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.
Jin et al. (Wed,) studied this question.