A single frame spectrum signal detection approach based on deep learning

Auto signal detection is an important technique for spectrum monitoring and management, but the detection accuracy and robustness might degrade when the signal is weak and the disturbance is heavy. In this paper, we researched the effectiveness of a deep-learning based method for robust single frame spectrum signal detection. Firstly, a deep convolutional neural network structure is designed for single frame spectrum signal detection. Then, in order to alleviate the information scarcity in single spectrum frame, an enhanced input data representation manner is presented to provide additional information and clue to the neural network. The experiments show that the convolutional neural network is effective and robust for single frame signal detection, and the enhanced input data representation can help the deep-learning based classifier to improve the detection performance.