Signal Source Location Algorithm Based on Deep Learning and MVDR

In this paper, we introduce an azimuth estimation algorithm that utilizes deep learning and deconvolution. The algorithm combines traditional azimuth estimation techniques for feature extraction, integrating Minimum Variance Lossless Response (MVDR) and deconvolution algorithms to extract azimuth features. The covariance matrix is then processed through an encoder and convolution network to perform error correction. Subsequently, the MVDR output and deconvolution output undergo deep learning processing. Finally, the modified data is aggregated with the deconvolution results. Experimental results illustrate that the proposed method attains high resolution and accuracy.