A Spectrum Sharing System: Visualized Interference Classification and CNN Approach
Hyelin Nam, Kyuwon Han, Jihoon Park, Han Cha, Seong‐Lyun Kim · 2021
The scarcity of the low-frequency band has demanded Dynamic Spectrum Access (DSA). We implemented DSA system that detects channel usage of primary users and controls secondary users through a deep neural network in real-time. With an LTE base station that adopts Citizens Broadband Radio Service (CBRS) protocol, we implement the Navy’s active sensing radar signal through a Universal Software Radio Peripheral (USRP). We obtain visualized image dataset for CBRS system, called ViC dataset consisting of 12 classes with 500 samples per class and visualize it through spectrograms. CNN detects each class of primary-existence scenarios with an accuracy of 96.37%.