A Cognitive Radio Spectrum Sensing Algorithm to Improve Energy Detection at Low SNR

Abstract

Energy detection is among the most popular spectrum sensing method for spectrum sensing due its low complexity. Unfortunately, its performance is poor at low SNR. In this paper we proposed a spectrum sensing method for cognitive radio network that improves the performance of energy detection. The proposed method based on distribution analysis using kurtosis as test statistic. This comes from the fact that distribution of received signal when a channel is occupied will be different from vacant channel. Noise tends to have a Gaussian distribution. Signal which faces multipath fading during the transmission way will have non Gaussian distribution. Sensing algorithm was tested using captured DTV signal. Result shows that our method performs well at low SNR. It achieves probability of detection of 90 % for 10 % Probability of false alarm for low SNR, below -20 dB.