Journal DOI 10.68956/journal.human-performance-futures

Creative Commons source reading · Movement biomechanics and kinetics

Dual-Model Artificial Intelligence Framework Integrating AI-Based Markerless Motion Capture for Dynamic Gait Prediction and Health-Status Classification

Edder Jair Rodríguez-Granados; Guillermo Urriolagoitia-Sosa; Beatriz Romero-Ángeles; Jorge Alberto Gomez-Niebla; Jonathan Rodolfo Guereca-Ibarra; Maria de la Luz Suarez-Hernandez; Manuel Nazario Rocha-Martinez; Eduardo Enrique Carmona-Hernández; Luis Itzcoatl Lugo-Chacon; Gabriela Ramirez-Sanchez

Originally published in Diagnostics (Basel) · 2026-08-16

From the article

Selected source text. JournalOS removes markup and shortens excerpts where indicated. The full article provides the complete context.

Abstract

Background/Objectives: Human gait analysis is essential for identifying biomechanical alterations associated with pathological conditions. However, conventional laboratory systems that combine optical motion capture and force plates remain costly, space-demanding, and difficult to implement in routine or accessible settings. This study proposes a dual-model artificial intelligence framework designed to bridge kinematics and dynamics and subsequently support gait health-status classification from motion data. Methods: The first model was developed to estimate ground reaction forces (GRF) and center of pressure (CoP) signals from kinematic inputs. Public datasets containing synchronized kinematics and dynamics were used to train and evaluate long short-term memory (LSTM) and one-dimensional convolutional neural network (CNN1D)…

4. Discussion

This study proposed and evaluated a dual-model artificial intelligence framework for gait analysis, designed to connect kinematic acquisition, dynamic prediction, and gait health-status classification within a single workflow. The main findings can be summarized in five points. First, Model 1 demonstrated that ground reaction forces (GRF) and center of pressure (CoP) can be estimated from kinematic inputs with acceptable error levels, particularly when using a long short-term memory (LSTM) architecture. Second, Model 2 confirmed that normalized dynamic signals can be used for gait classification, and that the inclusion of metadata improves classification performance. Third, the bridge block allowed the two models to be connected operationally by transforming continuous predicted dynamics into the 101 × 10 representation required by…

5. Conclusions

This study presented a dual-model artificial intelligence framework for gait analysis that connects kinematic acquisition, dynamic prediction, and gait health-status classification within a single workflow. The results showed that Model 1 can estimate bilateral ground reaction forces (GRF) and center of pressure (CoP) from kinematic inputs, with the Long Short-Term Memory (LSTM) architecture providing the best performance among the tested dynamic prediction models. The classification stage confirmed that dynamic signals contain useful information for distinguishing healthy-control and non-healthy-control gait patterns. Although the rich-metadata configuration achieved the highest classification performance, the permutation importance analysis showed that much of its additional contribution was associated with acquisition and clinical…

Editorial reading guide

  • How was movement measured, and what error or calibration information is reported?
  • Can the finding transfer from the tested task and population to the intended sport setting?