Crowdsensing Based Spectrum Database with Total Variation Regularization

A concept of crowdsensing based spectrum database has a great deal of attention due to their ability to support largescale and low-cost spectrum monitoring by collecting spectrum data from a swarm of user devices. However, for accurate spectrum database construction, we still have the problem caused by crowdsensing, i.e., the low measurement accuracy of user devices and the measurement noise due to the multipath fading effect. As a remedy for the measurement issue, we develop a novel database reconstruction problem, referred to as TVR+, based on a path loss estimation and total variation regularization (TVR). Further, we propose an iterative algorithm with Split Bregman method to solve the TVR+ problem. Extensive simulations demonstrate that the reconstructed database with TVR+ significantly improves the accuracy.