Research On The MUSIC Spectral Peak Searching Based On The Genetic Algorithms

MUSIC algorithms as a classic super-resolution algorithm for spatial spectrumestimation has important application in the multi-dimensional parameter estimationof array signal. But because the calculation of searching spectral peak process isheavy, it can’t meet the requirement of high real-time. As genetic algorithms has theadvantages of parallelism, global optimization and simplicity etc., and the widelyapplication in the complex function optimization, make it a good solution for theabove problem. But it also has some problems. In this paper, we improved thegenetic algorithms to the premature convergence phenomena and the problem ofpoor local search ability.For global optimum problem, this paper presents a double mutation operatorgenetic algorithms, it combines orthogonal mutation operator and multilocusmutation operator together. The orthogonal mutation operator is used for balancingthe ratio of each locus genes’ gene value in order to maintain the diversity of thepopulation to overcome the premature convergence, and the little mutationprobability multi-locus mutation operator is used to improve the local search abilityof the algorithm. Simulation experiment indicates that the improvement measuresare effective.For the problem of seeking more than one local optimum, an improveddeterministic crowding niche genetic algorithms and a new hierarchical gradientoperator for two-dimensional parameter estimation are proposed. The migrationstrategy, hierarchical gradient operator and the genetically balanced thinking areapplied to the niche genetic algorithms. The improved niche genetic algorithmswhich can better maintain the diversity of population, strengthen the local searchability, and the accuracy of the solution is higher. The effectiveness of this algorithmis verified by the sample application.Finally, the improved genetic algorithms is applied to the MUSIC algorithmsspectral peak searching process. The performance of the improved geneticalgorithms for the MUSIC spectrum peaks in the case of single-source andmulti-source search are simulated. Compared genetic algorithms and the traversalalgorithms, genetic algorithms can largely reduce the computation of the MUSICspectrum peak search.This study not only significance for the application of the MUSIC algorithms inreal-time occasion, but also has important reference value for other complexfunctions optimization.