In the complex electromagnetic environment, the traditional direction of arrival(DOA) estimating technology is restricted by multipath effect, interference and other factors, and the accuracy of DOA estimation is facing challenges. This paper focuses on the DOA estimating algorithm of emitter array based on deep learning, and explores a new path to improve the performance of direction finding. In this paper, a deep learning model based on multi-label classification is constructed. The continuous Angle space was discretized into 181 grid points, the features of the received signal covariance matrix were extracted by Convolutional Neural Network (CNN), and the Direction of Arrival classification prediction was realized by combining the binary cross-entropy loss function. Experimental results show that the accuracy of the model can reach 0.97 and the Root Mean Square Error (RMSE) is stable at about 0.23 under a specific SNR, which has certain direction finding ability. However, the generalization ability of the data with SNR deviation from the training condition is limited.