Nonuniform Sampling Weak Signal Detection in Frequency Domain Based on ICA
Wang Zheng-ying · Journal of Tianjin University Science and Technology · 2007
Nonuniform sampling can acquire signal of which the frequency exceeds the threshold in Nyquist theorem when the sample rate of analogue-to-digital device is lower than usual.When two signals are sampled by nonuniform sampling,and the amplitude of one signal is above 10% smaller than that of the other,the weak signal cannot be detected from the mixed signal′s spectrum because of sampling′s fake-randomicity.Because independent component analysis(ICA) was immune to signal′s amplitude,the weak signal could be detected from the mixed signal′s spectrum with two input source signals.One input source signal was the mixed signal itself,and the other was a constructed signal of which the frequency was equal to that of the strong signal.The weak signal,of which the amplitude was five million times smaller than the strong one,was separated from the mixed signal′s spectrum by the FastICA algorithm.A phase-scanning method using cross correlation was adopted to solve the strong signal′s phase matching.A weak signal,of which the amplitude was 0.06% smaller than that of the strong signal,was separated from the mixed signal,as the relative error of phase matching was 0.70%.