Signal Classification Based on Spectral Correlation Analysis and SVM in Cognitive Radio

Spectrum sensing is one of the most challenging functions in cognitive radio system. Detection of the presence of signals and distinction of the type of signals in a particular frequency band are critical for cognitive radios to adapt to the radio environment. In this paper, a novel approach to signal classification combining spectral correlation analysis and support vector machine (SVM) is introduced. Four spectral coherence characteristic parameters are chosen via spectral correlation analysis. By utilizing a nonlinear SVM, a significant amount of calculation is performed offline, thus the computational complexity is reduced. Simulations indicate that the overall success rate is above 92.8% with data length of 1000 when SNR is equal to 4 dB. Compared to the existing methods including the classifiers based on binary decision tree (BDT) and multilayer linear perceptron network (MLPN), the proposed approach is more effective in the case of low SNR and limited training numbers.