Bad Sitting Posture Detection and Alerting System using EMG Sensors and Machine Learning

Abstract

Poor sitting posture can lead to a variety of serious diseases raging from spinal disorders to psychological stress. This paper aims to design a sitting posture monitoring system that detects improper postures and notifies the user in real time through a mobile application. The system leverages the use of low-cost EMG sensors, and relies on energy-efficient communication via Bluetooth Low energy (BLE). To ensure bad posture detection, different machine learning algorithms are tested and compared, namely support vector machine (SVM), K-nearest neighbours (KNN), decision tree (DT), random forest (RF), and multi-layer perception (MLP). We formulated the problem as a binary classification (good vs. bad posture) and multi-class classification (good, tilted to the front, right and left). The results of the training performed on a real dataset showed that KNN have the best accuracy (91% accuracy) and execution time (0.0066 ms).