Knee Kinematic Signals Clustering for the Identifcation of Sagittal and Transverse Gait Patterns

The purpose of this study is to investigate knee kinematic signals clustering by principal component analysis. The aim is to identify meaningful patterns in normal gait knee flexion/extension and tibial internal/external rotation signals. To preserve all of the information contained in these kinematics signals, the analysis uses the entire angle curve over a gait cycle rather then a few features extracted from this curve as done traditionally. To reduce processing complexity, the data dimensionality is reduced without loss of relevant information by projecting the gait curve onto a subspace of significant principal components (PCs). Gait patterns are then extracted by a discriminant analysis of the set of training data based on the PCs sign. The analysis identified two representation patterns for each of the flexion/extension (sagital plane) and the tibial internal/external rotation (transverse plane). These patterns were validated both by the clustering silhouette width and clinical interpretation.