Applications of Compressed Sensing in Modern Wireless Communication Systems

Abstract

This thesis presents research contributions to the developing field of Compressed Sensing (CS) and its applications in wireless communications. The novel approach of CS has opened new venues to tackle wireless communication challenges. Peculiar nature of problems while communicating through wireless medium provides opportunities for CS based solutions. In the first contribution, a detailed survey on sensing matrices and wireless applications specific recommendation is provided. An analysis of its main characteristics and sparse signal recovery guarantees based on these properties is provided. Moreover, a qualitative comparison of sensing matrices in real and complex generated through different techniques is carried out. Furthermore, an analysis based on their application specific desirable features is performed. Algorithms for coherence optimization of Rank-1 Grassmannian codebooks are provided in the second contribution. The contribution reduces processing time for state of the art algorithms, namely Best Complex Antipodal Spherical Code (BCASC) and Coherence Based Grassmannian Codebook (CBGC) for generating deterministic sensing matrices achieving coherence lower bound. The trimming of runtime is performed by preventing the algorithms to remain stagnant or divergent during the process. The proposed modifications preserve the low coherence quality of the generated matrices in orders of magnitude lesser time. Reduction in processing time allows generation of larger codebooks which are required in many wireless applications. The third contribution provides coherence optimized channel estimation for mm-wave massive multiple input multiple output (MIMO) communication. Millimeterwave (mm-wave) communication is inherently sparse due to its peculiar propagation characteristics with limited diffraction, penetration and small number of scatterers. Based on these characteristics, a 2-dimensional non-uniform quantized azimuth and elevation angle grid antenna array response is proposed. Later coherence minimized training vectors generation algorithm is proposed by extending the CBGC algorithm. The proposed training vectors minimize the coherence with respect to the antenna array response. Open-loop channel estimation of the mm-wave channel performed by the proposed method improves normalized mean squared error performance and spectral efficiency in comparison to the existing techniques. Frequency hopping (FH) radio network identification in a wideband using CS is performed in the fourth contribution. Radio networks operating in overlapping wideband spectrum are sensed using an established CS technique of multi-coset sampling. The acquired features are then used to isolate and differentiate between different radio networks based on their respective attributes.