A study on enhancement of speech signal using separated bandwidth and non-uniform sampling

Abstract

There are various methods to improve signal quality under the environments of noised signal in the speech signal processing. In the case of the majority of signal restoration, it will be concentrated in signal quality for deterioration of low frequency as they are also converged into voiced signal from the characteristic of speech quality deterioration. Therefore, in the case of deterioration of voiced signal section, great efforts have been contributed for quality improvement and restoration. Other case of clipped signal section is for signal reconstruction. In this paper, the method to improve voiced and clipped signal will be proposed. Voiced signal in noisy and clipped section is created by utilizing the fact that differentiation and clipped signal are supposed to give small influence for the evaluation of recognition level. For the signal which is obtained by stressing amplitude at voiced signal section where most of them are distributed, the method of emphasizing synthesis with original signal was used. In this paper we propose the method which enhances the voiced signals by dividing the clipped signal and the low frequency signal. We differentiate the signals and use the fact that signals have little effect on perception to make voiced part, and isolated the high frequency part through low pass filter. © 2012 Springer-Verlag.