Model-aided Deep Learning-Based Target Detection for Channel Matrix-Based Cognitive Radar/sonar

Data driven based approaches to signal processing including deep neural networks (DNN) have shown promise in various fields. Such techniques tend to require significant training for good convergence. Model-based approaches, however, provide practical solutions often with insightful and intuitive interpretations. A hybrid approach that employs data driven techniques aided by knowledge from model-based approaches may help reduce required training and improve convergence rates. This work investigates the potential of deep learning techniques to detect targets while accelerating the learning process via use of expert/domain knowledge for the channel matrix based cognitive radar/sonar framework. The channel matrices characterize responses from target and clutter/reverberation. We will leverage our previous work [Ali et al. (2021); (2022)] on model-based adaptive detection approaches for cognitive radar/sonar. We compare the detection performance of model aided deep learning-based algorithms with that of traditional model-based techniques using receiver operating characteristic (ROC) curves from Monte Carlo simulations. We also study and compare the robustness of these techniques by changing the signal-to-interference plus noise ratio (SINR), the number of targets and clutter sources, and the amount of available training data. The effects of choice of hyperparameters and loss functions are also studied.