Fast converging methods for multichannel blind equalization or separation of multipath mixtures
R.H. Lambert, Chrysostomos L. Nikias · 2002
Algorithms for adapting FIR filters can also be applied to adaptation of FIR separation matrices (with multipath) for source separation. Fast converging algorithms robust to eigenvalue disparity are of interest, such as the exact least squares methods from the area of adaptive filters. Time domain adaptive multichannel blind recursive least squares (MBRLS) is presented. Three frequency domain algorithms based on the Bussgang property are given, two of which exhibit improved performance compared to the traditional Bussgang blind algorithms. The serial update method is studied as a fast blind update.