The Noise Reduction of Speech Signals Based on RBFN

The speech signals in the transmission are often accompanied with a lot of unnecessary disturbing noise. Usually, sound-absorbing cotton or directional microphones are applied on the hardware to suppress the background noise that caused by the unwanted signals, such as sound of wind and human voice. In addition, the channel effects are generated among various voice-processing systems when voice signals are transferred. This paper takes advantages of the Radial Basis Function Networks (RBFN), in which the noise signals can be figured simply in time period without the transfer through frequency domain, by matching the features of the sources of speech signals within the samples speech frame, the authentic speech signals can be restored. The simulation results show that the propose RBFN method works well to reduce the noise signals and provides the same quality as the original speech signals under Additive White Gaussian Noise (AWGN).