Ilmu Komputer & AI editorial
Open AccessOA2025
A Smart System for Continuous Sitting Posture Monitoring, Assessment, and Personalized Feedback
User-tailored pressure sensing and GPT-4o powered feedback for healthier sitting habits
D. F. Odesola; J. Kulon; Shiny Verghese; A. Partlow; Colin Gibson· Italian National Conference on Sensors· 2025· DOI 10.3390/s25185610
The core problem
Prolonged sitting and poor posture are widespread among adults and the working population, contributing to musculoskeletal disorders and long-term health conditions. Existing approaches often rely on generalized datasets from multiple healthy participants, which may not capture individual physiological and musculoskeletal differences. This study addresses this gap by proposing a novel smart-sensing chair system that collects data from a single individual to tailor posture classification and feedback. The system aims to encourage better postural habits and promote well-being through continuous monitoring and actionable insights.
Innovation
Among the five models, CNN achieved the highest classification accuracy of 98.29% for the 19 sitting postures. The other models (DT, RF, SVM, KNN) were also evaluated, but CNN outperformed them. The system successfully collected and processed pressure data from the two 32×32 sensor mats. The SitWell platform, comprising mobile and web applications, was developed to provide posture classification, duration analytics, and sitting quality assessment. The integration of GPT-4o enabled personalized insights and recommendations based on users' historical posture data.
Prolonged sitting and poor posture are widespread among adults and the working population, contributing to musculoskeletal disorders and long-term health conditions. Existing approaches often rely on generalized datasets from multiple healthy participants, which may not capture individual physiological and musculoskeletal differences. This study addresses this gap by proposing a novel smart-sensing chair system that collects data from a single individual to tailor posture classification and feedback. The system aims to encourage better postural habits and promote well-being through continuous monitoring and actionable insights.
The proposed system integrates two 32×32 pressure sensor mats into an office chair to collect postural data. A user-tailored dataset was gathered from a single individual to account for unique physiological characteristics. Five machine learning models were trained to classify 19 distinct sitting postures: Decision Tree (DT), Random Forest (RF), Support Vector Machine (SVM), K-Nearest Neighbors (KNN), and Convolutional Neural Networks (CNN). The CNN model achieved the highest accuracy of 98.29%. The system architecture is illustrated below:
Why it matters
The user-tailored approach demonstrates the potential of personalized machine learning models for posture monitoring, as it accounts for individual physiological differences that generalized datasets may overlook. The high accuracy of CNN suggests that deep learning is well-suited for pressure map classification. The SitWell platform's integration of GPT-4o enhances user engagement by delivering actionable, personalized feedback, which is crucial for long-term behavior change. However, the study's reliance on a single individual limits generalizability; future work should validate the approach across larger and more diverse populations. Additionally, the system's effectiveness in improving posture over time and its impact on health outcomes warrant further investigation. The proposed architecture can be extended to other applications in ergonomics and remote health monitoring.
Who should read this
CS practitioners and researchers
Opening member content…