Development of a Smart Chair for Real-Time Posture Detection to Promote Ergonomic Habits in Sedentary Workspaces

Abstract

Prolonged sitting in academic and office settings is strongly linked to occupational musculoskeletal disorders, including lower back pain, neck tension, shoulder stiffness, and spinal strain. Traditional ergonomic interventions often fail because they rely on user self-regulation, which diminishes during periods of high cognitive load or fatigue. This study presents the design, empirical evaluation, and socio-technical assessment of a smart chair with sitting posture detection. The system combines six force sensitive resistor (FSR402) sensors mounted on a seat cushion, an ESP32 microcontroller, a 128x64 OLED display, and an active piezo buzzer for real-time edge computing and immediate corrective feedback. The system classifies six distinct sitting postures: slouching, forward leaning, side leaning right, side leaning left, sliding forward, and crossed legs. Using a K-Nearest Neighbors algorithm trained on 660 observations across 11 participants, the offline validation accuracy reached 93.8% with an average Receiver Operating Characteristic Area Under the Curve of 0.9627. Real-time testing on 5 unseen participants achieved a classification accuracy of 67.58% post-feature engineering. By integrating real-time sensory feedback into daily seating, this low-cost innovation advances United Nations Sustainable Development Goal 3 (Good Health and Well-Being) and provides an accessible framework for workplace and educational health management