Abstract
The filter-based turbo equalization scheme has been proposed in several papers to avoid the prohibitive complexity imposed by the trellis-based turbo equalization. In the existing literature,the filter-based approach has been solely implemented by a linear MMSE filter, the coeffi cients of which are updated to minimize the mean-square error for every output symbol of the equalizer. A new turbo equalization algorithm is introduced in this paper. It has a lower computational complexity compared to most of the existing MMSE filter-based turbo equalization schemes. The complexity reduction is accomplished by deriving log-likelihood ratios (LLRs) directly from the output of an interference canceler, thus avoiding the MMSE filtering and its inherent matrix inversion for each symbol estimate. Numerical results show that the proposed scheme enables ISI-free transmission for some frequency selective channels.