Work

Sitting Pressure Distribution Monitoring System Using K-Nearest Neighbor and Rule-Based Logic

Year: 2026

Type: article

Abstract: An embedded sitting-pressure monitoring system is developed to identify asymmetric pressure distribution and detect prolonged static sitting. Four FSR402 sensors connected to a 12-bit ADC on an ESP32 measure seatpressure, using Left_percent, Right_percent, and P_ratio as classification features. A K-Nearest Neighbor (KNN) classifier (K = 5) is implemented directly on the ESP32 to classify Balanced, Lean Left, and Lean Right postures. Classification results are combined with duration-based rules to determine NORMAL, WARNING, and HIGH RISK conditions, supported by vibration feedback, Firebase Realtime Database, and a Flutter application. The results show an average relative pressure measurement error of 1.05% (98.95% accuracy), with pressure shifts of 17.32–18.29%. KNN achieves 98.89% mean cross-validation accuracy and 97.78% test accuracy, with a macro F1-score of 0.9776, while end-to-end posture detection achieves 90% accuracy. Haptic feedback delays range from 0.784 to 1.127 seconds. These findings demonstrate the feasibility of implementing KNN-based posture classification directly on an ESP32 and integrating it with duration-based rules for real-time monitoring of asymmetric and prolonged static sitting.

Source: CHIPSET

DOI: 10.25077/chipset.7.02.164-168.2026

Authors: Athaya Thalita Jumeyza, Rizka Hadelina

Institution: Andalas University

Language: English

Topic: Context-Aware Activity Recognition Systems

Subfield: Computer Vision and Pattern Recognition

Field: Computer Science

Domain: Physical Sciences

SDG: Good health and well-being

Open Access status: diamond

Funder: Universitas Andalas