论文部分内容阅读
先讨论由语音信号反演声门激励函数,发现波沿非均匀声管传导既能产生新的频率(共振作用),也可抑制原有频率(滤波作用).这说明目前用于语音处理基础的全极点LPC模型并不能充分反映这些特征.然后研究如何直接从语音生成的偏微分方程模型出发,用反演方法提取一种新的语音特征参数代替LPC作为研究各种语音处理的基础.
Firstly, we discuss the glottal excitation function by the voice signal. It is found that the wave can generate new frequency (resonance effect) and suppress the original frequency (filtering effect) along the non-uniform acoustic conduction. This shows that the current all-pole LPC model used for speech processing does not fully reflect these features. Then we study how to extract a new speech feature parameter instead of LPC as the basis for studying various speech processing directly from the partial differential equation model of speech generation.