Wideband Spectrum Sensing With Multiple Nodes Based On Deep Learning

Wideband spectrum sensing technology is a hot topic in current research.The purpose is to sense unoccupied idle frequency bands in the wideband spectrum.However,the limitation of high sampling rate and hardware cost makes many spectrum sensing algorithms unable to be applied to wideband spectrum sensing.Compressed sensing can reduce the requirements for sampling rate and is an effective technical means to solve the above limitations.However,due to shadows and multipath fading,the perception of a single node is unreliable.Compressed sensing to improve perceptual accuracy.In the traditional methods of using distributed compressed sensing to solve the problem of wideband spectrum sensing,most methods need to know the sparsity of the signal in advance,but in practical applications,the sparsity of the signal cannot be known in advance.In addition,many traditional methods have low probability of correct estimation of the spectral support set,and the obtained results are not the global optimal solution.Moreover,most traditional methods are proposed for a specific signal scenario,only applicable to some specific scenarios,and not universal.These are the problems that need to be solved urgently in wideband spectrum sensing.According to the research status of wideband spectrum sensing,it can be found that the current wideband spectrum sensing problem is mainly limited by high sampling rate,and the accuracy of spectrum support set estimation is not high,the real-time performance of sensing algorithm is not strong,and the universality of sensing method is not high.and poor practicality.In view of the above problems,the research content and main results of this paper are as follows:First,this paper analyzes and summarizes the problems faced by wideband spectrum sensing.In order to solve the constraints of high sampling rate on wideband spectrum sensing tasks,according to the theory of compressed sensing,a model for wideband spectrum sensing based on distributed compressed sensing is established.Firstly,multiple nodes are used to compress and sample the signal,and the data is transmitted to the fusion center.In the fusion center,the sensing algorithm is used to estimate the wideband spectrum support set to complete wideband spectrum sensing.The requirement of sampling rate and the cost of data transmission are reduced,and the accuracy of perception is guaranteed.Secondly,a distributed compressed sensing reconstruction algorithm based on deep learning is proposed.The key of the multi-node fusion wideband spectrum sensing algorithm is how to use the multi-node undersampled data to estimate the signal support set of the wideband spectrum,and the core is the support set estimation problem of distributed compressed sensing.Considering that the influence of different node signals on the reconstruction results is different,based on deep learning and attention mechanism theory,this paper proposes a distributed compressed sensing reconstruction algorithm(LA-DCS)based on LSTM and attention mechanism,and based on GRU and improved attention.Algorithms such as Distributed Compressed Sensing Reconstruction Algorithm(GA-DCS)with force mechanism.Simulations show that the mean square errors of the algorithms proposed in this paper,such as LA-DCS and GADCS,are lower than those of the classical algorithms,and the reconstruction performance is better.Finally,a non-reconstructive wideband spectrum sensing algorithm based on distributed compressed sensing is proposed.Each node undersamples the wideband signal and performs joint sensing at the fusion center.When sensing,based on the proposed distributed compressed sensing technologies such as LA-DCS and GA-DCS,it is not necessary to completely reconstruct the original signal,but only need to estimate the support set of the wideband signal.Simulations show that the wideband spectrum sensing algorithm proposed here is better than the traditional sensing algorithm,the spectral support set estimation accuracy of this algorithm is higher,and the sensing effect is better.