Method for Separating Multiplicative Noise Sharing a Common Band with Desired Signal

It is hard for the traditional homomorphic filtering technique to deal with the multiplicative noise effectively as its frequency band is overlapped with the one of desired signal. A novel method is proposed that is able to separate multiplicative noise from observed signal even though they share a common band completely. The method integrates the basic idea of homomorphic transform with the advantages of partial time-domain coherence algorithm in separating linear-mixed signals. It works by taking logarithmic operation first to observed signal, a mixture of the signal and noise, and then applying partial time-domain coherence algorithm to it in order to analyze the power proportion of each separated component. The results of computer simulation experiments have shown that the separation of orthogonal multiplicative signals can be very accurate, and the separation of partial nonorthogonal signals or random multiplicative noise are still acceptable.