CS-AdaBoost: One-Bit Compressed Sensing for Direction-of-Arrival Estimation

Using One-bit analog-to-digital converters (ADCs) instead of high-precision counterparts for direction-of-arrival (DOA) estimation is a promising alternative to considerably reduce power consumption and manufacturing transceiver costs. However, these benefits come at the cost of information loss, as one-bit ADCs retain only the sign of the signals and discard the amplitude information. Consequently, developing customized one-bit DOA estimation methods is necessary. In this work, we propose a learning-based one-bit DOA estimator referred to as compressed sensing adaptive boosting (CS-AdaBoost). The proposed method employs a two-stage weak classifier within AdaBoost framework. It first discretizes the DOA angular interval and builds an overcomplete dictionary for the array steering matrix and then estimates the corresponding source signal matrix in an iterative manner. In each AdaBoost iteration, an approximate weighted least <tex>$\ell_{2}$</tex>-norm estimation is used as the first stage, followed by hardthresholding for imposing sparsity at the second stage. Numerical simulations demonstrate the superiority of the CS-AdaBoost method over other existing methods, especially in the face of closely spaced and correlated sources.