A SPECTRUM PREDICTION METHOD BASED ON TWO-DIMENSIONAL HIDDEN MARKOV MODEL FOR UAV COMMUNICATIONS

With the development of unmanned aerial vehicle, the spectrum resources for communication become increasingly scarce. Spectrum sharing can alleviate this problem effectively, and spectrum prediction is the key step. The previous spectrum prediction models only consider the time domain variation and ignore the time-frequency correlation. In order to change this defect, we propose a new spectrum prediction model, which extends the traditional one-dimensional HMM model to the two dimensions of time-frequency. In addition, homotopy method is introduced to solve the strict dependence on historical spectrum data during UAV flight. Through simulation experiments, we verify that the new two-dimensional model has better prediction effect than the one-dimensional model, and has a good performance in the absence of sufficient prior information.