Classification-Based one-Bit DOA Estimation for Sparse Arrays

Nowadays, high-resolution DOA estimation techniques have been widely used in many fields such as sonar, communication, radio astronomy and biomedicine etc. Unfortunately, conventional algorithms cannot achieve high precision due to the limitation of the array aperture, and usually suffer from high computational complexity. In this paper, we innovatively model the DOA estimation problem of incoherent signals as a binary classification problem and greatly reduce physical complexity by virtue of sparse arrays. We first propose a classification framework for DOA estimation by benefiting from one-bit quantification. In this framework, any classification algorithm can be exploited to estimate DOA, such as logistic regression used in this paper. And then, sparse array is considered to reduce the exceeding number of antennas. Moreover, an iterative grid refinement procedure is presented to achieve more accurate DOA estimation and further reduces the actual number of physical antennas. Simultaneously, we derive Cramer-Rao bound (CRB) for the proposed algorithm. Finally, simulations are conducted for correctness and validation and the results illustrate the significant performance and reduction of complexity in hardware and computation over the existing methods.