A New Approach to Improve Signal Classification in Low SNR Environment in Spectrum Sensing

The performance of cognitive radio is sensitive to the accuracy of signal classification. The proposed method can increase the accuracy of existing methods on the certain degree at SNR=0 dB and below. In simulation, we classify five types of signals which are AM, BPSK, FSK, MSK and QPSK. The experiments show that above 99.9% received signals are correctly classified at SNR=-12 dB and above.