Decentralized Expected Consistent Signal Recovery For Quantization Measurements

Signal recovery through coarse quantization of a linear transform output has many applications in engineering, such as channel estimation and signal detection in massive MIMO systems. A recently proposed scheme, known as generalized expectation consistent signal recovery (GEC-SR), can achieve Bayesian inference and exhibit better robustness than many existing methods. However, recovering signals with large transform matrices continue to present a computational burden for GEC-SR. In this study, we develop a novel decentralized architecture by leveraging the core framework of GEC-SR called “deGEC-SR.” deGEC-SR offers excellent performance as GEC-SR and runs tens of times faster than GEC-SR. We derive the theoretical state evolution of deGEC-SR and demonstrate its accuracy using numerical results.