【摘要】 针对粒度母体混合分布识别中参数优化求解问题,为进一步提高识别效率,利用一种改进的微粒群算法对粒度母体混合分布的参数进行优化。方法通过设置检验值,判断算法是否陷入局部最优解,并让陷入局部最优的粒子进入下一次迭代,避免微粒群算法在搜索过程中陷入局部最优的缺陷问题。在仿真实验部分,将方法估计的高斯混合模型的参数与迭代EM算法估计的模型参数做比较,结果表明,得到的模型参数接近真实的分布,使得粒度母体混合分布的识别率进一步提高。更多还原

【Abstract】 To solve the problem of parameter optimization of granularity mother mixed distribution,This article used an improved particle swam optimization algorithm.It is tested by setting the check to determine whether the algorithm get into a local optimal solution,and the particles of a local optimum access to the next iteration,which avoids particle swarm optimization the shortcomings of local optimum in the search process.By comparing the parameters of Gaussian mixture model of this method with the parameters based on EM algorithm,the simulation show that this mothed of the model parameters is closer to the true distribution,and it improves the recognition of the mixed distribution of granularity mother.更多还原