Research On Signal Detection And Fusion Based On Chaotic Noise Background

The detection of weak signal in chaotic noise background is a new detection method based on nonlinear system.It has been a research hotspot in the field of signal detection.The existing researches on weak signal detection under chaotic noise background are almost all carried out under the single-sensor observation mechanism.This paper considers the distributed detection and fusion of weak signals in chaotic noise background under the multi-sensor observation mechanism.First of all,by modeling the chaotic noise and the target signal,the distributed detection fusion problem in chaotic noise background is abstracted into a hypothesis testing problem.As the key to solve this problem lies in the processing of chaotic noise signal,the problem is further divided into two steps,that is,chaotic signal processing and detection and fusion of distributed signals.Then,the phase space reconstruction technology is introduced to establish the chaotic prediction model,and the chaotic signals are stripped from the observed signals to obtain the one-step prediction errors.The linear autoregressive model(LAR)and the local weighted regression model(LOESS)are considered in the establishment of chaotic prediction model.The simulation results show that when the observed signal only contains chaotic signal and target signal,the chaotic signal can be successfully stripped out by the two models.And the target signal can be clearly reflected by the one-step prediction errors.When the observation signal contains chaotic signal,target signal and observation white noise signal,the two models can only strip out chaotic signal,but the influence of observation white noise cannot be eliminated.It is difficult to detect the target signal through one-step prediction error directly.The performance of LOESS model is obviously better than that of LAR model at this time.Finally,on the basis of chaotic signal preprocessing by using LOESS model,the one-step prediction error is regarded as a new observation signal.The target signal is detected and fused from these new observation signals.The distributed detection fusion problem is studied under parallel and serial architectures respectively.Under the framework of Bayes risk theory,the optimization models were established to minimize the Bayes risk of fusion center,in which the decision rules of local sensors and fusion rules of fusion center were the decision variables.An iterative algorithm is designed to solve the problem by using the Gauss-Seidel method.The simulation results show that the proposed method and model can effectively detect the target signal,which is submerged in chaotic signals and white noise signals.The LOESS model is successful in processing chaotic signals,which can remove chaotic signals and minimize the influence of observation noises.The resulting one-step prediction error preserves the information of target signal.The detection performance of fusion center is generally better than that of local sensors,and this superiority will be further expanded with the increase of the number of local sensors.