Study On Control Mechanism Of Deep Sleep Process Based On Eeg Signals Time-frequency Analysis

With the rapid development of society,people are facing tremendous pressure in their work and life,accompanied by the phenomenon of sleep disorder,which makes great disturbance to human's mind and body,including the decline of memory,the descent of immunity and the distraction in work.According to the latest medical research,about 300,000 people commit suicide each year in China,of which over 89% have had insomnia,anxiety and depression.Therefore,research on improving sleep has been paid much attention.Sleeping therapy devices has been appeared on the market,but they all have deficiencies as follows:(1)lack of guidance on sleep-related theories;(2)fail to monitor the status of sleepers in real time;(3)unable to form a closed-loop feedback for the sleep control process.In this paper,on the basis of deep sleep,a sleep control system based on the feedback control idea is proposed,which uses EEG as feedback,human sleep state as control object and hypnotic sound as control variables.Firstly,the sleep state is identified by non-linear time-frequency analysis.Secondly,the rules of sleep control are formulated according to the principle of biological resonance and the sound-induced library of sleep state is constructed.Finally,the body is induced to enter deep sleep under the action of sound.The specific works are as follows:(1)Study on identification methods of sleep state.Firstly,sleep mechanism of human is analyzed,and analysis is carried out to the main components of EEG signals of each sleep state according to the different wave components of physiology and EEG signal,then preprocessing is conducted using wavelet threshold and Butterworth filter.Finally,the sleep state is identified by two time-frequency analysis methods---short-time power spectrum estimation and WVD.(2)Study on sound induction of human sleep state.By studying hypnotic principle--theory of biological resonance,it is proposed that human sleep state can be induced to alter by exploiting the rhythm sound of different frequency band,scilicet five kinds of sounds including ?,?,high ?,low ? and pink noise is respectively used to induce the transition from sleep to deep sleep according to the different proportion of EEG signal components in different sleep states.(3)Study on control mechanism of deep sleep process.On the basis of studying human sleep state and process,a sleep control rule is set up which uses current sleep state as feedback input and five kinds of induction sounds as output control variables.Then the current sleep state can be identified by this control rule.Furthermore,the induce sounds are played to accomplish deep sleep base on the optimization of the sleep "path".(4)Experiment and effect analysis of sleep state induction.EEG signals of each sleep state of experimenters are collected in real-time with the aid of Emotiv EPOC helmet.And the next steps are as follows: using the sleep state identification method proposed in this paper to identify sleep state,taking the identification result as the feedback of deep sleep control system and playing different induction sounds according to the sleep control rules to realize the deep sleep state migration of experimenters.A large number of repeated tests are carried out,and the experimental results are analyzed.Through the above work,study on the control mechanism of deep sleep process that carrying out time-frequency analysis to EEG signals has been realized.Experimental results show that the accuracy of sleep state identification of these two time-frequency analysis methods can reach to 88.89% and 91.6%,and its stability and robustness have been significantly improved;what's more,sleep control system based on the control rules can gradually control human body into deep sleep,so as to improve sleep quality significantly.