Direction Finding in Time Modulated Linear Arrays Using Supervised Machine Learning

In Time Modulated Arrays (TMAs), the spectrum characteristics of the time modulated signal vary with each incident angle. In this paper, a novel idea of the feasibility of Machine Learning (ML) approach for the Direction of Arrival (DoA) estimation in a two-element Time Modulated Linear Array (TMLA) is proposed. The Fourier components of the received TMLA signal are used as the input data for training. The DoA estimation of signal from a single source is formulated as a multiclassification problem with single label. The performance of the classification based method is tested under various SNR conditions. The regression based method for estimating off-grid angles is also studied.