Research on the Influence of Different Feature Parameters on Speech Recognition Rate

It is necessary to extract feature parameters that can reflect the essential characteristics of the speech signal, so as to recognize speech signal. In this paper, three feature parameters have been introduced, namely, LPCC, MFCC, and RAS_MFCC, which are common in speech recognition. The recognition rates of the three feature parameters under different signal-to-noise ratios with or without differences are compared, the WFBA method is applied to MFCC. The experimental results demonstrate that the MFCC parameter performs best at low signal-to-noise ratios; the recognition rate of the RAS_MFCC feature parameter is better when the signal-to-noise ratio is higher than 5dB. The effect is significantly improved greatly when the feature parameters have differences. Furthermore, the recognition rate of the RAS_MFCC feature parameter with difference can reach 98.2% at the highest when MFCC has been enhanced by the DWFBA method.