Research on the Algorithm of Noisy-Robust Tibetan Speech Recognition Based on RBF

Aiming at the problem of Tibetan speech recognition under the condition of resistance from noise, a kind of Tibetan speech recognition algorithm, combining RBF network with auditory feature was presented in this paper. The description for the Tibetan speech signals was carried out with Mel frequency cepstrum constant (MFCC), and the recognition classifier was designed based on RBF network with the property of supervised learning based on gradient descent. The simulation on the presented algorithm was run, and the results illustrates that the presented algorithm is valid.