Constrained optimization methods for direct blind equalization

Constrained optimization techniques are studied in this paper for direct de sign of linear multichannel equalizers. Novel blind algorithms are derived by minimizing the equalizer's output variance subject to appropriate constr aints. The constraints are chosen to guarantee no desired signal cancellati on, and their parameters are jointly optimized to maximize the signal compo nent at the output. The resulting blind algorithm was observed to have near optimal performance at high signal-to-noise ratio, i.e., close to the perf ormance of the trained minimum mean-square-error receiver, Also, the propos ed method is not sensitive to the color of the transmitted sequence. Analyt ical expressions are derived to quantify the algorithm's performance.