稳态噪声背景下加权Myriad滤波的遗传算法实现

该文利用遗传算法具有全局最优搜索的特点,同时,考虑到运算复杂度问题,提出了用十进制代替传统的二进制进行算子操作的遗传算法,来获得加权Myriad滤波器输出。另外,针对在原自适应权值训练算法中权值处于稳定区后波动仍较大的问题,分析了波动存在的原因并对其进行了改进.仿真结果表明,该方法在较多极值点的情况下,所有染色体均可快速收敛至全局极值点,并且改进后权值估计可使滤波效果有明显改善。

Weighted Myriad Filter Implementation with Genetic Algorithm in -Stable Noise Environments

A novel algorithm is proposed based on the genetic theory, which has global optimum searching characteristic. Meanwhile, considering its computational cost, decimal code, not binary one is used in operation of the operators. Moreover, a modified adaptive weights training algorithm is also used to resolve the weights fluctuant problem in stable range. Numerical simulations demonstrate that the genetic algorithm can converge at the global extremum quickly, and that weights estimation with the modified adaptive weights training algorithm have a more stable range, and the filtering performance has been improved obviously.