Initializing Blind Adaptive Equalizers Using A-Priori Environmental Knowledge

This paper addresses the problem of initializing blind adaptive equalizers. Due to the absence of information, the commonly used strategy is single spike initialization. Usually, this means starting the adaptive process far from the minimum mean-squared error, leading to a slow convergence. Blind equalization methods additionally suffer from local minima of the cost function in which the adaptive process might get trapped. These local minima, however, correspond to a large amount of residual intersymbol interference. Therefore, proper initialization might help to overcome both problems. We propose to incorporate coarse environmental knowledge to obtain an estimate of the channel impulse response. This estimate is then used to initialize the equalizer. Tests with two widely used blind equalization methods and experimental data confirm our claim.