Digital Network Echo Cancellation Using Genetic Algorithm and Combined GA-LMS Method

Echo cancellation for digital networks according to ITU-T G.168 is performed by using genetic algorithm (GA). A new procedure is proposed to estimate the coefficients of the adaptive filters used in echo cancellation with combination of the GA and least-mean-square (LMS) method. The results are compared for various values of LMS step size and two types of crossovers, heuristic and microbial, which are all satisfactory. The algorithm can estimate an echo path with the desired length of impulse response. Results show that the proposed combined GA-LMS method operates more satisfactorily than simple GA in terms of the number of generations needed to achieve a particular amount of echo cancellation. In similar cases, heuristic crossover causes faster convergence than the microbial one. Results are also compared with LMS algorithm. Although LMS is faster, but its solutions are less precise and diverge in some cases, while the proposed method converges in more cases