Channel Quality Prediction Based On A Hidden Markov Model

Spectrum prediction is a key technology in cognitive radio, which analyzes the law of the spectrum through the history information of each channel, to find out the characteristic of the users’utilization and occurrences of the spectrum hole in spectrum. According to the result of the spectrum prediction, cognitive uses can access the optimal channels beforehand and evacuate the channels before primary users could utilize. Meanwhile accurate spectrum prediction can aid in the procedure of channel allocation to decrease the probability of the false sensing, diminish the interference and delay actively, promote the efficiency of spectrum usage and the throughput of the whole network.Focused on the problem of the first-order hidden Markov model that it cannot take full advantage of the history sequence to improve accuracy, in this paper, the traditional first-order hidden Markov model has been transformed to get the high-order transmission probability and emission probability. However most current literatures researched on the high-order hidden Markov model only applies to the assumption that the durations of channel state are exponentially distributed, which has been validated to be inadequate in nature. Our proposed high-order prediction algorithm is suitable for all kinds of distributions of the channel state, which maintains the performance of prediction while the distribution of channel state is not exponential necessarily. Furthermore, based on the result of prediction, a channel quality metric involving the sensing accuracy and the predicted channel idle probability is defined, which can facilitate SUs to select better channels to get accessed. Finally, the performances of our strategy have been proved to be valid under different circumstances.