汉语语音基音轨迹的层叠滤波和平滑

为了获得更加精确的汉语语音基音周期轨迹,研究了对汉语语音的基音周期轨迹进行层叠滤波和平滑的方法.基音是指发浊音时声带振动所引起的周期性,而声带振动频率的倒数就是基音周期.语音信号处理中精确提取特定语音的基音周期有重要的意义,基音周期的检测无论在语音信号的合成、编码还是识别方面,都起着非常重要的作用.本文依据数字语音信号的特点,设计了一种用于语音处理的层叠滤波算法,合理选择结构元素构筑了一组用于汉语语音基音轨迹平滑的层叠滤波器,仿真实验结果表明这种方法对于随机误判点与传统的基于中值与线性平滑的组合平滑相比效果更佳,同时也证明了层叠滤波器用于语音信号研究的可行性.

Stack filtering and smoothing to pitch contrail of Chinese speech

For the sake of getting more accurate pitch period,a method of stack filtering and smoothing was investigated.The pitch is the vocal vibratile periodicity of voiced speech,the reciprocal of vocal vibratile frequency is the pitch period.It is important to get accurate pitch period in speech signal processing,and the detection of pitch period is important for the synthesis,coding and recognition of speech.Based on the characteristic of speech signals,a new stack filtering algorithm is proposed and a new stack filter with suitable structuring elements is designed to get the preferable accurate pitch contrail of speech signals.Simulation results show that this new approach is better than the combination filter based on the median filter and linear filter on the stochastic incorrect value,and furthermore,it has been proved that the stack filter is feasible for the speech signal research.