Blind and adaptive reconstruction approach for non-uniformly sampled wideband signal

Abstract

In 5G, wideband communication receivers (WCRs) capable of digitizing signal ranging from 400 MHz to few GHz are desired to support various data intensive services. The design of such WCRs is a challenging task due to huge area, high cost, limited speed and dynamic range of analog-to-digital converters (ADCs) as well as poor reconfigurability of analog front-end. Recently, non-uniform (or sub-Nyquist) sampling techniques have been envisioned to digitize sparse wideband signal using existing ADCs. However, subsequent digital reconstruction works well only when the number of active users in the received signal are known i.e. they are not blind. To overcome this, a new Adaptive Orthogonal Matching Pursuit (AOMP) blind reconstruction approach has been proposed in this paper. The term adaptive means that the parameters of the AOMP are dynamically tuned based on learned spectral occupancy and channel quality statistics. The novelty of the AOMP is the use of online learning algorithm to estimate spectrum occupancy (or sparsity). Extensive simulation and complexity results indicate the superiority of the proposed AOMP approach over existing approaches for wide range of SNRs and different levels of sparsity. Numerically, AOMP offers as high as 64.5% improvement in normalized mean square error over existing approaches with slight penalty in terms of computational complexity. In the end, performance comparison of various reconstruction approaches for automatic modulation classification application is presented.