Small Sample Electromagnetic Signal Recognition based on Time Series Data Augmentation
Xinrui Wang, Yun Lin · 2021 8th International Conference on Dependable Systems and Their Applications (DSA) · 2021
With the rapid development of science, technology and industry, the electromagnetic environment becomes more and more complex. The acquisition of information first needs to identify the signal. However, the number of electromagnetic transmitting equipment is large, and the signal is unstable, which easily leads to a small number of available signal samples. As a result, over fitting is easy to occur in the training process, which reduces the accuracy of recognition. In this paper, the complex network model is used for the electromagnetic signal recognition. Aiming at the problem of less dataset, the time series data augmentation method is used for the processing of this dataset, so as to enhance the accuracy of electromagnetic signal recognition. According to the characteristics of time series, four methods are used to augment the data, including time domain flipping, amplitude inversion, amplitude scaling and noise disturbance. The experimental results display that the four augmentation methods can enhance the recognition accuracy, and the mixed form of different augmentation methods can better enhance the effect.