Feature extraction using wavelet packets strategy

In this paper, improved wavelet packets (WPs) decomposition coefficients of the frame are applied in the feature extraction method. In the proposed speech recognition system, the static WPs coefficients+dynamic WPs coefficients of the frame were employed as a basic feature. The framework of linear discriminant analysis (LDA) is used to derive an efficient and reduced-dimension speech parametric vector space for the speech recognition system. Using the continuous hidden Markov model (HMM) as the speech recognition model, the speech recognition system was successfully constructed. Experiments are performed on the speaker independent isolated-word speech recognition task. It is found that the improved WPs method achieves better recognition performance than the most popular Mel frequency cepstral coefficients (MFCC) feature extraction method in a noisy environment.