Multiple Evolutionary Neural Networks In Speech Recognition

As the traditional speech recognition algorithm is complicated and difficult to realize, and besides, the human beings ear-hearing model is not been adequately comprehended, the neural network, an analog of black box, is widely adverted as a newly rising speech recognition tool. Neural network has the ability of well self-adaptation, self-study, self-form and it's easy to realize. BP algorithm is the most widely used learning algorithm of neural network. It has two shortages:The convergence speed is low and being easily plunge into local optimization solution. The GA algorithm with strong global searching ability is thus brought in to improve the BP algorithm. It is a newly thriving aspect of research to combines the two theories of bionics.BP algorithm is combined with GA algorithm and applied in the research of speech recognition in this article. The advantages and disadvantages of genetic algorithm and BP algorithm are introduced, the significance and feasibility of combining GA and BP algorithm is deeply analyzed, a new recognizing algorithm based on evolutionary neural network is proposed according to the characteristics of speech recognition, and the steps of realizing that algorithm is listed. In the end of this article, a speech recognition system based on evolutionary neural network is realized in simulation system, its feasibility and good efficiency is proved by comparing it with the speech recognition system based on normal BP algorithm.