Performance Analysis of Prediction Based Spectrum Sensing for Cognitive Radio Networks

Cognitive Radio (CR) was anticipated as a solution to the severe concern in wireless communication i.e. scarcity of accessible spectrum. Spectrum sensing is considered to be the most substantial part in CR. Due to the trade-off between spectrum sensing and throughput of the CR network, the licensed users’ transmission activity prediction is considered as a potential alternative to spectrum sensing. The present work considers a neural network based multilayer perceptron model to predict the availability of licensed spectrum. Performance of this model is evaluated in different traffic load. Stand-alone prediction model and prediction before sensing model, both are analysed with respect to sensing parameters, false alarm and misdetection probability. MATLAB software is used for simulation.