Detection Of Weak Pulse Signal Under Chaotic Noise Based On Long Short-Term Memory Neural Network

There are a large number of weak pulse signals in daily life.Generally,these signals are masked by noise with chaotic backgrounds,and the signal strength of weak pulse signals is relatively weak compared to noise signals,making it difficult to detect these signals.In practical applications,detecting the presence of weak pulse signals in observed signals has become the focus of research in various fields,such as fault diagnosis,geological exploration,and abnormal detection of cardiac and brain signals.Due to the extreme sensitivity of chaotic sequences to initial values,they exhibit a seemingly irregular random motion.How to accurately depict the trajectory of chaotic signals has become an important breakthrough in solving such problems.After a systematic and in-depth study of the problem of weak signal detection,considering the motion characteristics of chaotic systems and the advantages of Long ShortTerm Memory(LSTM)neural networks,this paper proposes a weak pulse signal detection model based on Long Short-Term Memory memory neural networks under chaotic noise.Firstly,through reading relevant literature,summarize and analyze the ideas and methods of weak signal detection.Then,it elaborates the relevant theoretical knowledge studied in this article,such as chaos theory,methods of phase space reconstruction,structures and algorithms of cyclic neural networks and short-term memory neural networks.Finally,the implementation steps of the detection model are analyzed,mainly including: reconstructing the phase space of the observed signal;Establish a short-term and short-term memory neural network fitting model for the reconstructed observation signal;Finally,the target signal is further obtained from the prediction error.Reconstructing the phase space of chaotic signals can effectively extract chaotic information from sequences,and constructing LSTM neural networks can effectively learn long-term correlations and important features in time series.The gating structure in the LSTM unit can remember important information and filter out unimportant information,which allows the LSTM model to predict observed signals with larger errors at signal points and smaller prediction errors at non signal points.This can significantly distinguish between signal points and non signal points,effectively detecting weak pulse signals in chaotic backgrounds,and reducing the detection threshold.In order to further illustrate the feasibility,effectiveness,and advantages of the model in this paper,four simulation experiments have been conducted.The experimental results show that:(1)When detecting weak impulse signals in Lorenz chaotic system and Rossler chaotic system,the model has high detection accuracy,and can still maintain detection performance when the signal-to-noise ratio is low.(2)Compared with other machine learning models and deep learning models the detection accuracy of LSTM models is higher than other comparable models under different signal-to-noise ratios.When the signal-tonoise ratio is low,such as in the Lorenz system,where the signal-to-noise ratio is below –84d B,the other models cannot detect the target signal or have low detection accuracy,the LSTM model still has strong detection performance and good stability,with a detection accuracy of 0.9987.(3)The model successfully detects weak pulse signals in the sunspot sequence and achieves the purpose of diagnosing fault signals in rolling bearings.These results show that the model can accurately detect weak pulse signals in chaotic backgrounds with low signal-to-noise ratios,and is suitable for dealing with weak signal detection problems in chaotic noise backgrounds in real life and fault diagnosis problems in engineering applications.This can not only reduce the detection threshold for weak signal detection,but also expand the scope of application for weak signal detection.