Deep Neural Based Beamforming Techniques for Direction of Arrival (DOA) Estimation

Most of the applications with SONAR are relied on the Direction of Arrival (DOA) estimation. It is an inevitable part in the scenarios of target detection and localization where the location of target is retrieved from the received sensor data. The paper models novel deep learning architectures for DOA es-timation and compares their performance with the existing DOA estimation algorithms like conventional beamforming, MVDR beamforming and MMSE estimation. The deep learning frame-works like Deep Neural Network (DNN) and Deep Convolutional Neural Network (DCNN) are deployed to recover the signal radiated from passive targets using multiple sensor arrays. The estimation is carried out for the SNR conditions ranging from −5dB to 20dB and the results are analyzed using MSE v/s SNR plot, waterfall diagrams and the time taken for execution.