Gaussian Moment for Noisy Independent Component Analysis
Jun Xiao · Jisuanji fangzhen · 2006
The Independent Component Analysis as a method widely used in blind source separation is a hotspot in signal processing. But most of the ICA algorithoms ignore the noise infection of separated signals. In fact, actual signals have noise more or less. And when the signal-to-noise rate is under some value, separated signals will be bad. This paper defines the Gaussian function with different scale parameters, and shows how the Gaussian moments of a random variable can be estimated from noisy observations. This enables us to use Gaussian moments as one-unit contrast function that have asymptotic bias even in the presence of noise. To implement efficiently the maximization of the contrast functions based on Gaussian moments, a modification of our FastICA algorithom-noisyICA is introduced. Simulation experiments prove its feasibility and robustness.