用混合编码遗传算法实现匹配追踪算法

Matching Pursuit via Genetic Algorithm Based on Hybrid Coding

在分析总结浮点数编码和格雷码编码各自特点的基础上,提出了一种用浮点数和格雷码混合编码的遗传算法来实现匹配追踪算法.该算法有机结合了遗传算法和匹配追踪算法的优点,不仅能够得到较高精度的最佳匹配参数,而且有效地降低了计算量,克服了匹配追踪算法由于计算量太大而不能广泛应用的缺点.计算机仿真结果表明,该算法提取相位的精度和提取时间均有明显改善,证实了该算法的准确性.最后,将该算法应用于转子实验台的冲击信号特征提取中,提取结果证明了它的实际应用价值.

A novel algorithm for decomposing any signal into a linear expansion of elementary functions with a redundant dictionary is proposed. The algorithm is based on the conventional matching pursuit (MP) and the genetic algorithm (GA), where MP is implemented via GA with hybrid coding efficiently combined float coding with Gray coding. Consequently, the computational cost and computational error, compared with conventional MP, are obviously reduced. The properties of the proposed algorithm are investigated. Its ability to decompose signals is illustrated by simulated and practical data, and the results verify the reliability and accuracy of this algorithm.