An improved GA based modified dynamic neural network for Cantonese-digit speech recognition

Abstract

In this chapter, a dynamic neural network tuned by an improved GA (Lam et al., 2004) is proposed. New genetic operations (crossover and mutation) will be introduced. Rules have been introduced to the crossover process to make offspring widely spread along the domain. A fast convergence rate can be reached. A different process of mutation has been applied. This chapter is organized as follows. The genetic algorithm with improved genetic operations will be briefly described in section 2. The specific structure of the proposed dynamic neural network will be presented in section 3. In section 4, a Cantonese-digit speech recognition system will be discussed. The results for recognizing thirteen Cantonese digits and a conclusion will be given in section 5 and 6 respectively.