INTELLIGENT DUAL-LEG WEARABLE FOR EARLY ARTHRITIS SCREENING VIA GAIT ANALYSIS AND ON-DEVICE MACHINE LEARNING

Early identification of arthritis can markedly improve long term mobility and clinical outcomes. This paper reports the design and proof-of-concept evaluation of a low-cost, 3D-printed, knee-mounted wearable that screens for early signs of arthritis from gait. The system comprises two compact modules one on each leg each integrating an ADXL335 tri-axial accelerometer and an Arduino Nano 33 BLE microcontroller. The modules acquire and wirelessly stream bilateral gait signals in real time, enabling continuous assessment of walking behaviour. Models are trained with Googles Tiny Motion Trainer and deployed for on-device inference to classify gait into three clinically meaningful categories: No Arthritis, Moderate Arthritis, and Major Arthritis. A mobile interface provides live visualization and logging, while multimodal biofeedback via a buzzer, vibration motor and NeoPixel indicator alerts users to abnormal patterns and encourages corrective action during everyday ambulation. The prototype demonstrates stable data acquisition, reliable Bluetooth Low Energy transmission and responsive, real-time classification, highlighting the feasibility of combining wearable sensing, TinyML and biofeedback for preventive musculoskeletal care.This work motivates larger studies to validate diagnostic performance across diverse populations and to refine feature extraction and model tuning for robust, at-home screening and rehabilitation support.