计算机工程与应用 ›› 2009, Vol. 45 ›› Issue (15): 158-159.DOI: 10.3778/j.issn.1002-8331.2009.15.045

• 数据库、信号与信息处理 • 上一篇    下一篇

一种改进的倒谱基音提取算法

朱 颖,钱盛友   

  1. 湖南师范大学 物理与信息科学学院,长沙 410081
  • 收稿日期:2008-03-20 修回日期:2008-05-19 出版日期:2009-05-21 发布日期:2009-05-21
  • 通讯作者: 朱 颖

Pitch detection using improved algorithm based on cepstrum

ZHU Ying,QIAN Sheng-you   

  1. College of Physics and Information Science,Hunan Normal University,Changsha 410081,China
  • Received:2008-03-20 Revised:2008-05-19 Online:2009-05-21 Published:2009-05-21
  • Contact: ZHU Ying

摘要: 介绍一种基音检测的新方法。该算法首先对分帧语音采用Burg算法进行LPC分析,获取最佳预测系数构造逆滤波器;然后通过逆滤波得到预测误差序列代替原输入语音序列进行倒谱分析;最后在恰当的范围里测出最大峰值,进行基音检测。实验结果表明,与传统方法相比,该方法鲁棒性好,在信噪比低的情况下仍可获得较高的准确率。

Abstract: A new algorithm of pitch extraction is proposed.In this algorithm,LPC analysis of segmented speech by Burg algorithm is performed to construct inverse filter with the best predictor coefficient.Then cepstrum analysis is proceeded by replacing the former speech sequence with the LPC prediction error sequence,and the maximum peak is tested in an appropriate scope to detect the speech pitch.The experimental results indicate that the proposed pitch extraction method is feasible and more robustly than the conventional method.Using the improved algorithm,a high accuracy can be obtained at low signal-to-noise ratio.