摘要
近年来,盲源分离算法由于其良好的去噪效果在信号处理领域得到了广泛应用,但传统独立分量分析方法存在着未考虑噪声干扰及未充分利用已知信息等弊端。提出基于GAR模型的变分贝叶斯独立分量分析算法,将源信号的时间结构与系统噪声进行融合研究,基于GAR模型近似建模语音信号的时间结构特征,应用变分贝叶斯学习方法分离带噪声的语音信号。通过与标准变分贝叶斯独立分量分析算法的仿真对比,证明改进后的算法有较好地实际分离效果,有效解决了ICA算法无法在噪声环境下直接进行盲源分离问题。算法可用于减轻铁路列车司机通信时的听觉疲劳。
In recent years, the blind source separation algorithm has been widely used in the field of signal processing due to its good denoising effect. However, the traditional independent component analysis method has the disadvantages of not considering noise interference and underutilizing known information. A variational Bayesian independent component analysis algorithm based on GAR model is proposed. The time structure of the source signal is integrated with the system noise. The time structure of the speech signal is approximated based on the GAR model. The variational Bayesian learning method is applied. Separate the noisy speech signal. Compared with the standard variational Bayesian independent component analysis algorithm, it is proved that the improved algorithm has better practical separation effect, which effectively solves the problem that ICA algorithm can not directly perform blind source separation in noisy environment. The algorithm can be used to alleviate the communication hearing fatigue of railway train drivers.
作者
徐岩
王雷
XU Yan;WANG Lei(School of Electronic and Information Engineering, Lanzhou Jiaotong University, Lanzhou 730070,China)
出处
《铁道学报》
EI
CAS
CSCD
北大核心
2019年第9期88-93,共6页
Journal of the China Railway Society
基金
国家自然科学基金(61461024)
关键词
AR模型
独立分量分析
变分贝叶斯
盲源分离
autoregressive model
independent component analysis
variational Bayesian
blind source separation
