线性预测反滤波声门图

Inverse filering glottography by means of linear prediction technique

本文研究了四种分别从语音或气流信号中获取声门气流体积速度波形的方法，即语音单通道，语音──电声门图双通道、气流──电声门图双通道和语音──气流混合式线性预测反滤波声门图方法。因而，将原来仅适用于正常发声的线性预测反滤波声门图方法推广应用到正常与病变发声测试中。合成元音与人体发声实验结果证实了反滤波声门图算法的有效性；提取的声门气流体积速度波形和特征参数反映了嗓音源特性及其变化。

In this paper, four linear prediction inverse fitering glottographys (LPIFG) are studied, by which the glottal air flow volume relocity waveform can be drawn from speech or air flow signal. The four methods are one channel of speech,dual channels of speech-electroglottogram (EGG), dual channels of air flow-EGG and speech- air flow mixing LPIFG. As a result, the method of LPIFG, which was only suitable for normal phonation previously, is extended to normal and pathological phonation testing. The experimental results of synthesized vowels and human phonations verified that the theory and algrithms of LPIFG are valid and the obtained glottal air flow volume velocity waveforms and parameters can reflect voice source properties and their change.