Gridless Variational Line Spectral Estimation with Multiple Measurement Vector from Quantized Samples
Jiang Zhu, Qi Zhang, Benzhou Jin, Zhiwei Xu · arXiv (Cornell University) · 2019
Utilizing multisnapshot quantized data in line spectral estimation (LSE) for improving the estimation accuracy is of vital importance in signal processing, e.g., channel estimation in energy efficient massive MIMO systems and direction of arrival estimation. Recently, gridless variational line spectral estimation (VALSE) treating frequencies as random variables has been proposed. VALSE has the advantage of low computation complexity, high accuracy, automatically estimating the model order and noise variance. In this paper, we utilize expectation propagation (EP) to develop multi snapshot VALSE-EP (MVALSE-EP) to deal with the LSE from multisnapshot quantized data. The basic idea of MVALSE-EP is to iteratively approximate the quantized model as a sequence of simple multiple pseudo unquantized models sharing the same frequency profile, where the noise in each pseudo linear model is i.i.d. and heteroscedastic (different components having different variance). Moreover, the Cramer Rao bound (CRB) is derived as a benchmark performance of the proposed algorithm. Finally, numerical results demonstrate the effectiveness of MVALSE-EP, in particular for the application of direction of arrival (DOA) problems.