虚拟噪声通道在基于ICA消噪过程中的应用

Application of Virtual Noise Channels in the Processing of Signal De-noising with ICA

考虑应用独立分量分析方法进行信号的消噪处理时,要求观测通道的个数大于或等于独立信源的个数的条件,以及实际应用中的噪声源的随机性,尝试引入虚拟噪声通道的概念,克服难以确定噪声源的个数,因而很难满足独立分量分析方法应用要求的限制,根据具体的检测条件和被测对象的特性,引入相应的虚拟通道数,从而提高独立分量分析方法在消噪方面的应用性能。实验表明,利用该方法可以得到很好的消噪结果,有效提高信号的信噪比。

In the application of Independent Component Analysis (ICA), it is usually needed that the number of observed channels should be lager than or equal to the number of independent signal sources. However, because of the randomicity of noise source, the number of noise is often uncertain in practical applications. To improve the application performance in signal de-noising of ICA, the concept of virtual noise channel was introduced and the needed virtual channels were imported based on the concrete detected conditions and the characteristics of detected objects. The experiments indicate that good de-noised results can be obtained and the signal-noise ratio can be improved effectively with the method.