Improvement of accuracy of linear prediction analysis by processing of the variable extension of pitch period and processing of windowing

摘要

In this paper, we proposed a method to improve the accuracy of analysis of high pitch speech by processing of the variable extension of pitch period by linear prediction analysis and synthesis, and processing of windowing to separate waveform every pitch period, and then doing linear prediction analysis of these waveforms collectively. We show that by these processing the influence of the short pitch period is reduced and the analysis accuracy is improved. And we investigated the tolerance to noise of the proposed method.