Research And Implementation Of Intelligent Technologies Of Digital Spectrum Monitoring System

With the research and development of 5G technology,The Internet of Things technology with satellite,and the commercialization of drones and software radio equipment,more and more military,commercial,and personal radio equipment are connected to wireless networks.In order to ensure that these spectrum resources are not misused,it is necessary to monitor the radio equipment that accesses a certain section of the spectrum,and then determine whether they are legal access equipment.In order to ensure the legal use of spectrum resources in the military and commercial fields,and to protect some key safety buildings from the influence of radio frequency monitoring equipment,research on key technologies of digital spectrum monitoring systems is of great significance.However,currently there are mainly spectrum monitoring systems that can only deal with a set of predefined modulations,and when a new kind of modulation is added,it cannot quickly adapt to solve this problem.Moreover,the software architecture design and software deployment of the spectrum monitoring system are still relatively complicated,which affects the scalability of the subsequent system.For this reason,this paper designs a intelligent digital spectrum monitoring system,and focuses on the research of modulation recognition algorithm based on neural network.It designed the overall architecture of the digital spectrum monitoring system,realized the functional modules for parameter estimation,and optimized the procedure of software deployment of the system.The main research work of this paper is as follows:1.In-depth research on modulation recognition algorithm based on neural network,and an improved two-stage neutral network-based method is proposed.First,according to whether the neural network itself extracts signal features,the modulation recognition algorithm is divided into two neural network algorithms that require preprocessing and those that do not require preprocessing.First,the feature-based modulation recognition algorithm is compared with the neural network-based modulation recognition algorithm that requires preprocessing in different dimensions.It is found that when the training samples are sufficient,the performance of the modulation recognition algorithm based on neural network is better.However,because the combination of different features has a great impact on the modulation recognition performance,this modulation recognition algorithm cannot quickly adapt to the data set scene of the newly added modulation mode.For this reason,a neural network-based modulation recognition algorithm without preprocessing is studied to solve this problem,but compared with the former,the overall recognition performance needs to be improved.Aiming at the data set used in this article,a two-stage neural network-based modulation recognition algorithm is proposed to improve certain performance.2.Research on the spectrum scanning problem of multi-carrier separation in the monitoring system,and propose an improved spectrum scanning method without smoothing.The main function of spectrum scanning is to roughly locate the position of each carrier in the entire channel.And then provide calculation anchor points for the estimation of the number of carriers,carrier center frequency,carrier bandwidth and other parameters.Compared with the traditional maximum-minimum method,which is affected by the smoothness of the power spectrum,this article hereby discusses the results of downloading wave parameter estimation in different smoothing methods.In this regard,a spectrum scanning method based on the least square method is proposed.The main advantage of this method is that even if the power spectrum is not smoothed,it still has a very high accuracy.3.Using USRP X310 and Keras deep learning framework to design a set of front-end and backend separated intelligent spectrum monitoring system.First of all,for all parameter display and some low-calculation parameter estimations,all are handed over to the front-end implementation.Then for the functions of large amount of calculation,the computing service router is designed,so that the front and back ends can interact through the network protocol,which greatly improves the scalability of the system.And in terms of system deployment,Docker container technology is used to simplify software deployment.Finally,a certain index is proposed for the system,and the function test is carried out in conjunction with the Pluto equipment,and the result is in line with the proposed index.