PosTer: A Smart Lumbar Posture Correction Vest Using a Fully Connected Neural Network Algorithm
TL;DR
PosTer is an effective assistive tool for real-time lumbar posture monitoring and correction in short-term, non-clinical, sedentary settings, although its restricted vest sizing and dependence on correct sensor placement limit its suitability for long-term or clinical therapeutic use.
Abstract
Maintaining proper lumbar posture is important for preventing musculoskeletal disorders among sedentary individuals, yet conventional interventions such as exercise and physical therapy are often impractical for sustained daily use, and existing wearable posture aids remain limited in sensor integration, adaptability, and real-time intelligence. This study developed and evaluated PosTer, a smart lumbar posture correction vest that combines Inertial Measurement Unit (IMU) and flex sensors with an ESP32-based web interface and a Fully Connected Neural Network (FCNN) for real-time posture classification, providing visual, auditory, and automatic corrective feedback. A quantitative developmental design was used; the system was tested by 126 purposively selected office-based employees of a Business Process Outsourcing (BPO) company in Calamba City, Philippines, representing individuals in low-intensity, sedentary work. Sensor readings and an ISO/IEC 25010-based survey were used to evaluate classification performance and system quality. The FCNN achieved 99.43% accuracy with precision, recall, and F1-score of 0.985, outperforming Logistic Regression (93.40% accuracy) and Support Vector Machine (99.33% accuracy). ISO/IEC 25010 evaluation showed strong mean ratings for functional suitability (3.84), performance efficiency (3.91), usability (3.83), reliability (3.68), and safety (3.80). Participants reported that the web interface was clear and that the vest responded promptly to posture deviations. These findings indicate that PosTer is an effective assistive tool for real-time lumbar posture monitoring and correction in short-term, non-clinical, sedentary settings, although its restricted vest sizing and dependence on correct sensor placement limit its suitability for long-term or clinical therapeutic use. The study provides a foundation for further work on sensor adaptability, extended-use evaluation, and mobile-based posture analytics.