DATA-DRIVEN PHASE SEGMENTATION AND LIMB COORDINATION ANALYSIS OF WRESTLING LEG ATTACK TECHNIQUES: DIFFERENCES BETWEEN ELITE AND SUB-ELITE ATHLETES

Author(s): XU, N., LIU, W., MA, Y., ZHOU, Y., BAI, Y., WU, H., Institution: BEIJING SPORT UNIVERSITY , Country: CHINA, Abstract-ID: 2780

INTRODUCTION:
Leg attacks are fundamental offensive techniques in wrestling, characterized by high-speed, aperiodic movement. Traditional analyses rely on subjective phase segmentation and isolated kinematics, failing to capture temporal structure and inter-limb coordination underlying skilled performance. From an ecological dynamics perspective, skilled movement emerges through functional coupling of body segments into synergies [1]. However, objective methods to quantify phase structure and coordination in combat sports remain limited [2]. This study aimed to: (1) establish data-driven phase segmentation using Hidden Markov Models (HMM) [3]; (2) extract coordination patterns using Non-negative Matrix Factorization (NMF) [4] ; and (3) compare differences between elite and sub-elite wrestlers.
METHODS:
Forty-six female freestyle wrestlers (24 elite, 22 sub-elite) were paired within skill levels and performed single-leg and double-leg takedowns (5 trials each) under partner-present cooperative contexts. Full-body kinematics were captured using inertial motion capture (Xsens Link, 240 Hz). A 5-state Gaussian HMM was fitted for each task using 14 kinematic features [5]. Within each state, NMF (k = 8) extracted coordination patterns from segment angular velocities [4]. Activation time-series were parameterized by onset, duration, amplitude, and CV. Group comparisons used Welch’s t-tests.
RESULTS:
HMM segmented both tasks into five phases: Preparation-Dive-Grab-Lift-Control. Elite wrestlers showed higher CoM height during Grab (0.57 ± 0.05 vs 0.53 ± 0.04 m, p<0.001) , greater L5S1 flexion (12.8 ± 3.2 vs 8.1 ± 2.9deg, p<0.001), and stronger head rotation toward attack side (-9.4 ± 4.1 vs -0.3 ± 5.2deg, p<0.001). NMF revealed proximal-to-distal coordination with components for propulsion, trunk control, and grasping. Elite wrestlers showed: (1) higher activation amplitude (20-90% greater, p<0.001); (2) earlier onset timing (35-50% reduction); (3) longer Grab duration (0.92 ± 0.08 vs 0.82 ± 0.1, p<0.001); and (4) a “stable core-flexible periphery” CV pattern.
CONCLUSION:
The HMM-NMF framework provides an objective approach to characterize temporal structure and coordination in non-periodic combat techniques. Elite wrestlers exhibit distinctive motor signatures: earlier anticipatory head orientation, greater activation amplitude, compressed sequences, and differentiated stability-flexibility organization. These findings extend coordination research to partner-present cooperative contexts [2] and offer indicators for talent identification.