Multiple-input multiple-output (MIMO) systems have recently emerged as one of the most significant wireless techniques, as they can greatly improve the channel capacity and link reliability of wireless communications. These benefits have encouraged extensive research on a virtual MIMO system where the transmitter has multiple antennas and each of the receivers has a single antenna. Single-antenna receivers can work together to form a virtual antenna array and reap some performance benefits of MIMO systems. The idea of receiver-side local cooperation is attractive for wireless networks since a wireless receiver may not have multiple antennas due to size and cost limitations. In this thesis we investigate a virtual-MIMO wireless system using the receiver-side cooperation with the compress-and-forward (CF) protocol. Firstly, to perform CF at the relay, we propose to use standard source coding techniques, based on the analysis of its expected rate bound and the tightness of the bound. We state upper bounds on the system error probabilities over block fading channels. With sufficient source coding rates, the cooperation of the receivers enables the virtual-MIMO system to achieve almost ideal MIMO performance. A comparison of ideal and non-ideal conference links within the receiver group is also investigated. Considering the short-range communication and using a channel-aware adaptive CF scheme, the impact of the non-ideal cooperation link is too slight to impair the system performance significantly. It is also evident that the practicality of CF cooperation will be greatly enhanced if a efficient source coding technique can be used at the relay. It is even more desirable that CF cooperation should not be unduly sensitive to carrier frequency offsets (CFOs). Thus this thesis then presents a practical study of these two issues. Codebook designs of the Voronoi VQ and the tree-structure vector quantization (TSVQ) to enable CF cooperation at the relay are firstly described. A comparison in terms of the codebook design complexity and encoding complexity is presented. It is shown that the TSVQ is much simpler to design and operate, and can achieve a favourable performance-complexity tradeoff. We then demonstrate that CFO can lead to significant performance degradation for the virtual MIMO system. To overcome it, it is proposed to maintain clock synchronization and jointly estimate the CFO between the relay and the destination. This approach is shown to provide a significant performance improvement. Finally, we extend the study to the minimum mean square error (MMSE) detection, as it has a lower complexity compared to maximum likelihood (ML) detection. A closed-form upper bound for the system error probability is derived, based on which we prove that the smallest singular value of the cooperative channel matrix determines the system error performance. Accordingly, an adaptive modulation and cooperation scheme is proposed, which uses the smallest singular value as the threshold strategy. Depending on the instantaneous channel conditions, the system could therefore adapt to choose a suitable modulation type for transmission and an appropriate quantization rate to perform CF cooperation. The adaptive modulation and cooperation scheme not only enables the system to achieve comparable performance to the case with fixed quantization rates, but also eliminates unnecessary complexity for quantization operations and conference link communication.