The classification of sleep EEG with partial least squares regression method

The aim of this study is to put forward the success of the classification carried out by using partial lest squares regression in order to get sleep initiation early and to differentiate these two stages from each other by using brain activity situation observed during sleep and wake cycle observed in EEG sleep record. In order to make a comparison, k-nearest neighbor and bayes classification methods were applied with the same data and when the results were compared, classification carried out by using regression method was found 90 % successful and was seen more advantageous in terms of time and processing load. In the study, EEG records that belong to seven different healthy individuals were used and the EEG signs are gathered from the recordings that belong to sleep scoring studies present in sleep EDF database in physioBank. These records were exposed to pre-treatment composed of normalization and filter and autoregressive modelling method was used to exctract the feature. It is supposed that this study can be used in clinical applications and in sleep warning systems.All of the results in the study were gathered through MATLAB program.