Clustering Kinematic Patterns of Runners: A Comparative Study of Hierarchical, K-Means, and Deep Temporal Clustering Based on PCA and End-to-End Learning

Document Type : Original article

Authors

1 Sports Biomechanics and Motor Behavior, Sports Science faculty, Mazandaran University, Babolsar, Iran

2 Faculty of Sport biomechanics/ Sport Sciences department/ University of Mazandaran/ Babolsar/ Iran

3 Unit of Computer Sciences, Department of Mathematical Sciences, University of Mazandaran, Babolsar, Iran

10.22037/sjrm.2025.117446.3363

Abstract

Background and Aims: Current biomechanical research lacks comprehensive investigation of deep learning clustering algorithms for identifying homogeneous running movement patterns. This study compares PCA-based hierarchical clustering and K-means algorithms with an end-to-end deep temporal clustering approach, using ankle joint kinematic data to group runners with similar movement patterns.

Materials and Methods: Three-dimensional ankle joint angles were obtained from 108 healthy adults (age: 22.45 ± 2.42 years, height: 1.69 ± 0.11 m, body mass: 64.64 ± 9.54 kg, sex: 55 males, 53 females) during barefoot running. Deep temporal clustering, hierarchical clustering, and K-means algorithms were trained using ankle joint angles during running. Following clustering, the performance and accuracy of each algorithm in identifying homogeneous movement clusters were evaluated using clustering validation metrics.

Results: In cluster separation, the deep temporal clustering algorithm demonstrated superior performance and output accuracy compared to the other two algorithms, achieving a silhouette score of 0.74 and a DB index of 0.35. This algorithm identified three distinct clusters with a clustering inconsistency rate of 2%.

Conclusion: The findings indicate that the deep temporal clustering algorithm outperforms traditional methods such as hierarchical clustering and K-means in identifying homogeneous movement patterns among runners. Its higher accuracy and lower learning error make it a suitable choice for analyzing kinematic data in biomechanical research. These results can enhance the understanding and analysis of movement patterns and contribute to the development of effective strategies for prescribing targeted interventions.

Keywords: Running, Ankle joint angles, Deep temporal clustering, Hierarchical clustering algorithm, k-means clustering.

Keywords

Main Subjects

The Scientific Journal of Rehabilitation Medicine

Articles in Press, Accepted Manuscript
Available Online from 21 May 2025

Files

History

  • Receive Date: 04 May 2025
  • Revise Date: 07 May 2025
  • Accept Date: 21 May 2025
  • First Publish Date: 21 May 2025

Share

How to cite

Statistics