Cognitive radio is a new technology, which aims to achieve the user’s dynamic spectrum access and improve the efficiency of the spectrum. Spectrum sensing is the basic of cognitive radio, and spectrum sensing is used to discover the spectral holes in cognitive radio networks, which allows secondary users(SU) to communicate without causing harmful interference to primary users(PU). It requires good perceiving performance even at very low signaltonoise ratios to ensure that the perceived user communicates without affecting the primary user′s communication in spectrum sensing. This paper presents an efficient sensing algorithm based on stochastic resonance (SR) and support vector machine(SVM). Firstly, the detecting signal through the Stochastic resonance system to improve signaltonoise rate and strength the feature of signal. Secondly the FAM algorithm is used to extract the cyclostationary feature of signal. Finally, we use support vector machine to classify the feature to get the detection results. The experimental results show that the method proposed in this paper has higher detection reliability than the traditional detection method at low SNR.