基于Bessel函数展开的语音信号ICA-R算法

The algorithm of voice ICA-R based on Bessel function expansion

为了解决语音信号参考独立分量分析(ICA with Reference,ICA-R)算法中参考信号选取难的问题.依据语音信号传播机理和Bessel函数展开系数对语音信号的表征能力,提出了基于Bessel函数展开的语音信号建模方法,利用Bessel函数展开系数作为变换系数,用少量的系数构建ICA-R中的参考信号来实现对期望语音信号的提取,并将该方法运用到语音信号隐藏技术中.仿真实验结果表明该方法能够较好地从混合语音信号中提取期望语音信号,并能有效地实现语音信号的隐藏与恢复.

According to propagation mechanism of voice signal and characterization capabilities of Bessel function expansion coefficient, this paper presents a method of modeling voice signal. The method bases on Bessel function expansion, and it makes choosing reference signal easily in voice Independent Component Analysis with Reference （ ICA with Reference, ICA - R） algorithm. The reference signal in ICA - R is constructed by coefficient, and voice signals are extracted in this way. Results of simulation experiment indicate that this method can extract expected signal, hid and recover voice signals effectively.