Research On Multi-sensor Network Information Processing Algorithm Based On Complex Value

The main principle of multi-sensor network technology is that each node in the network(i.e.the sensors)can obtain instantaneous information from its neighboring nodes.The Mean Square Error(MSE)criterion has long been widely studied and applied in this distributed estimation and has achieved good results.However,most of the related research is based on the circular error signal.In fact,many signals in nature,such as 16-QAM signals,fMRI signals are complex-valued signals,will show different degrees of noncircular characteristics.The mean square error criterion is not ideal for dealing with these non-circular signals.The multi-sensor cooperative network machine under the influence of noncircular signal is studied in this thesis.The main work is as follows:First of all,the basic model of multi-sensor network is introduced,modeling the multi-sensor non-cooperative network.The principle is described by mathematical language,and the cost function is deduced by mean square error criterion.Then,two kinds of cooperative network algorithms,i.e.Adapt-Then-Combine(ATC)algorithm and Combine-Then Adaptation(CTA)algorithm,are introduced in detail,and the multi-sensor cooperative network model is studied.Then,the basic theory knowledge of complex domain signal is introduced,and the definition of noncircular signal,circularity coefficient and Gaussian entropy is given,which may laid a theoretical foundation for the follow-up studies.Then,based on the multi-sensor non-cooperative network,the linear and generalized linear filters are discussed,including MSE-based linear filter and Gaussian entropy-based linear filter.Based on the study of Gaussian entropy,the closed-form solution of the linear filter and the closed-form solution of the generalized linear filter are deduced.The simulation results show that the entropy algorithm is similar to the solution of MSE with circular signal.When the degree of non-circularity of the signal is 0.7,the steady-state mean squared deviation of the original MSE closed-form solution is-25.8dB,and the entropy closed-form solution is-29.2dB.As the degree of non-circularity increases,the steady-state error is smaller.Finally,based on the multi-sensor cooperative network algorithm,this thesis analyzed the distributed network model of N nodes and derives the global optimization based on entropy.Based on the ATC algorithm and CTA algorithm,a multi-sensor cooperative network algorithm based on Gaussian entropy is deduced,and the Gaussian entropy is introduced into the multi-sensor cooperative network algorithm.Then it analyzed the performance of the algorithm,including the mean stability analysis and mean variance performance analysis.The simulation shows under the influence of noncircular noise,when the degree of non-circularity is 0.7,the steady-state error of the traditional cooperative algorithm is-26 dB,the algorithm proposed in this thesis is-36 dB.As the degree of non-circularity increases,the steady-state error is smaller.