Fast and reliable sensing using a background process for noise estimation

This paper presents an efficient way to ensure a good detection performance by implementing algorithms running in background a reliable noise estimation process. The proposed solution operates at two different time scales: a slow time scale to determine in adjacent sub-bands the supposed slowly varying noise level, and a faster time scale to determine in the band of interest the presence of signal, using a reliable energy detection solution. In order to identify the free bands where the noise variance can be estimated, the paper describes several blind and semi-blind strategies based on the statistical properties of the received signal. One of the benefits of the proposed solution is that the output of the described algorithm can populate the database of free/occupied bands, which classically needs to be regularly updated in a cognitive radio architecture.