Joint DoA Estimation and AMC for Overlapped Signal via Multi-Task Deep Learning

In this paper, we propose a deep learning (DL)-based joint direction of arrival (DoA) estimation and automatic modulation classification algorithm to identify overlapped signals. The proposed algorithm utilizes independent component analysis for signal separation to estimate two characteristics of each signal within an overlapped signal. Each separate signal is processed by MoDANet, a multi-task deep learning model for joint characteristic estimation. The model estimates DoA and modulation type and maps these characteristics to the appropriate signal. The simulation results show that the proposed algorithm performs classification and estimation accurately.