Matrix-MCE based fuzzy neural network for speech recognition

Matrix-MCE (MMCE) based fuzzy neural network (FNN) for speech recognition is proposed in this paper. The environment noises usually degrade the performance of speech recognition. To reduce the effect of noises, MMCE is applied to minimize the classification error of two-dimension-cepstrum (TDC). Then the template matching employs FNN. To evaluate the performance, the speech data used for our experiments are a set of isolated Mandarin digits. Experimental results indicate that MMCE-based FNN works better than the other methods.