论文部分内容阅读
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，其有效性已被实验结果所证明
A method of high-quality linear predictive analysis of speech is proposed. Firstly, the glottal closure time (GCI) is detected by wavelet transform of speech signals, and the interval of two adjacent GCIs is the pitch period, so that the pitch period synchronization analysis Secondly, we choose the speech samples with weak excitation outside the glottis closing time to perform linear predictive analysis. This method significantly reduces the prediction error of selected samples and greatly improves the accuracy of channel parameter estimation. The validity of this method has been proved by experimental results