基于改进LS-SVM的来波方位估计

Incoming Wave Direction Estimation Based on Improved LS-SVM

提取已知方位信号的协方差矩阵的上三角部分作为样本特征,构建方位估计模型。针对最小二乘支持向量机最优参数难以选定的问题,采用实值编码的启发式遗传算法,以模型的来波方位估计性能为目标,实现基于高斯核函数的SVM参数优化,提高了来波方位估计精度。实验结果表明,该方法估计精度较高、实时性好,在无线电测向领域具有广阔应用前景。

This paper extracts the upper triangular half of the covariance matrix of knowing direction signals to construct the direction estimation model. Aiming at the problem that the best parameter of Least Squares-Support Vector Machine（LS-SVM） is hard to select, it uses real-coded heuristic genetic algorithm. Aiming at the approximate performance estimation of model incoming wave direction, it optimizes the parameters of LS-SVM with Gauss kernel function. The estimation precision is improved. Experimental results show that this method has high precision, high real-time performance and a broad application future in wireless direction finding field.