Machine learning techniques for spectrum sensing when primary user has multiple transmit powers

In this paper, we propose a machine learning based spectrum sensing framework for a new cognitive radio (CR) scenario where the primary user (PU) operates under more than one transmit power level. Different from the existing algorithms where the primary transmit power levels are assumed to be known, the proposed approach does not require much prior knowledge of either the primary user or the environment. Before sensing, the cognitive user will first be aware of the environment from a learning phase, where the unsupervised learning algorithm (e.g., K-means clustering) is applied to discover PU's transmission patterns as well as its statistics. Then, the supervised learning algorithm (e.g., Supporting Vector Machine) is implemented to train the CR to distinguish PU's status based on energy feature vectors. Simulations clearly demonstrate the effectiveness of the proposed machine learning based spectrum sensing framework.