基于核函数的混合音频多路信号盲分离方法

Blind Separation Method of Mixed Audio Multichannel Signal Based on Kernel Function

基于核函数的混合音频多路信号盲分离方法可提升音频信号的分离精度,实现高精度、高效多路混合音频信号盲分离。通过结合核函数与独立分量分析获得核函数独立分量分析方法,运用均值化与白化预处理初始混合音频多路信号,简化盲分离运算过程,运用核函数独立分量分析方法盲分离预处理后的混合音频多路信号,获得多路分离音频信号输出,实现混合音频多路信号的盲分离。结果表明,所得多路分离音频信号与源音频信号具有较高相似系数和信噪比,二者几乎完全吻合,可获得良好的分离效果,具有较高的实际应用价值。

The blind separation method of mixed audio multi-channel signals based on kernel function can improve the separation accuracy of audio signals and achieve high-precision and efficient blind separation of multi-channel mixed audio signals.By combining the kernel function and independent component analysis,the kernel function independent component analysis method is obtained,the initial mixed audio multi-channel signal is pre-processed by averaging and whitening,and the blind separation operation process is simplified.The kernel function independent component analysis method is used to blindly separate the preprocessed mix multiple audio signals,obtain multiple separated audio signal output,and realize the blind separation of mixed audio multiple signals.The results show that the obtained multi-channel separates audio signal,and the source audio signal has high similar coefficients and signal-to-noise ratio.The two are almost completely consistent.Hence,a good separation effect can be obtained,and it has high practical application value.