Spectrum Sensing Algorithm Based on Self-Supervised Contrast Learning

The traditional spectrum sensing algorithm based on deep learning requires a large number of labeled samples for model training, but it is difficult to obtain them in the actual scene. This paper applies self-supervised contrast order solve this problem, and (SSCL) proposed. mainly includes two stages: pre-training fine-tuning. In stage, according characteristics communication signals, data augmentation methods are designed pre-trained positive sample pairs, features pairs unlabeled extracted by feature extractor. fine-tuning parameters extraction layer frozen, small used update classification layer, labels connected get classifier. simulation results demonstrate that SSCL has better detection performance over semi-supervised energy algorithm. When only 10% supervised SNR higher than −12 dB, probability 97%, which slightly lower