Direction of Arrival Estimation Based on Machine Learning

The accuracy of DOA estimation can be affected by scenarios with unknown target numbers and low-precision quantization processing. This study centers on crafting a scheme for DOA estimation through low-precision ADC quantization signals, leveraging machine learning techniques in scenarios where the number of targets is unknown. We deploy a framework based on Deep Neural Network (DNN) to restore the low-precision quantized signal to its high-precision counterpart received signal, and subsequently employ a Convolutional Neural Network (CNN) model to estimate the DOA of the restored signal. To attain precise DOA estimation when the target number is unknown, this study formulates the DOA estimation problem as a classification task. Specifically, the neural network provides DOA estimation results ranging from 0 to 180 degrees. The simulation results demonstrate that the proposed algorithm exhibits high accuracy in estimating the direction of arrival (DOA) across varying quantization levels, numbers of antennas, and signal-to-noise ratios (SNRs).