Real-Time Sleep Apnea Monitoring and Alert Mechanism using Machine Learning and Embedded Systems

Apnea is a sleeping disorder that is common and serious because it happens when one has breathing interruptions when asleep. When left untreated, it may cause severe health problems, such as heart disease, tiredness and thinking difficulties. The classic methods in the diagnosis include polysomnography (PSG), which proves to be costly and cannot be utilized in application in long term continuous monitoring as well as home-based monitoring because they require clinical monitoring in addition to the use of complex machinery. In order to overcome these constraints, this paper presents an internalized electronics and machine learning (ML) integrated AI based real-time sleep apnea sensors and warning system. However, in this system a camera is involved in the examination of the facial features of the user during the night and eye blinks and yawns are the behavioral parameters of the user defined to generate the sleep irregularities. The captured video frames are then subjected to the classifier which is an ML based classifier that determines the abnormal blink frequency and the yawning pattern typical of the apnea episodes. Upon detection, an AB serial command is sent to an Arduino microcontroller to activate a vibration motor and buzzer to provide instant haptic and sound feedback and a warning message is displayed on an LCD. The design used is non-invasive and thus without the need of wearable sensors, The system supports real-time operation with low latency under controlled conditions and can be extended to provide cloud-based analytics for long-term analysis of sleep behavior. Overall, this inexpensive and user-friendly solution demonstrates the feasibility of AI-assisted, non-intrusive sleep health risk monitoring.