Nonlinear Dynamics Analysis On EEG Signals And The Application To Sleep Stages

Nowadays, people pay more attention to health. And WHO has suggested that researches should focus on human’s health. With effective monitoring and analysis on daily physiological and psychological status, abnormal state could be found out in time. Brain is our control center, containing rich information. Applying the newly-developing nonlinear dynamic methods into EEG analysis and characteristic parameters extraction, a deeper and better way will be discovered to help with the study in bioscience. And if the methods are used for sleep stage classification, it will be of great importance to sleep researches and may provide new means for better classification.EEG extracted through electrode is the reflecting of electrical activity in nerve system. EEG is a weak signal and surrounded by a lot of interference. So before feature extraction, the original signal should be filtered to remove the baseline drift, highfrequency and power line interference.In recent years, many studies show EEG has chaotic characteristic. So in this paper, 3 kinds of nonlinear dynamic method are chosen with best parameters combination using the sleep EEG in MIT-BIH database. And to compare with traditional method, frequency band energy ratio parameter and wavelet multi-resolution analysis parameter are also chosen to compare the final result. 7 parameters are chosen with 3 nonlinear dynamic parameters called complexity, approximate entropy and correlation dimension, 2 frequency band energy ratio parameters called ratio(α) and ratio(δ),and 2 wavelet multiresolution analysis parameters called D3-1 and D3-2. Then we compare these parameters with the experts’ sleep stage result for sleep EEG, and calculate the correlation coefficient between the parameters and the expert’s result. Combining with computation time, individual variation and significance in sleep stage, we can choose the best parameter as correlation dimension, and apply it into sleep EEG analysis and sleep stage.Correlation dimension will decrease with deeper sleep. We make quantification for the relation between correlation dimension and every different sleep stage, and get a range of the correlation dimension value for each stage. Then the value is used to analyze the sleep EEG signal actually collected in our laboratory of our schoolmates. And an approximate sleep stage can be got for the subjects.