A Physiotherapy Video Matching Method Supporting Arbitrary Camera Placement Via Angle-Of-Limb-Based Posture Structures
Jiunn-Wu Lin, Yao-Sheng Chou, YUN PIN HUANG, Min-Hsiung Hung, Ming-Hung Kao, Jia Ji, Lin-Yi Jiang, Pi Wei Chen, Chao-Chun Chen
AI summary
Problem
Remote physiotherapy assessment is hindered by viewpoint discrepancies when patients record exercises from arbitrary angles, making it difficult to accurately compare patient movements to mentor demonstrations. Existing methods rely on fixed camera setups and lack robustness for real-world, unconstrained recording conditions.
Approach
The method transforms 2D/3D keypoints into a camera-angle-free, angle-of-limb-based posture structure (ALPS) and uses a three-phase matching algorithm to align mentor and patient videos without requiring synchronized viewpoints.
Key results
- ALPS representation for viewpoint-invariant posture encoding
- CAFE transformation to eliminate camera-angle dependencies
- TALMA algorithm for efficient, structure-aware video matching
- Sub-0.07-second alignment error against expert annotations
Why it matters
Enables reliable, scalable remote physiotherapy assessment for aging populations and Hospital-at-Home initiatives by removing the need for controlled recording setups.
Abstract
The “Hospital at Home” initiative transforms medical service automation through modern technologies. This paper revisits remote physiotherapy, allowing convalescents to record exercises using mobile devices from arbitrary angles. To address this, we propose a physiotherapy video matching method that accurately aligns movements from unconstrained viewpoints. The task is formulated as an optimization problem and solved using a modular pipeline. We introduce the Angle-of-Limb-based Posture Structure (ALPS) and the Camera-Angle-Free (CAFE) transformation to counter camera-angle differences. We also develop the Three-phase ALPS Matching Algorithm (TALMA) for matching movements between mentor and convalescent videos. Real-world experiments show our method outperforms existing solutions in both precision and practicality, with a time deviation of less than 0.07 seconds from expert anno- tations. The prototype and datasets are publicly available at: https://github.com/NCKU-CIoTlab/TALMA-on-ALPS/.