Laboratory access recognition based on MAML and radio frequency fingerprint

Addressing the efficiency and accuracy issues of radio frequency fingerprint recognition in laboratory access control, this paper presents a recognition model based on model-agnostic meta-learning and radio frequency fingerprint. By integrating short-time Fourier transform and constellation diagram, the proposed approach significantly enhances identification performance. The experimental results show that in the simulation run experiment, the average recognition accuracy of the proposed model before and after training is 0.59 and 0.97, respectively. When the training sample size is 10 and 5, the average recognition efficiency of the proposed model is 93.80% and 90.15%, respectively. When the model adds the short-time Fourier transform and constellation diagram, its recognition accuracy and efficiency increase by 0.19 and 13.01%, respectively. In addition, in actual model performance experiments, the proposed model shows the most stable performance, with an average of 90.47% for recognition accuracy and 0.91 for efficiency. Moreover, the recognition accuracy of the model for different types of signals can reach up to 0.91. In environments with signal-to-noise ratios of 5 and 0, the average recognition efficiency of the proposed model is 83.20% and 62.00%, respectively. The original contributions of the study lie in the optimization of feature extraction and model training processes, which significantly enhance the accuracy and efficiency of radio frequency fingerprint recognition, especially in scenarios with limited sample sizes. This addresses the issues of low accuracy and poor efficiency in traditional radio frequency fingerprint models. The study is capable of improving the efficiency and accuracy of radio frequency fingerprint recognition and enhancing the level of laboratory safety management and control.