A Study of Angle of Arrival Estimation of a RF Signal with FPGA Acceleration

This paper presents a comprehensive study on angle of arrival (AoA) estimation techniques for wireless communication systems. In particular, signal processing techniques including Multiple Signal Classification (MUSIC) and Estimation of Signal Parameters via Rotational Invariance Techniques (ESPRIT), as well as machine learning (ML) techniques such as artificial neural networks (ANNs) and convolutional neural networks (CNNs), are evaluated. The performance of these techniques is compared using both synthetic and over-the-air (OTA) test scenarios. Additionally, the impact of multipath fading on the performance of these techniques is investigated. Experimental results show that while the MUSIC algorithm exhibits superior accuracy in synthetic scenarios, it suffers from performance degradation in the presence of multipath fading. On the other hand, ML algorithms demonstrate robustness and stability under varying conditions, albeit with a slightly higher error rate compared to simulations. Furthermore, suggestions for improving the ML algorithm in OTA scenarios are discussed, including updates to the synthesis model and the incorporation of OTA samples for training.