一种语音信号基音检测的功率谱二次处理方法  

A Power Spectrum Reprocessing Algorithm for Pitch Detection of Speech

作　　者：张天骐[1] 张战[1] 林孝康[1] 权进国[1] 

机构地区：[1]清华大学深圳研究生院,深圳518055

出　　处：《计算机工程》2006年第1期1-3,59,共4页Computer Engineering

摘　　要：对基音检测的倒谱法进行了细致的分析,指出其在基音检测计算及其数字实现时存在不足。为了克服倒谱法的不足之处,提出了一种全新的功率谱二次处理基音检测方法,该方法在克服了倒谱法在基音检测中不足之处的同时还提高了计算速度,对基音轨迹曲线的估计准确性也得到了较大改善。This paper analyses the cepstrum method of pitch detection carefully, and points out its several shortcomings when it is being realized digitally. Furthermore, it presents a new pitch detection algorithm based on digital spectral analysis. It has the characteristics of simplification, and is easy to be realized by digital signal processing. The algorithms overcome the shortcomings of the presented conventional cepstrum algorithms. It uses the power spectrum reprocessing results of speech to extract its pitch contour. The results of computer simulation show that this algorithm can work well at high speed.

关 键 词：功率谱二次处理 倒谱 基音检测和基音轨迹 

分 类 号：TN912.16[电子电信—通信与信息系统]