一种基于支持向量机的目标定位方法

为了提高声纳在浅水域的性能，提出了一种基于统计学习理论的目标识别器的目标定位方法.该方法选择支持向量机（SVM）作为学习算法的核心.从已知训练样本得到多通道数据的协方差矩阵，将得到的矩阵转化为SVM的输入多维特征向量，并训练SVM而获得权向量.利用此权向量和SVM输出估计，可以得到目标位置信息.理论推导和仿真结果表明，与多重信号分类（MUSIC）算法相比较，该方法具有高的定位精度和快的收敛速度.该方法能有效地对在平面波模型下的目标进行测向，并具有鲁棒性.

Support vector machine-based algorithm for target localization

To improve sonar performance in shallow water,a novel method for target localization based on a statistical learning theory's recognizer was proposed.The support vector machine(SVM) was selected as the kernel of learning algorithm in the algorithm.After the covariance matrix of the multi-channel data was obtained by the known training samples,the matrix was transformed into multidimension feature vectors as the input of the SVM and the weight vector was computed by training SVM.The target position information was acquired by the SVM estimation output and the weight vector.The theoretical deduction and experimental result reveal that compared with the multiple signal classification(MUSIC) algorithm,the proposed method has the higher location precision and the faster convergence speed.The algorithm can effectively estimate the target's bearing and have good robustness by the model of plane wave.