Soft Computing Techniques to Estimate FIR Filter Weights in an Adaptive Channel Equalizer: A Comparative Study

An adaptive channel equalizer is ubiquitous in combating the effect of Inter-Symbol Interference (ISI) and other impairments caused by cross talk, additive noise, electronic components present in transceivers and the nature of medium, on the digital data transmitted in a communication system. Techniques to estimate the weights of finite impulse response (FIR) filter, which forms an integral part of the equalizer, using few popular soft computing techniques are presented here. Parameterized expression of the filter cost function under adaptation and corresponding training set is created. Thereafter, global minimization process is executed using Genetic Algorithm (GA), Particle Swarm Optimization (PSO), Bacterial Foraging Optimization (BFO), Wind Driven Optimization (WDO) and the benchmark Least Mean Square (LMS) algorithm. A design example is undertaken to verify the effectiveness of each technique in estimating the weights of FIR filter based adaptive channel equalizer. Extensive simulation is carried out on linear and nonlinear channel using the five popular and successful algorithms at 10dB additive noise. WDO achieves minimum error of -9.574dB at 35 iterations as compared to -5.153dB at 20 iterations, -5.993dB at 33 iterations, -8.66dB at 23 iterations for GA, PSO and BFO respectively. It also shows better performance in terms of bit error rate (BER), which is 800 bits as compared to 5717, 13640, 3630 and 8508 bits in 10 samples for LMS, GA, PSO and BFO respectively. This paper presents a good and comprehensive set of results and states arguments for the merits and demerits of each of the technique. On careful observation, it is revealed that WDO emerges as the fastest optimization algorithm for nonlinear channel.