Modified 2-D Cepstrum Using Soft-Computing

摘要

We proposed a method which can solve the low speech recognition rate problem under noisy environment. in our system, the method is GAS-based Speech Recognition using Two Dimensional cepstrum. Two dimensional cepstrum (TDC) can simultaneously represent several kinds of information contained in the speech waveform: static and dynamic features, as well as global and fine frequency structures. from analysis, an utterance only some TDC coefficients will be selected to form a feature vector. Hence, it has the advantages of soft computation and less storage space. However, it is quite sensitive to background noise. in order to solve this problem, we propose the GAS-based M-TDC method in our system to improve the performance of TDC under noisy condition. from the experiments with five noise types, we found that the GAS-based M-TDC have better recognition results than the TDC under the noisy environments.