Study on the Influence of Clipping on Low Bit Speech Coding

Clipping can affect the quality of speech coding, especially under the condition of low bit speech coding. Clipped speech no longer conforms to human voice model, and the quality of speech coding for clipped speech is seriously degraded. In order to study the effect of clipping on low bit speech coding, two aspects of the parameters extraction of clipped speech and the quality of clipped speech coding are analyzed. Deviation degree is adopted to measure the accuracy of parameter extraction, and the parameters including LPC, pitch and voicing decision. The error distribution, error types and error causes of various parameters are analyzed under different clipping degrees. The coding quality of clipped speech are evaluated by objective perceptive speech quality assessment (PESQ). In order to solve the problem that the performance of the traditional declipping methods is decreased when the clipping degree is large, it is proposed to use two improved declipping methods to repair the clipping speech. The experimental results show that the improved declipping method can effectively improve the performance of low speed speech coding when the clipping degree is large.