Monitoring and Classification of Human Sleep Postures, Seizures, and Falls From Bed Using Three-Axis Acceleration Signals and Machine Learning

作     者：Intongkum, Chawakorn Sasiwat, Yoschanin Sengchuai, Kiattisak Booranawong, Apidet Phukpattaranont, Pornchai 

作者机构：Department of Electrical and Biomedical Engineering Faculty of Engineering Prince of Songkla University Songkhla 90110 Thailand 

出 版 物：《SN Computer Science》 (SN COMPUT. SCI.)

年 卷 期：2024年第5卷第1期

页      面：104页

基　　金：Faculty of Engineering  Prince of Songkla University  PSU ENG 

主　　题：3-Axis accelerometer Classification Falls Monitoring Sleep postures 

摘      要：A system for monitoring and classifying human activity and sleep postures in bed using three-axis acceleration signals is presented in this paper. In this low-cost system, a three-axis accelerometer sensor using a single GY-521 sensor is placed on the abdominal muscles of the human body to measure human activity and sleep posture signals. The sensor is connected and communicates with the Arduino Mega for processing. Focused activities and sleep postures in bed, including (a) sleeping on his back, (b) turning around to sleep on his side, (c) sleeping on his side, (d) turning around and falling from the bed, (e) lying on the ground, and (f) seizure sleeping, are tested and evaluated. Finally, signal feature extraction using thirty-five features from three groups of calculations and classification using a K-nearest neighbors (KNN) algorithm are applied. Experiments are conducted in a laboratory, and results indicate that the proposed system could automatically monitor different human postures in bed and falls in real-time. Acceleration signals measured from activities and sleep postures have their own patterns and characteristics. Additionally, the average classification accuracy using the best four features obtained 93% for the postures a) to f) of 97.5%, 73.9%, 98.4%, 88.7%, 98.4%, and 91.1%, respectively. Here, before falls, falls, and seizures can be accurately detected. Our system and results can thus be used to support caretakers, physicians, and medical staff in the evaluation, planning, and treatment of the elderly and patients in healthcare systems. © 2023, The Author(s), under exclusive licence to Springer Nature Singapore Pte Ltd.