Demonstrating the estimation accuracy of ground reaction forces, joint moments, and joint angles without optical motion capture or force plates — joint research with Saitama Prefectural University

AcceleBody Co., Ltd. (Headquarters: Toshima-ku, Tokyo; Representative Directors: Haruo Aoki [CTO] and Ryota Nakamura [CEO]), developer of the gait analysis system GaitForce, is pleased to announce that a paper from our joint research with Saitama Prefectural University was published online on July 10, 2026 in Gait & Posture (Elsevier), a peer-reviewed international academic journal. The study validated that three-dimensional gait motion, joint angles, ground reaction forces, and lower-limb joint moments can be estimated using only a single RGB-D camera* and physics-based optimization* — without dedicated optical motion capture* or force plates*. This provides objective evidence that the mechanical analysis of gait, which has traditionally required large-scale equipment, may be performed with a far simpler setup.

Background

“Mechanical analysis” — quantifying three-dimensional motion, ground reaction forces, and joint moments — is increasingly used in research, yet in clinical practice the assessment of movement and posture still largely relies on visual observation. Accurately measuring three-dimensional gait motion, ground reaction forces, and joint moments has traditionally required large-scale measurement environments such as optical motion capture and force plates, along with specialized expertise. The cost, installation space, and measurement effort involved have been barriers to wider adoption.

To address these challenges, this study examined whether reasonable accuracy can be achieved with a simple camera-based setup, without any of this dedicated equipment.

Study Design and Results

  • Method: Three-dimensional gait motion captured with a single RGB-D camera was analyzed using physics-based optimization grounded in a mechanical model — rather than AI-based inference — to estimate ground reaction forces, lower-limb joint moments, and joint angles.
  • Validation: The estimates were compared against measurements from optical motion capture and force plates, the standard measurement methods.
  • Results: The study confirmed that joint angles, ground reaction forces, and lower-limb joint moments can be estimated with reasonable accuracy without this dedicated equipment. (For details on the accuracy metrics and how they were computed, please refer to the paper.)

Note: This paper validates an early version of our technology. GaitForce is now offered with even higher accuracy, thanks to AB Cam — a new camera developed in-house — and a fully revamped analysis software stack (AB Cam is already available; please see our website for details).

Method overview

Ground reaction force comparison

Outlook

Making joint loads visible and measurable by anyone is at the core of our founding vision of extending healthy life expectancy. Having the accuracy of this underlying technology validated in a peer-reviewed international journal is an important step toward a new infrastructure that connects research and clinical practice.

We will start by providing easy access to this data — without dedicated equipment — not only to researchers working on motion analysis, but also to physical therapists and other healthcare professionals who want to quantitatively assess their patients’ movement and posture in clinical settings. As measurement becomes accessible across both research and clinical practice, the insights gained will flow back into the field. We believe that growing this cycle is what will ultimately lead to our long-term vision of extending healthy life expectancy.

Comment from Haruo Aoki, Representative Director & CTO

We have been working toward a world where anyone can perform motion analysis, anywhere. Having the effectiveness of our technology objectively evaluated in this paper is a major step toward that goal. The study has also clarified the remaining challenges and the direction for improvement. We will keep improving the technology by leveraging AB Cam, our in-house-developed camera, and by developing new algorithms. We would like to express our sincere gratitude to the researchers at Saitama Prefectural University for their dedication to this study.

Paper Information

  • Journal: Gait & Posture (Elsevier)
  • Title: Force-Plate-Free Estimation of Ground Reaction Forces and Lower-Limb Joint Moments During Gait Using a Single RGB-D Camera and Physics-Based Optimization
  • Authors: Keisuke Kubota, Haruo Aoki, Ryota Nakamura, Riku Suzuki, Naohiko Kanemura (in the order listed in the paper)
  • Published: July 10, 2026 (online)
  • DOI: https://doi.org/10.1016/j.gaitpost.2026.110285

*Free access link, available until September 6, 2026 (no registration or fee required): https://authors.elsevier.com/c/1nSUn3RoQ8uBmZ

Glossary

  • *Optical motion capture: A system that tracks markers attached to the body with multiple cameras to measure movement in three dimensions with high accuracy. It is the accurate, standard measurement method, but it requires a dedicated measurement environment, and both measurement preparation and post-measurement analysis take considerable time and effort.
  • *Force plate: A dedicated device that measures the forces applied to the body from the ground (ground reaction forces). Its large size imposes constraints on installation space and cost.
  • *Ground reaction force / joint moment / joint angle: The ground reaction force is the reaction force from the ground acting on the foot; a joint moment is the rotational force generated when moving a joint (the knee adduction moment (KAM) is a representative metric known to be associated with knee osteoarthritis). All of these are fundamental for assessing joint loads and movement.
  • *RGB-D camera: A camera that captures not only color images (RGB) but also the distance to the subject (Depth).
  • *Physics-based optimization: An approach that estimates the forces consistent with the observed motion based on physical models (equations of motion) and mathematical optimization theory, rather than AI that relies on training data. Unlike many AI methods, it is not a black box — its computation process and results are easy to interpret.

Company Overview

  • Name: AcceleBody Co., Ltd.
  • Address: 1-25-18 Senkawa, Toshima-ku, Tokyo, Japan
  • Representatives: Haruo Aoki, Representative Director & CTO; Ryota Nakamura, Representative Director & CEO
  • Founded: December 2023
  • Business: Development and provision of the gait biomechanics analysis system GaitForce
  • URL: https://www.accelebody.com/

GaitForce is a gait biomechanics analysis system that applies robot control technology. With a single in-house-developed AB Cam camera and a measurement of about one minute, it estimates metrics such as three-dimensional motion, ground reaction forces, joint moments (including KAM), and joint angles. It has been filed as a general medical device with Japan’s Pharmaceuticals and Medical Devices Agency (PMDA).