Simplified Blind Algorithms based on Cross-Information Potential

초록

The learning algorithm for blind signal processing developed from the performance criterion of cross-information potential (CIP) has superior compensation performance for intersymbol interference induced by channel distortion, even under impulsive noise. One of the drawbacks of the CIP algorithm is a heavy computational complexity caused by considering all the interactions between N (sample size) output samples and the symbol points where a large sample size is preferable to guarantee a desired level of accuracy in distribution estimation. In this paper, the idea of taking only the current output sample into their interactions instead of all output samples is proposed under the assumption that the information the current sample has is the most useful of all other samples; this leads to the computational complexity of the proposed algorithm not being related to the sample size. The simulation results show that the proposed algorithm significantly reduces the computational complexity by approximately 21 times for N=20 without noticeable loss of learning performance, which indicates that the proposed criterion and algorithm are more suitable for practical implementations than the conventional CIP algorithm.

키워드