overcome the shortage of FNN, in this paper, we use SEFNN (Structure Equivalent Fuzzy Neural Network) and optimized its parameters with Firefly algorithm. Firefly Algorithm (FA), which is usually used in optimization problems is a stochastic population-based algorithm inspired by intelligent collective behavior of fireflies in the nature. The parameters of SEFNN trained by FA were used in speech recognition system to improve the ability of generalization of FNN. Results shows that the SEFNN optimized by FA for speech recognition system have higher recognition rate in compare of FNN trained by PSO method.