Research On Signal Analysis Technology Of Fiber Optic Perimeter Security System Based On Deep Neural Network

At present,the "Internet of Everything" has become an indispensable symbol of the "smart city",the Internet of Things has gradually appeared in every corner of people’s lives,and the development of society and technology has promoted the technological transformation of various industries.For the security system,most scenes still adopt the traditional way of physical isolation and security personnel patrol,which not only cannot achieve real-time protection of the security area.Achieving high quality security means higher personnel costs.Optical fiber perimeter security system is concerned by the industry because of the advantages of optical fiber sensing technology.Fiber optic vibration sensor perception of external effects in the signal parameters,because the fiber has high sensitivity,anti-electromagnetic interference,high security characteristics,so that the fiber perimeter security system has become the focus of the current perimeter security field,is the future perimeter security field essential security technology.In the fiber perimeter security system,the recognition of intrusion behavior with high accuracy and low delay is realized by improving the signal analysis method and designing the signal recognition algorithm.In this paper,the optical fiber vibration sensing signal analysis technology in the optical fiber perimeter security system is studied as follows:Firstly,the signal characteristics of fiber optic vibration sensing are screened and improved,and the input data form of signal recognition algorithm is improved.According to the characteristics of fiber optic vibration sensing signals,feature extraction and simulation analysis were carried out.Finally,the features that can distinguish different types of signals were obtained and used as data samples for model training and testing.In the deep learning algorithm,MFCC is selected as the input data of the algorithm model,which not only reduces the amount of computation,but also ensures the accuracy of signal recognition.Secondly,the gated recurrent unit(GRU)algorithm and the Gaussian Mixture Model and Hidden Markov Model(GMM-HMM)has good technical advantages and recognition effect in speech signal recognition area.Therefore,GRU algorithm and GMM-HMM algorithm are proposed as the recognition algorithm of fiber optic vibration sensing signal.Thirdly,GRU algorithm and GMM-HMM algorithm are used to identify and classify the same data set of fiber optic vibration sensing signals,and the advantages and disadvantages of the two algorithms are compared through simulation experiments.The simulation results show that the model recognition accuracy of GRU algorithm is 97.44%,and the model recognition accuracy of GMM-HMM algorithm is 98.40%.It is concluded that both GRU algorithm and GMM-HMM algorithm are suitable for realizing different types of fiber vibration sensing signal recognition in fiber perimeter security system.