基于深度卷积网络的二维波达方向估计方法OA北大核心CSTPCD

To improve the real-time and convenient feature of the direction of arrival(DOA)estimation technology,a deep convolution network(DCN)is proposed.The covariance matrix of received signal obtained from the uniformed circular array(UCA)is regarded as an image which contains the real part channel and imaginary part channel.By using the covariance matrix as the input tensor of the convolutional network,it is possible to extract the subtle feature of the DOA implied in the covariance matrix,therefore,the DOA information of multi-signal can be estimated quickly and precisely.The simulation results show that the proposed DCN can achieve the DOA estimation of two dimensional signals well.Compared with the traditional method based on the sub-space calculation,the proposed network can obtain more accurate result and the algorithm is relatively stable,therefore,the network has some potential applications in engineering.