Research On Anomaly Detection Method Of Electromagnetic Environment Via Deep Learning

With the development of wireless communication technologies,wireless communication devices have been widely used in the civil and military fields.However,the use of a large number of wireless communication devices poses great difficulties and challenges for the monitoring of the electromagnetic environment.As part of electromagnetic environment monitoring,how to effectively detect the anomalies of complex electromagnetic environment has important significance that can not be ignored.In order to solve the problem of anomaly detection in complex electromagnetic environment,this paper makes a theoretical modeling and method research on a new method of anomalous electromagnetic environment detection based on depth learning.The main contents of this paper can be summarized as follows:First of all,this paper defines the abnormal electromagnetic environment and gives the corresponding mathematical model.At the same time,this paper analyzes the characteristics of anomaly in the electromagnetic environment,and points out the problems and challenges in the anomaly detection in the electromagnetic environment.Secondly,this thesis studies the related theories and algorithms of neural network and deep learning,and focuses on the auto-encoder algorithm in deep learning,which provides the theoretical support for the subsequent research work.Then,based on the characteristics of the electromagnetic environment anomaly and the related theory of auto-encoder in deep learning,this paper presents an unsupervised detection method of anomaly for the electromagnetic environment,which is suitable for processing large amounts of high-dimensional data,that is,anomaly detection method based on auto-encoder reconstruction.The simulation experiments and experimental results show that the anomaly detection method based on auto-encoder reconstruction can not only correctly detect the anomaly occurring in the electromagnetic environment,but also can effectively improve the detection accuracy compared with other algorithms.Finally,considering the two problems existing in the anomaly detection method based on auto-encoder reconstruction,namely,the high reconstruction cost and the difficulty of threshold selection,this paper improves the method and proposes a new method based on auto-encoder and one-class classification detection method.Simulation experiments and experimental results show that the detection performance of the anomaly detection method based on auto-encoder and one-class classification is superior to other algorithms.By changing the network structure of the automatic encoder,the detection accuracy can be further improved.Moreover,this method can enhance the autonomy of anomaly detection model and provide a new idea for the design of intelligent communication system.