Abstract: Cognitive Radio (CR) is an efficient way to solve the problem of the lack of the spectrum resource. In a Cognitive Radio Network the unlicensed users (secondary users) must incessantly monitor the spectrum for the presence of the licensed users (primary users) to avoid the interference to primary users. In this study, a spectrum sensing scheme based on adaptive optimal SVM (support vector machine) is proposed. A prototype system and the simulation experiments show that in low SNR the algorithm can also get a reasonable probability of detection and a low probability of false alarm.