Discrimination of gender-, speed-, and shoe-dependent movement patterns in runners using full-body kinematics

Abstract 

Changes in gait kinematics have often been analyzed using pattern recognition methods such as principal component analysis (PCA). It is usually just the first few principal components that are analyzed, because they describe the main variability within a dataset and thus represent the main movement patterns. However, while subtle changes in gait pattern (for instance, due to different footwear) may not change main movement patterns, they may affect movements represented by higher principal components.

This study was designed to test two hypotheses: (1) speed and gender differences can be observed in the first principal components, and (2) small interventions such as changing footwear change the gait characteristics of higher principal components.

Kinematic changes due to different running conditions (speed – 3.1m/s and 4.9m/s, gender, and footwear – control shoe and adidas MicroBounce shoe) were investigated by applying PCA and support vector machine (SVM) to a full-body reflective marker setup.

Differences in speed changed the basic movement pattern, as was reflected by a change in the time-dependent coefficient derived from the first principal. Gender was differentiated by using the time-dependent coefficient derived from intermediate principal components. (Intermediate principal components are characterized by limb rotations of the thigh and shank.) Different shoe conditions were identified in higher principal components.

This study showed that different interventions can be analyzed using a full-body kinematic approach. Within the well-defined vector space spanned by the data of all subjects, higher principal components should also be considered because these components show the differences that result from small interventions such as footwear changes.

Keywords: Gait analysis, Principal component analysis, Support vector machine, Principal movement patterns, Full-body marker setup