In the recently proposed collaborative compressive sensing, the cognitive radios (CRs) sense the occupied spectrum channels by measuring linear combinations of channel powers, which is more efficient than sweeping a set of channels sequentially. The measurements are reported to the fusion center, where the occupied channels are recovered by compressive sensing algorithms. In this paper, we study a method of dynamic compressive sensing, which continuously measures channel powers and recovers the occupied channels in a dynamic environment. While standard compressive sensing algorithms must recover multiple occupied channels, a dynamic algorithm only needs to recover the recent change, which is either a newly occupied channel or a released one. On the other hand, the dynamic algorithm must recover the change just in time. Therefore, we propose a least-squares based algorithm, which is equivalent to ℓ0 minimization. We demonstrate its fast speed and robustness to noise. Simulation results demonstrate effectiveness of the proposed scheme.
展开▼
