Decision Feedback Equalization using Particle Swarm Optimization

It is well-known that the Decision Feedback Equalizer (DFE) outperforms the Linear Equalizer (LE) for highly dispersive channels. For time-varying channels, adaptive equalizers are commonly designed based on the Least Mean Square (LMS) algorithm which, unfortunately, has the limitation of slow convergence specially in channels having large eigenvalue spread. The eigenvalue problem becomes even more pronounced in Multiple-Input Multiple-Output (MIMO) channels. Particle Swarm Optimization (PSO) enjoys fast convergence and, therefore, its application to the DFE merits investigation. In this paper, we show that a PSO-DFE with a variable constriction factor is superior to the LMS/RLS-based DFE (LMS/RLS-DFE) and PSO-based LE (PSO-LE), especially on channels with large eigenvalue spread. We also propose a hybrid PSO-LMS-DFE algorithm, and modify it to deal with complex-valued data. The PSO-LMS-DFE not only outperforms the PSO-DFE in terms of performance but its complexity is also low. To further reduce its complexity, a fast PSO-LMS-DFE algorithm is introduced.