Analysis For Sleep EEG And Study Of Sleep Stages

ABSTRACT:Sleep is an important physiological activity of human. It is a very close relationship with health, work, learning etc. It is the basis for reasonable sleep stages to study the sleep quality and diagnose sleep disorders. Electroencephalogram (EEG) is the most significant and most intuitive signal in describing sleep, so sleep EEG is important and powerful tool about study sleep. Numerous studies indicate that EEG is a chaotic signal. As the complexity of brain electrical activity, using non-linear approach may have better outcomes. This article explains the background of sleep and sleep stages. It describes the principle of EEG produced in detail and several common brain waves and how collected EEG. It explains several key non-linear dynamics:Lyapunov index, complexity, correlation dimension, approximate entropy etc. And describes their treatment of sleep EEG, finding in the different sleep period, the calculated results showed some variation.In addition, the article described the genetic algorithm and support vector machines. In order to improve the performance of SVM, the article proposed an improved SVM algorithm. Using genetic algorithm (GA) and traditional SVM algorithm constructed a parameters optimal evolution SVM (GA-SVM), SVM model using radial basis functions (RBF) as a core function. In this paper, the characteristics of the value (correlation dimension, complexity, Lyapunov indices, approximate entropy) were recognized using GA-SVM pattern by experiments. We used 8 personal(Slp01a,Slp01b,Slp02a,Slp02b,Slp03,Slp04,Slp14,Slp48) data in the experiment. Data samples are 100 groups (4-dimensional feature vector). Slp01a, Slp01b, Slp02a, Slp02b, Slp03, Slp04, Slp14, Slp48's classification of correct recognition rate are:95.33%, 100%,100%,100%,87.33%,95.33%,100% and 95.33%. We put all the data for the six categories and SVM correct identification rate is 78.9%. According to the experimental results showed the model of choice is reasonable. EEG parameters of nonlinear can express different physiological states of the brain, so nonlinear dynamics can be applied in the study of sleep EEG...