Blind Equalization based on Time-independency Principle
Hiroyoshi Oda, Yoichi Sato · International Symposium on Information Theory and its Applications · 1994
Blind equalization is the problem restoring the transmitted signal only by observation of the received signal over the non minimum phase channel. Traditional adaptive algorithms adjust the equalizer parameters to minimize the cost function based on the distribution matching principle. We investigate the blind equalization problem from the view point of the time-independency principle: under the assumption that the transmitted signal is independent identically distributed (IID) sequence, the transparency of the total system combined the channel with the equalizer can be achieved, if the equalizer output is also IID sequence. Firstly, for this strategy, we aim to lay the theoretical foundation based on the mutual information. Secondly, practical off-line algorithms are proposed by using the statistics such as higher order cumulants in order to measure the time-independency at equalizer output. The simplification of algorithms is also realized by the zero-forcing principle. Moreover, the off-line algorithm yielding the most precise estimate of equalizer tap weights is obtained by the design of the adjusting term. Finally, on-line versions of the proposed algorithms are derived.