A decision feed-back blind equalization algorithm based on a cluster formation technique

This paper proposes a decision feed-back blind equalization (DFBE) algorithm to reconstruct a QPSK constellation degraded by a stationary multipath channel and additive white Gaussian noise. The equalizer structure is composed of two filters, a feed-forward filter and a feed-back filter. The inputs to the feed-forward and feed-back filters are the received signal sequence and the detected symbols, respectively. Both filters are updated by a stochastic gradient algorithm in proportion to an estimated error based on a Gaussian cluster formation technique. During equalization, the QPSK symbol constellation is modeled as Gaussian clusters centered around the transmitted symbols. Both equalizer filters are updated to minimize the variance of the Gaussian clusters, thereby reconstructing the transmitted symbols. The blind nature of the DFBE algorithm makes it very attractive for mobile communications applications where the reliable transmission of signals through rapidly changing multipath fading channels with deep spectral nulls is a challenging problem yet to be resolved.