Abstract:

Spectrum prediction is one of the key technologies in cognitive radio（CR）systems. This technology can reduce considerable energy consumed by spectrum sensing, and improve the overall system's spectrum utilization. Aiming at the low accuracy and invalid prediction problems of spectrum prediction in cognitive radio, a new prediction method was proposed by integrating support vector machine（SVM）which was based on statistical learning theory（SLT）and structural risk minimization principle（SRM）. The channel status is forecasted by utilizing the excellent forecasting performance of the model in small sample and nonlinear data of SVM. The results show that by avoiding invalid prediction, the spectrum utilization can also be improved, and the forecasting accuracy is better than model based on back propagation（BP）, thus the proposed algorithm is practicable and flexible for spectrum prediction in cognitive radio.

Key words: cognitive radio, spectrum prediction, support vector machine