# EEG Sleep Stages Classification Based on Time Domain Features and Structural Graph Similarity(2016) - ### Abstract - The proposed model uses graph theory and extracts the features from EEG and uses K-Means algorithm to classify the sleep stages. - ### Model - The model classifies EEG into 6 sleep stages(RK method). - Each segment(sleep epoch) of EEG signal is broken down to m sub segments, whose features are extracted and mapped to a graph. - THe structural similarity properties are extracted and put into a classifier. - ![](https://i.imgur.com/sZ2uRjz.png) - The 12 key features to represent the graph are median, mazimum, minimum, mean, mode, range, first qaurtile, second quartile, standard deviation, variation, skewness, kurtosis. - The features extracted are degree distribution, clustering coefficient andJaccard similarity coefficient. - K-means clustering algorithm used for classification. - ### Evaluation - The dataset used was Sleepedf. - Evaluation was done taking top 6, top 10 and all 12 features from the graph as mentioned above. - Classifaction based on 12 features gave the best result with an accuracy of 95.15%.