Scholarlink

<p>The analysis of human movement data in sports science is often challenged by the inherent variability in movement speed and rhythm, which results in gait time-series data of inconsistent lengths (dynamic dimensionality). This poses a significant obstacle for traditional optimization algorithms in constructing accurate motion templates for performance analysis and rehabilitation. To address this, we propose a novel Dynamic Dimension Warping (DDW) algorithm specifically designed for efficient search in dynamic multidimensional spaces. DDW integrates a Cross-Dimensional Mapping (CDM) mechanism, fusing Dynamic Time Warping and Euclidean distance to enable comparison between variable-length sequences, and an Optimal Dimension Collection (ODC) method to break fixed-dimension constraints. When applied to the task of optimizing human gait templates from experimental data, DDW demonstrated superior performance against 31 benchmark algorithms, reducing average fitness to 9.16 (41% below mean) and achieving rapid convergence within 10 generations. The algorithm also attained global optima in 52.17% of classical function tests, confirming its robustness. This work establishes DDW as an effective optimization framework for complex, dynamic-dimensional problems, with direct methodological value for gait analysis and biomechanical motion assessment.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Optimizing gait template generation from variable-length data: a dynamic dimension warping approach

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月13日

查看全文
  • Dongnan Jin,
  • Yali Liu,
  • Qiuzhi Song,
  • Zhenpeng Guan,
  • Xunju Ma,
  • Yue Liu,
  • Dehao Wu

摘要

The analysis of human movement data in sports science is often challenged by the inherent variability in movement speed and rhythm, which results in gait time-series data of inconsistent lengths (dynamic dimensionality). This poses a significant obstacle for traditional optimization algorithms in constructing accurate motion templates for performance analysis and rehabilitation. To address this, we propose a novel Dynamic Dimension Warping (DDW) algorithm specifically designed for efficient search in dynamic multidimensional spaces. DDW integrates a Cross-Dimensional Mapping (CDM) mechanism, fusing Dynamic Time Warping and Euclidean distance to enable comparison between variable-length sequences, and an Optimal Dimension Collection (ODC) method to break fixed-dimension constraints. When applied to the task of optimizing human gait templates from experimental data, DDW demonstrated superior performance against 31 benchmark algorithms, reducing average fitness to 9.16 (41% below mean) and achieving rapid convergence within 10 generations. The algorithm also attained global optima in 52.17% of classical function tests, confirming its robustness. This work establishes DDW as an effective optimization framework for complex, dynamic-dimensional problems, with direct methodological value for gait analysis and biomechanical motion assessment.