A Novel Genetic Algorithm for Parameter Estimation of Sinusoidal Signals

초록

In this article, a novel genetic algorithm (NGA) for estimating parameters of sinusoidal signals is presented. Estimation of the relative parameters from periodic time series is a classical research topic in engineering and applications. To get accurate results, high-efficient techniques are increasingly demanded. For this, we propose several enhancements for traditional genetic algorithm (GA) in this research to improve the performance. Comprehensively considering the advantages and disadvantages of different fitness individuals, a new selection mechanism is posted to preserve population diversity. For improving the information exchange between individuals, a twostep crossover operator is designed. In addition, we establish an adaptive mutation to avoid premature convergence. In order to examine the performance of the proposed algorithm, benchmark function based numerical experiments and parameter estimation of sinusoidal signal are carried out. As a result, the NGA outperforms GA and particle swarm optimization (PSO) in terms of accuracy and robustness.