This paper proposed a source number estimation algorithm based on support vector machine (SVM) for improving estimation performance in low ratio of signal to noise, small snapshots and color noise environment. This algorithm decomposes the received signal data covariance matrix to get the signal vector and the noise vector at first, and then abstracts the property sort of signal and noise with Gerschgorin circle algorithm by the orthotropic property of the antenna array manifold and the noise vectors. At last, the algorithm constructs and trains the SVM to get the number of the incidence signals. The paper compares the algorithm with some classic algorithms by simulation experiment in low ratio of signal to noise, small snapshots and color noise environment, and results show that the algorithm has higher precision than that of classic algorithms.%为了改善在低信噪比、小快拍、色噪声环境下盖氏圆准则信源数估计算法的估计性能,提出了基于支撑矢量机(SVM)的信源数估计算法.基于支撑矢量机的信源数估计算法应用天线阵列接收数据协方差矩阵经特征值分解后,噪声的特征矢量与天线阵列的阵列流型正交的特性,通过盖氏圆算法提取信号和噪声的分类特征,再构造和训练两类分类矢量机,将天线阵列接收的数据分为信号子空间和噪声子空间.通过仿真实验比较了本算法与其他经典算法在低信噪比、小快拍、色噪声环境下的信源数估计性能,结果证明本算法对信源数的估计精确度要高于其他经典算法.
展开▼
