Chaotic Time Series Prediction And Weak Pulse Signal Detection Based On BEL Learning Network-GA Algorithm

Chaotic time series is the time series observed from chaotic system.Chaotic time series is widely used in weather,hydrology,earthquake,signal and other fields.Weak signal is the signal submerged in the background of strong noise,which is difficult to be detected by traditional detection methods.Weak signal has been applied in many engineering fields.The application of chaotic time series and weak signal in these fields is shown in meteorological forecast,flood warning,earthquake early warning and encrypted communication in real life.In order to improve people's living standard and quality and ensure life safety,it is necessary to study chaotic time series prediction and weak signal detection.In this paper,chaos theory,chaotic time series prediction theory and weak signal detection technology are combined to study the prediction of chaotic time series and weak pulse signal detection under chaotic noise background.Firstly,the affective learning network model(BEL)was proposed by improving the affective learning model.The improvement method was to change the activation function and learning weight of two linear structures in the model,the amygdala and the orbitofrontal cortex,into a nonlinear structure,and the learning proportion weight was applied in the output part to describe the different influence degrees of the two structures.Then design a dynamically changing adaptive genetic operator and optimize the parameters of BEL model are by adopting adaptive genetic algorithm.The parameters of affective learning network model(BEL)were encoded into chromosomes.Through the dynamic crossover probability and mutation probability to control the crossover and mutation process.The optimal parameters of the current generation of chromosomes were selected by evaluating the chromosomes with fitness function,and the optimal parameters of BEL were obtained through iteration.Finally,combining the emotional learning network model(BEL)and adaptive genetic algorithm,the emotional learning network model-adaptive genetic algorithm(BEL-AGA)prediction model is proposed.In the research of weak pulse signal detection under chaotic noise background,firstly,based on the short-term predictability of chaotic signal and phase space reconstruction technology,the detection problem of signal is transformed into the detection problem of prediction error.Then the observation signals were reconstructed in phase space,and the prediction errors were obtained by the affective learning network model-adaptive genetic algorithm(BEL-AGA)prediction model,and the weak pulse signals were detected from the prediction errors by hypothesis testing.Finally,R?Matlab software is used to predict the model and signal detection experiments.The results show that:(1)the chaotic time series based on Rossler chaotic system and sunspots time series prediction,BEL-AGA model predicts the mean square error(MSE)of 0.0733 and 1.8872,emotional learning model of mean square error(MSE)of 0.2784 and 12.2187,the contrast can be found that the BEL-AGA model prediction effect is better than the original emotional learning model,better fitting of chaotic time series has higher prediction accuracy;In the comparison of prediction results under different sample sizes,the deviation of prediction error of BEL-AGA model is small when the sample size changes,while the deviation of prediction error of affective learning model is large when the sample size changes,and the BEL-AGA model has stronger prediction stability.The rationality of the improved method proposed in this paper is verified,and the superiority of the model compared with the affective learning model is verified.(2)the chaotic noise under the background of weak pulse signal of the signal-to-noise ratio(SNR)is 65.8145,the BEL-AGA detection model of the detection accuracy(ACC)is 1,can effectively detect the weak pulse signal,the signal-to-noise ratio(SNR)corresponding to 73.4031,78.7045,81.8117,86.8380,the detection accuracy of the model(ACC)is 0.9988,0.9852,0.9617,0.9103,can effectively detect the weak pulse signal.Under the same signal-to-noise ratio,the detection performance of the model is better than that of LM model and BP neural network,which verifies the rationality of the detection model proposed in this paper and the excellent detection performance.