高质量线性预测分析语音的一种方法

A Scheme for High Quality Linear Prediction Analysis of Speech

提出了高质量线性预测分析语音的一种方法：首先用语音信号的小波变换检测声门闭合的时刻（ＧＣＩ），而两个相邻ＧＣＩ的间隔即为基音周期，从而可进行基音周期同步分析；其次，选择声门闭合时刻附近以外的声道受到弱激励的语音样本进行线性预测分析．该方法使选择样本的预测误差明显降低，大大提高声道参数估计的精度。

Because of the existence of pseudoperiodic excitation impulses, the conventional Linear Prediction ( LP ) analysis is often inaccurate for estimating vocal tract parameters of voiced speech. To solve the problem, a scheme is proposed for high quality Linear Prediction analysis of speech. The scheme consisits of two main stages. First, the glottal closure instant ( GCI ) is detected using wavelet transform of speech signals, the interval between the adjacent two GCI's is a pitch period and the LP analysis can be executed pitchsynchronously on one or several pitch periods. Second, samples which are far from GCI are selected and LP analysis is executed on these selected samples rather than all samples in an analysis frame because the excitation impulses are much stronger around GCI than those of the selected samples in a pitch period. The prediction errors of the selected samples become very small. Thus, the accuracy of estimating vocal tract parameters is improved significantly. The validity of the proposed method is confirmed by comparing the results obtained from the proposed scheme and those of the conventional LP analysis method.