基于语音波形外观的基音提取算法  

Pitch EStimation Algorithm Based on Profile of Speech Waveform

作　　者：王兵[1] 苏恩泽[1] 

机构地区：[1]空军工程学院系统工程与计算机应用系

出　　处：《数据采集与处理》1995年第1期73-76,共4页Journal of Data Acquisition and Processing

摘　　要：在对人用视觉提取基音周期过程模拟的基础上，提出一种基于语音波形外观形状的时城基音周期提取算法。该算法利用语音波形的一次峰值点和二次峰值点的幅度和位置以及后峰到前峰的距离等几种属性，来判断决定基音周期值，具有算法简单、运算量小、能准确定位各基音周期位置的特点。此外，由于不需加窗处理，免去了加窗处理所带来的平均效应，该算法在处理相邻基音周期差值较大的语音信号时具有明显优势。该算法比较适合于做各种基音同步算法的前处理。On the basis of the simulation of human pitch extraction Process with eyes,a new pitch estimation algorithm hased on the profile of speech waveform is put forward.This algorithm gets pitch periods by applying several attributes, such as the amplitude and position of Peak and quadric Peak of speech waveform and distance from one Peak to the former peak. The algorithm has the following advantages; (l ) it is simple, and with high accuracy (2 ) it doesn' t need too much calculation and (3) itcan locate the pitch peridds. In addition, without window processing, the average effect brought aboutby windowing is absent. When handling speech signals of large difference between two pitch periods,this algorithm has its apparent advantages, so the algorithm is fit for working as a pre--processor of thepitch--synchronous algorithm.

关 键 词：语音信号处理 基音估值 基音提取 语音波形 

分 类 号：TN912.3[电子电信—通信与信息系统]