Spectrum Hole Prediction Based On Historical Data: A Neural Network Approach
This work addresses spectrum underutilization in wireless communication by enabling opportunistic sharing without interference, though it appears incremental as it combines existing neural network and genetic algorithm methods.
The paper tackles the problem of predicting spectrum holes (idle channels) in cognitive radio networks using historical occupancy data, achieving high prediction accuracy across all frequency bands considered.
The concept of cognitive radio pioneered by Mitola promises to change the future of wireless communication especially in the area of spectrum management. Currently, the command and control strategy employed in spectrum assignment is too rigid and needs to be reviewed. Recent studies have shown that assigned spectrum is underutilized spectrally and temporally. Cognitive radio provides a viable solution whereby licensed users can share the spectrum with unlicensed users opportunistically without causing interference. Unlicensed users must be able to sense weather the channel is busy or idle, failure to do so will lead to interference to the licensed user. In this paper, a neural network based prediction model for predicting the channel status using historical data obtained during a spectrum occupancy measurement is presented. Genetic algorithm is combined with LM BP for increasing the probability of obtaining the best weights thus optimizing the network. The results obtained indicate high prediction accuracy over all bands considered