非均匀噪声下基于BOOTSTRAP和特征空间投影的信源数估计

Source detection based on characteristic subspace projection and Bootstrap technique in nonuniform noise

针对阵列系统接收数据有限且存在空间非均匀噪声的信源数估计问题,提出一种基于Bootstrap和特征子空间投影的新方法。该方法将阵列信号的协方差矩阵分别投影到信号的特征子空间和噪声的特征子空间,提出了以特征空间投影为基础的一系列假设检验。数据分布未知,应用Bootstrap技术估计零假设下检测统计量的分布。通过仿真证明了所提方法在信源等功率、不等功率以及噪声功率不等的情况下有较好的性能,尤其在低信噪比和小快拍数的情况下有较好的性能。

Considering the problem of source number estimation in the presence of unknown spatially nonumiform noise and the limitation of the received data,a now method based on Bootstrap technique and characteristic subspace projection is proposed.The estimation projects covariance matrix estimate of array signal into signal eigen subspace and noise eigen subspace respectively.A sequential hypothesis test is formulated based on characteristic subspace projection.No assumption is made in the distribution of the data and the Bootstrap is used to estimate the distribution under the null of the proposed test statistics.Simulation results show that the proposed method has good performances when the signal power is equal or different and under the condition of space-nonuniform noise environment,especially in the low signal-to-noise ratio and small snapshots.