A PORTABLE GUI BASED SLEEP DISORDER SYSTEM CLASSIFICATION BASED ON CONVOLUTION NEURAL NETWORKS (CNN) IN RASPBERRY PI

 Atianashie Miracle A.¹, Elisha D'Archimedes Armah², and Nasiru Mohammed³

¹ Department of Computer Science, Catholic University College Fiapre, Sunyani Ghana

²Department of Computer Science and Technology, Cape Coast Technical University Ghana

³ Department of Artificial Intelligence, Beijing University of Civil Engineering and Architecture

View Full Article

Abstract: Sleep Disorder is the most well-known clinical issue experienced in medication and psychiatry that can affect your sleep quality. Sleep apnea and Insomnia are the most common disorders these both disorders impact the quality of life later on in health and also lead to Heart strokes, hypertension, etc. Sleep hygiene is primary an essential component of human life and important to maintain a good diet and exercise. Without the proper required amount of sleep, we cannot work properly the next day onwards and are disturbed mentally as well as physically. In this paper, we developed a new model, an automatic framework to detect the sleep disorder using Convolution Neural Network (CNN) then developed a Graphical User Interface for the classification of the disorder. The Proposed model is deployed on a raspberry pi processor board for real-time prediction. We trained our developed model on the most popular databases MIT – BIH Polysomnography (PSG) and Cyclic Alternating Pattern (CAP) and yields 92 % accuracy. The use of single-channel EEG to classify sleep disorders is described in an article as remarkable. The feature extraction and feature selection algorithms are not required. Physicians will be able to spot certain sleep patterns, such as exhaustion, sleepiness, and sleep disorders, with the proposed technique. In this article, we introduced a novel single-level architecture for sleep disorder classification. The proposed model is developed with a Deep learning algorithm i.e., Convolution Neural Networks these models were mostly developed for the classification of images (2D or 3D) where the model understands an interior representation of a 2D input, this process is known as feature learning. It is capable of working with signal processing also such as 1D signals.