Adaptive Deep Knowledge Framework for classifying Sleep Stage using Deep Feature Learning

Conventional wisdom holds that sleep is a universal, simultaneous event that impacts every part of the brain. An electroencephalogram (EEG) is a non-stationary and nonlinear way to monitor brain electrical activity. EEG signals have many applications, from researching the most fundamental parts of the sleep cycle to vital components of medical diagnosis. But up until recently, scientists didn't know much about the specifics of how distinct EEG characteristics relate to the stages of sleep. The feature extraction method was used immensely for accurately classifying EEG data throughout different stages of sleep. This research aims to use the Channel Based LSTM Convolution Network (CLCN) design to improve subject-independent classification accuracy. We compare this model's output to those of three other ML techniques. After comparing the suggested feature extraction method to the other available options, the findings show that it produces the best classification accuracy. Applying the SleepEDF EEG dataset to a five-class classification task yields the best classification accuracy of $\mathrm{9 5 . 7 8 \%}$.