基于改进蜂群算法的数字信号调制识别

A Modulation Identification Algorithm for Digital Signals Based on Modified Artificial Bee Colony Algorithm

针对传统人工蜂群(ABC)算法初始种群在解空间分布不均匀、收敛速度慢等缺点,文中提出一种基于二维均匀设计和欧氏距离的改进蜂群算法。改进蜂群算法在构造初始食物源时采用二维均匀设计使食物源在解空间均匀分布,提高了算法的全局搜索能力;在构造新食物源时采用欧氏距离法提高了算法的寻优效率。文中利用信号二阶以上累积量可以抑制噪声影响的特性,从二阶、四阶和六阶累积量中提取四个特征参数作为特征向量,采用支持向量机分类器,并用改进蜂群算法对支持向量机的惩罚因子和核函数参数进行优化,实现了2FSK、BPSK、QPSK、16QAM、64QAM五种调制方式的分类识别。仿真结果表明,改进蜂群算法具有更快的收敛速度,且改进ABC-SVM方法在信噪比-3 d B时具有更好的识别效果,平均识别率为92.9%;当信噪比超过4 d B时,改进ABC-SVM方法平均识别率达到99%。

In viewof the slowconvergence speed and non-uniform distribution of the initial food source of traditional Artificial Bee Colony（ ABC） algorithm,a modified ABC algorithm based on two-dimensional uniform design and Euclidean-distance has been proposed.Two-dimensional uniform design is used to make the food source uniformly distribute in the solution space when the modified ABC algorithm establishes the food source,which can improve the global search ability of the algorithm. Euclidean-distance is applied in constructing newfood source to improve the efficiency optimization. In this paper,four feature parameters which are picked up from second-order,fourth-order and sixth-order cumulants are obtained as a feature vector because second and higher order cumulants can suppress additive white Gaussian noise. SVMclassifier and modified ABC algorithm is used to optimize the penalty factor and kernel function parameter,realizing the identification of 2FSK,BPSK,QPSK,16 QAMand 64 QAM. The simulation results showthat the convergence speed of modified ABC algorithm is improved and the average recognition rate is 92. 9% when SNR is-3 d B,as well as over 99% when SNR is more than 4 d B.