Cognitive Radio Spectrum E volution Prediction using A rtificial Neural Networks based Multivariate T ime Series Modelling

Cognitive Radio (CR) is an efficient answer to spectrum scarcity as it can sense the spectrum steadily based on previous information about the spectrum evolution in time, thus predicting the future occupancy status. Framed within this statement, this paper proposes a new methodology for spectrum prediction by modelling licensed signal Radio F requency (RF) features as a multivariate chaotic time series, which is then given as input to Artificial Neural Network (ANN), that predicts the evolution of RF time series to decide if the unlicensed user can exploit the spectrum band. We exploit the inherent cyclostationarity in primary signals for Non-linear Autoregressive Exogenous T ime Series Modelling of RF features, which is an extremely challenging task due to interdependence of different R F features. Experimental results show a similar trend between predicted and observed values. We target this work at cognition incorporation in our designed Software Defined Radio (SDR) waveform.