DOA Estimation for 6G Communication Systems
Haya Al Kassir, Ioannis T. Rekanos, Pavlos I. Lazaridis, Traianos V. Yioultsis, Nikolaos V. Kantartzis, Christos S. Antonopoulos, George K. Karagiannidis, Zaharias D. Zaharis · 2023
The objective of this study is to analyze and compare different neural network (NN) architectures as multi-class classifiers to estimate the direction of arrival (DOA) using a uniform linear array (ULA). The study specifically investigates the prediction skills of three NNs: feed forward NN (FFNN), convolutional NN (CNN), and residual NN (ResNet) when estimating incoming signal DOAs in a realistic ULA of (M = 16) elements under noisy conditions. The NNs are trained on a correlation matrix generated by a ULA to estimate the DOAs. The results of the simulations indicate that ResNet performs better than FFNN and CNN in accurately estimating incoming signals.