A Study on Matching Pursuit Based on Genetic Algorithm
Qiang Gao, Duan Chen-dong, Fang Xiangbo, Liu Benchao · 2011
As a widely used adaptive signal decomposition method, a matching pursuit involves enormous computation cost because a greedy strategy is applied. The matching pursuit based on genetic algorithm has been proposed to reduce the computation cost. However, the numerical experiment results presented in this paper show that the genetic algorithm often converges to the local optimal solution when it is used in matching pursuit. Our studies indicate that a parameter of the time-frequency atom is very sensitive to the fitness and influences the global search ability of the genetic algorithm greatly. The characteristics of this parameter are investigated, and a logarithm scaling is used to solve this problem. The results show that the logarithm scaling can improve the global search ability of the genetic algorithm used in matching pursuit significantly.