Research On The Application Of Speech Recognition Based On Radial Basis Function Neural Networks

With the development of computer and communications technology , human more and more aspire to communion with any machine at any time , any where smoothly. Speech is the best one of the ideal man-machine conversation interfaces. So, Speech Recognition was born.The ascendancy of Artificial Neural Network in the field of Speech Recognition which made it the top research .Radial Basis Function is one of novelty and effective forward- feedback network among many Neural Network models. Its topology fixes on the learning course, which makes the applicability of network better, and active function is the gauss function, which has very good part approaching an ability , the person studies speed as far as the person trains an algorithm, quickly, optimum problem of part does not exists. Because that, Radial Basis Function Neural Network is applied for Speech Recognition in this paper.In this paper , firstly, analyzed the current situation of speech recognition , the aspect of this paper based on the problem of speech recognition; secondly, introduced voice recognition basal principle and technology and provided theory support for speech recognition simulates systematically ; thirdly , expatiated the model of radial basis function, clustering algorithm and the way of modeling. Considering that cluster-algorithm is when ascertaining whose centre node off-speed to K mean value, this paper introduced an improved clustering algorithm based on Iterative Self-organizing Data Analysis Techniques Algorithm. For the sake of proving the improved clustering algorithm is better than K-means clustering algorithm ,this paper used the appliance facility instruction to be background, aimed at Speaker Independent, applied RBFNN for Speech Recognition field , adopted Virtual C++ and Matlab program skills, implemented speech recognition system simulation experiment using the two algorithms respectively on PC, and computed the results and experiment analysis. The experiment shows that under the same environment, comparing with K-means clustering algorithms , there is great enhancement in the result when using the improved clustering algorithm ,and explains sufficiently that the latter has great effect on the enhancement of the performance of RBF neural network, and makes RBF network has stronger classification ability. This experiment supplies the theory analysis and simulation data. Especially, the influence of some factors, such as feature parameter, number of training samples, background noise is discussed.