计算机工程与应用 ›› 2009, Vol. 45 ›› Issue (9): 174-176.DOI: 10.3778/j.issn.1002-8331.2009.09.050

• 图形、图像、模式识别 • 上一篇    下一篇

基于SVM的非特定人声调识别的研究

肖汉光1,蔡从中2   

  1. 1.重庆工学院 数理学院,重庆 400054
    2.重庆大学 数理学院,重庆 400044
  • 收稿日期:2008-01-11 修回日期:2008-04-09 出版日期:2009-03-21 发布日期:2009-03-21
  • 通讯作者: 肖汉光

Study of speaker-independent tone recognition based on support vector machine

XIAO Han-guang1,CAI Cong-zhong2   

  1. 1.School of Mathematics and Physics,Chongqing Institute of Technology,Chongqing 400054,China
    2.School of Mathematics and Physics,Chongqing University,Chongqing 400044,China
  • Received:2008-01-11 Revised:2008-04-09 Online:2009-03-21 Published:2009-03-21
  • Contact: XIAO Han-guang

摘要: 在建立非特定人普通话四声语调语音数据库的基础上,采用Mel频率倒谱系数(MFCCs)对语音数据进行特征参数的提取,并利用支持向量机(SVM)对语音中的四种声调进行了训练和识别研究。实验结果表明MFCCs和SVM的结合得到的平均识别率达到了97.6%。

关键词: 声调识别, 特征提取, Mel频率倒谱系数(MFCC), 支持向量机

Abstract: A speaker-independent tone database of Chinese speech(putonghua) is established.The Mel-frequency cepstrum coefficients(MFCCs) are used for extraction of the tone feature parameters.The four recognizing models of four tones are trained by using support vector machine(SVM),and are tested by using the testing tone data.The results show that a recognition accuracy can reach 97.6% by combining MFCCs and SVM.

Key words: tone recognition, feature extraction, Mel-Frequency Cepstrum Coefficients(MFCCs), Support Vector Machine(SVM)