Smart Total Knee Replacement: ADL Recognition Using Dual IMU Sensors

Published in Sensors (2025), this study presents the first dual-IMU smart knee replacement capable of recognizing activities of daily living directly from sensors embedded inside the femoral and tibial components of a knee prosthesis.

Using a modified posterior-stabilized implant with powered IMUs embedded inside custom chambers, data was streamed at 50 Hz during simulated walking, sitting, standing, stair ascent/descent, jogging, and knee bending.

A novel dual-branch AI model combining LSTM (time-series) and CNN spectrogram analysis achieved 95.7% test accuracy, correctly identifying key activities with 100% accuracy for knee bending, sitting, standing, and stance. Walking and stair behaviors were recognized with F1 scores up to 0.98, demonstrating reliable pattern recognition even for dynamic tasks.

This proof-of-concept establishes the foundation for fully integrated, multi-sensor smart prosthetics capable of real-time functional monitoring and personalized post-operative rehabilitation.