Speech Recognition System Using a Wavelet Packet and Synergetic Neural Network

A recognition system is proposed, for audio and speech signal processing, which based on a Wavelet Packet (WP) and Synergetic Neural Network (SNN). Firstly, in the step of features extraction, the Mel-frequency cepstral coefficients (MFCC) algorithm is developed by using WP instead of FT. Secondly, we make use of SNN to pattern recognition. Due to the Kohonen rule,the SNN can be combined with improved MFCC very well in the system. Finally, the experiment results show that the speech recognition system has higher recognition rate and potential advantages in automatic speech recognition.