Reinforced HMM based learning model of dynamic spectrum allocation in UHF - ISM band of 902–928 MHz for Cognitive radio

A reinforced learning model based on HMM following urn and ball concept is suggested in this paper. Earlier works on HMM for Cognitive radio is towards predicting the channel state based on the activity of the primary user and on occupying the next available channel based on the power spectral density measurements. This work applies HMM techniques on real time spectral measurements and reinforces initial training based on the transition probabilities calculated from the measured power levels with the learning methodology based on the survival duration and packet loss data. The model is validated over the spectral measurements taken in the UHF-ISM band 902-928 MHz and shown to improve the reliability of the link decreasing the setting up delay of the link by selecting the best possible channel based on the interference level and the data on survival duration. The model is also analyzed with increased occupancy of the spectrum as more and more emerging users start occupying the band.