A Multiresolution-based Method to Filter Noise and Spikes in Speech Signal

The research of speech signal has been the significant progress recently. Regarding the convenience to the further analysis and application of speech signal, the elimination of meaningless periods in a given speech signal is an indispensable pre-procedure in the research topics related to speech signal. However, the aforementioned is still not to be much solved accurately, noise and spikes especially. Accordingly, we propose a simple and effective method rather than requiring complex procedures. A multiresolution-based method is proposed, which is able to filter noise and spikes in speech signal automatically. Experimental results demonstrate the reliability and feasibility of the proposed method.