A study on channel prediction for dynamic spectrum control aided spectrum sharing

This paper proposes a channel prediction scheme suitable for selfish dynamic spectrum control (Selfish DSC) in which each link dynamically chooses a necessary amount of spectrum having the highest signal to noise plus interference power ratio (SINR) from the entire shared band. For channel prediction in such a Selfish DSC, although it is mandatory to predict channel frequency response for the entire system band to choose the best spectrum for the next transmission, it is quite difficult because each link can measure only a part of the band. In other words, each user has to solve an ill-conditioned least square (LS) problem. Thus, the proposed scheme chooses a certain amount (which is smaller than the total amount) of spectrum with the highest SINR, followed by the selection of the rest of spectrum uniformly from the other part to mitigate the impact of ill-conditioned LS problem. Then data and reference signals are transmitted via the selected spectrum. Numerical analysis reveals that the proposed scheme can achieve satisfactory average channel capacity due to a good balance between frequency diversity effect by the dynamic spectrum selection and enhanced reliability by the selection of a certain amount of spectrum uniformly from the entire system band.