HYBRID VOICE RECOGNITION METHOD

摘要

PURPOSE: A hybrid voice recognition method is provided to generate an accurate optimum code book by combining the merits of a PNN(Probabilistic Neural Network) and a LVQ(Learning Vector Quantization), thereby enhancing performance. CONSTITUTION: A hybrid voice recognition method comprises steps of: generating an output vector X(t) using a PNN algorithm if voice recognition parameters are input; learning a LVQ algorithm by using the generated vector X(t) as the LVQ algorithm; generating an optimum code book if the output generated by learning the LVQ algorithm is agreed with preset value; and recognizing voice by executing a recognition algorithm using the optimum code book.