Research On Signal Processing And Sleep Staging Based On EEG

As modern society is advancing rapidly,the pace of life is speeding up,there are increasing threats of stress on people's physical-mental health.The most obvious performance of stress is that more and more people suffer from insomnia or poor quality sleep,which results in physiological or mental illness.Therefore,the study of sleep-EEG has increasingly aroused great concern among people.This paper gives a brief introduction to the basic characteristics and waveforms of sleep-EEG signals.The EGG signals have different characteristics in different periods of sleep,which provides the theoretical foundation for sleep staging.The sleep-EEG test data are obtained from the database of Massachusetts Institute of Technology.In this paper,we design the algorithm from three aspects: signal preprocessing,feature extraction and classification.Firstly,since there are a lot of background noises in the EEG-signal,it is necessary to select the appropriate denoising method.Through simulation experiment,wavelet packet denoising results are compared,wavelet basis function,wavelet decomposition layer and threshold function are selected.Secondly,Non-linear dynamics has been used to analyze characteristics of filtered EEG signals and correlation dimension,complexity,Lyapunov exponent and approximate entropy can be used to characterize sleep state.The Lyapunov exponent proves the chaotic nature of the brain and correlation dimension,complexity and approximate entropy showed a regular change in different sleep states.Finally,based on the analysis and comparison of the principle and algorithm of different classifiers,particle swarm optimization algorithm is used to optimize the parameters.Through the comparison of artificial neural network,least squares vector machine and particle swarm optimization algorithm based on the results of sleep staging,the effectiveness of the proposed optimization algorithm is verified.The results of experiment on sleep-EGG are as follows: among recognition experiments on eigenvalue of 7 * 400 groups of sleep-EGG signals,there are 338 correct groups and 31 incorrect groups during the awake period,and the accuracy rate is 91.60%;there are 411 correct groups and 47 incorrect groups during shallow sleep period,and the accuracy rate is 87.68%;there are 269 correct groups and 28 incorrect groups during rapid eye movement,and the accuracy rate is 90.57%.The final average accuracy rate can reach 90.00%.Therefore,the automatic staging systemdesigned in this subject is feasible and effective.