Blind Detection System Based On Spectrum Sharing For Machine-To-Machine Communication

Currently,supporting the spectrum sharing of the sporadic M2 M communications and the persistent conventional communications such as 5G cellular transmissions is an active research topic.Most spectrum sharing schemes use interference cancellation to separate mixed signals at the receiving end.Furthermore,most spectrum sharing schemes frequently and extensively use training sequences to pre-estimate the transmission channels which causes a decrease in spectrum sharing efficiency.Therefore,it is particularly important to use effective signal separation techniques and blind detection techniques without the use of training sequences in a spectrum sharing system.This article starts with the two important technologies mentioned above,and does the following innovative works based on M2 M communication spectrum sharing:(1)In order to avoid the problem of reduced spectrum sharing efficiency caused by the use of training sequences,the second chapter of this paper proposes an M2 M communication spectrum sharing blind detection system that uses the Complex Hopfield Neural Network(CSHNN)blind detection algorithm to detect conventional user signals.First of all,under the compressed sensing technology,the redundant characteristics of the transmitted signal are used to build the over-complete expression of conventional user and sparse model of M2 M communications.And then use the convex optimization to solve the M2 M communication signal.We remove the solved M2 M signal from the receiver,and leave the conventional users.Finally,we use the CSHNN blind detection algorithm recovers conventional user signals.Simulation experiments show that the redundancy in the transmitted signals can be exploited to separate the mixtures with compressive sensing techniques.The CSHNN algorithm can successfully implement blind detection of conventional user signals without training sequences which has the low time complexity,the short amount of data,and robustness to different channels.(2)In order to improve the anti-noise performance and bit error performance of the system,the third chapter of this paper proposes a Signal Space Cancellation-Complex System Hopfield Neural Network(SSC-CSHNN)blind detection algorithm.First,we restore the conventional user signal by CSHNN algorithm from the received signal.Then we use the SSC method which adds the first recovered conventional user signals to the noise space,and reconstructs a new complementary projection operator to construct a new quadratic function optimization problem with constraints.Finally,the blind detection algorithm of CSHNN algorithm is used to estimate M2 M communication user signals.Simulation experiments show that the SSC method can effectively separate two mixed signals with different characteristics,and the SSC-CSHNN blind detection algorithm can effectively detect conventional user signals and M2 M communication signals from aspect of error performance,data length,and anti-interference ability.(3)In order to further improve the system's anti-many performance,the fourth chapter of this paper proposes an SSC-CSHNN detection algorithm with improved activation function based on the third chapter.We select appropriate improved activation function through simulation experiment comparisons.At the same time,in order to further improve the convergence speed of the system algorithm,we introduce the double Sigmoid idea,a blind detection algorithm based on Signal Space Cancellation-Double Sigmoid Complex System Hopfield Neural Network(SSC-DSCSHNN)is proposed.We select appropriate double Sigmoid activation function through simulation experiment comparisons.Simulation experiments show that based on the successful separation of mixed signals using the SSC-CSHNN algorithm,the improved activation function-type CSHNN algorithm improves the anti-many performance of the network.Under the condition of the same signal-to-noise ratio,the SSC-DSCSHNN algorithm further accelerates the convergence speed of the algorithm on the basis of ensuring time efficiency,and greatly optimizes the neural network.