Research on KNN assisted hybrid energy detection model and its application

• 1、School of Information Science and Engineering of Yunnan University，Kunming 650091
• 2、Yunnan Key Laboratory of Spectrum Sensing and Frontier Radio Safety，Kunming 650091

Abstract：Energy Detection (ED) is widely used in radio monitoring. However, the performance of ED is damaged by noise uncertainty at low SNR. To solve the problem, this paper proposes a KNN (K-nearest Neighbor) assisted Hybrid Energy Detection model (HED), which is composed of three modules: SNR calculation, energy detection and KNN classification. Energy detection algorithm will be enabled at high SNR environment, and KNN classification algorithm will be enabled at low SNR environment. The simulation results show that whether noise uncertainty is considered or not, the detection probability of KNN is relatively high at low SNR, which is breaking the limitation of traditional energy detection at low SNR; ED fails, but high detection probability is obtained by using KNN algorithm when the noise uncertainty is 3dB and SNR is less than 4dB. Finally, in order to explore the application of HED model, 24-hour data was collected from the FM broadcast channel with the central frequency of 100MHz in the real environment, and a data set was made for experiments. The experimental results show that the HED model holds better availability and effectiveness, which provides a new idea and solution for radio monitoring in future.

Keywords： energy detection KNN uncertainty probability of detection SNR