Speech Recognition System Based On S-Transformand Improved Spectral Subtraction

Spectral subtraction and improved spectral subtraction are commonly used in speech enhancement methods,but both spectral subtraction and improved spectral subtraction are used in the process of Fourier transform,because the Fourier transform of a single conversion can not be effectively expressed Time-frequency domain common characteristics,dealing with speech signals such as non-stationary signal becomes "powerless." The S-transform can be used to represent the time-frequency characteristics of the signal.It is widely used in the non-stationary signal processing of seismic signals.In this paper,the non-stationary signal of speech is carried out by the combination of S-transform and improved spectral subtraction Noise treatment.Experiments show that the denoising effect is better than that of the basic spectral subtraction and the improved spectral subtraction.The effect of noise removal on the speech can improve the recognition rate of the speaker recognition system,which is very important for the realization of the speaker recognition system.