基于神经网络方法的窄带信号源波达方向估计

Narrowband Signal Source DOA Estimation Based on the Neural Network Method

文章研究了对窄带信号源进行神经网络方法的高分辨率测向，常规的高分辨波达方向估计都必须得到阵列输出的协方差矩阵，然后进行特征值分解，显然的缺点是处理时间比较长，我们采用Hopfield神经网络模型，提出了基于一些数据快拍点全互连对称突出权值和阈值的新方法，以此来提高波达方向估计的性能。该方法具有小运算量、低复杂度和大规模并行计算的特点，模拟仿真结果显示文章所提方法是实现信号源DOA实时处理的一条有效途径。

The high-resolution DOA estimation on the narrowband signal source based on the neural network method is studied in this paper. The conventional high-resolution DOA must be estimated by output array covariance matrix, and then eigen-decomposition, clearly that the processing time is relatively too long is the shortcoming of this method. Using Hopfield neural network model, a new method based on some data snapshot and whole point of interconnection symmetric outstanding value and the threshold value is advanced to improve the performance of the DOA estimation. This method has the characteristics such as less calculation mnount, low complexity and large scale parallel calculation. Simulation results show this method is an effective way to complete the real-time processing on signal source DOA.