Iterative Detection Based on Widely Linear Processing and Real-Valued Symbol Alphabets

In this paper, we consider an iterative or turbo receiver using a widely linear Wiener Filter (WF), i. e., the optimal widely linear filter based on the mean square error criterion. In communication systems where the transmitted symbols can be assumed to be realizations of a noncircular random variable, widely linear processing outperforms the linear one. Besides the non-circularity due to the a priori information fed back in the turbo scheme, we investigate real-valued modulation alphabets which are noncircular even if the a priori information is zero. We derive the widely linear WF for this case which has the same order of computational complexity as optimal linear processing. Moreover, we reveal the relationship between the proposed widely linear equalizer and interference cancellation based on the available a priori information. Simulation results of a frequency-selective system show the superiority of the proposed detector compared to the linear one although they have the same order of computational complexity.