计算机工程 ›› 2010, Vol. 36 ›› Issue (5): 162-164.doi: 10.3969/j.issn.1000-3428.2010.05.059

• 人工智能及识别技术 • 上一篇    下一篇

基于高斯过程模型的语音增强

沈 赟,张丽清   

  1. (上海交通大学计算机科学与工程系,上海 200240)
  • 收稿日期:1900-01-01 修回日期:1900-01-01 出版日期:2010-03-05 发布日期:2010-03-05

Speech Enhancement Based on Gaussian Process Model

SHEN Yun, ZHANG Li-qing   

  1. (Department of Computer Science and Engineering, Shanghai Jiaotong University, Shanghai 200240)
  • Received:1900-01-01 Revised:1900-01-01 Online:2010-03-05 Published:2010-03-05

摘要: 针对信号处理领域的语音活动探测问题,提出一种基于高斯过程先验假设的概率方法,用于增强语音。利用高斯过程模型的后验概率来估计纯净语音,使用在学习过程中得到的高斯过程模型的参数探测语音活动。实验结果表明,该方法对于在白噪声和有色噪声环境下的语音有较好的增强效果。

关键词: 语音增强, 高斯过程, 贝叶斯方法

Abstract: Aiming at the problem of Voice Activity Detection(VAD) in signal process area, this paper presents a probabilistic method employing Gaussian Process(GP) prior to deal with speech enhancement. Clean speech is estimated by posterior probability calculated with GP model, while VAD is solved by the scale of one hyperparameter of GP model estimated in learning process. Experimental results show good performance of this method for white and colored noise.

Key words: speech enhancement, Gaussian Process(GP), Bayesian method

中图分类号: