AI Powered Wearable for Posture Correction

Authors:

Harshitha K, VR Thilaka Sundar, Aaditi Bhale, Dr. Debarati Nath

Department of Electronics and Communication Engineering

Faculty of Engineering and Technology, SRM Institute of Science and Technology

Kattankulathur, Chennai, Tamil Nadu 603203, India Email: {hk3880, tv4345, ab6321, debaratn}@srmist.edu.in

Abstract—Prolonged sedentary behaviour due to occupational hazards is a major cause of musculoskeletal strain. Conventional posture monitors rely on rigid accelerometer thresholds, which causes frequent false-positive alerts during natural physical movements, as they are context unaware. We present a novel, 3D-printed wearable posture-correction system utilising TinyML- based edge computing on an ESP32 microcontroller with a single cervical MPU6050 6-axis IMU. The primary novelty of this work is its highly robust, user-agnostic machine learning model. Unlike traditional devices requiring manual calibration, our system demonstrates height-invariant accuracy in classifying three physical states: ergonomic alignment, kyphotic curvature, and dynamic movement. The system works on isolating and recognising differences between active movements and static postures. Through the BLE power of the microcontroller, the wearable also functions as a wireless Human Interface Device (HID) over Bluetooth. Upon detecting sustained slouching, it autonomously transmits a system-level lock command to the user’s workstation, replacing easily ignored auditory alerts with an active behavioural conditioning loop. By detailing our hard- ware design, data gathering, and model validation, we present a practical, scalable framework for everyday postural health.

Index Terms—Embedded systems, wearable computing, ESP32, MPU6050, inertial measurement unit (IMU), TinyML, edge computing, Bluetooth Low Energy (BLE), Internet of Medical Things (IoMT), posture correction.