SPECTRUM SENSING AND SIGNAL CLASSIFICATION IN COGNITIVE RADIO NETWORKS USING SUPPORT VECTOR MACHINE

Abstract

The need for wireless communication applications is increasing and the available electromagnetic spectrum band is getting crowded. Spectrum sensing helps to detect the spectrum holes (underutilized bands of the spectrum) in providing high spectral resolution capability. However the challenges being presently researched are devising the effective technique that can detect all forms of primary radios signals present in the cognitive radio environment and automatically classify the signal modulation formats. Spectrum occupancy spanning across the frequency range of 100 MHz – 2400 MHz is sensed at different times using Aaronia AG Spectran HF-6065 spectrum analyser and HyperLOG 7060 antenna to monitor the energy level of the received signal at different times according to traffic intensity. Key features of the received signal which are sensitive with modulation type and insensitive with Signal to Noise Ratio (SNR) variation were extracted from the dataset in order to obtain input features for the multi-class Support Vectors Machine (SVM) classifier that is constructed via one-against-all method. The key features extracted are derived from the instantaneous phase, amplitude and frequency. Different kernels of SVMs were compared using standard performance metrics. The models developed were simulated on a Computer running on Windows 10 platform and MATLAB R2019a and Python 3.7 programming language. Simulation results show that the radial basis function (RBF) kernel with grid search has better performance than other kernels. The RBF kernel with grid search technique showed accuracy, precision, recall and F1-Score of 0.947, 0.950, 0.950 and 0.950 respectively at 0 dB. The simulation was carried out over a broad range of 0 – 20 dB. The developed classifier has very high accuracy for digital signal classification even at 0 dB SNR. The developed system provides a promising solution and cost-effective signal detection probability and classification in cognitive radio networks.