An Improved Neural Network for Wideband Doa Estimates
基于神经网络的宽带相干DOA估计方法的改进

Coherent signal-subspace method (CSM) is an efficient method for wideband DOA estimates based on constructing focusing matrices. In this paper, a new BP neural network is proposed to complete the contruction procedure. The proposed network has much less neutrons in the input and output layers than before. The leraning procedure and convergence procedure are also accelerated. Further, the network is generlized to the case that the number of DOA's is unknown.