Methods for analysis of 3D movement of the trunk during treadmill locomotion

Abstract

Human locomotion is a common activity and although much progress has been made in the analysis and modeling of movement, we still have a rather limited understanding of the basic mechanisms of normal locomotion. A full understanding requires the integration of the underlying neural mechanisms with the 3D dynamics (kinetics + kinematics) of locomotion.

The trunk, the most proximal and massive body segment, clearly must play a key role in the dynamics of human locomotion. Most studies on the trunk have concentrated on its kinematics in the sagittal plane (2D). However. during normal locomotion the pelvis and thorax both show very complex spatial (3D) movement patterns. and thus planar analvsis is often inappropriate for extraction of pattern features. The underlying goal of this work was to obtain a better understanding of the role of the trunk (especially the pelvis and thorax) in locomotion.

This thesis has focused on one aspect of this goal: the development of methods to extract reliable 3D kinematic information on the movements of the pelvis and thorax during treadmill locomotion. A methodology was developed for the collection of the 3D kinematics of the pelvis and trunk. A footswitch device was constructed to measure the temporal relationship between events of the stride-cycle (e.g. foot-contacts) and the position and orientation of the trunk. A 2-camera Selspot system was used to capture 2D images of the 3D trajectories of both single markers and clusters attached to the trunk. The 2D images of the markers from both cameras were mathematically transformed into 3D trajectories, and used to estimate rotations and translatios of the clusters for different walking speeds. These 3D kinematic data showed that the total range of rotations for the thorax was approximatelv constant as walking speed changed which was quite different from corresponding rotational changes in the pelvis.

In addition it was shown that the fundamental period of gait can be extracted from the trajectory of a single marker (located near the center-of-mass) even when disturbances are present. A new method for the processing of multiple stride cycles. based on singular-value decomposition. was introduced and shown to be useful for the identification of dominant patterns in 3D trajectories. 3D kinematic data are: multidimensional, quasi-periodic, contain disturbances. missing data gaps, low-frequency trends (drift). and can be very large in size. The extraction of the actual underlying 3D movement patterns is thus a challenging signal processing task. Therefore, an important part of this study was the development of improved and effective algorithms for the processing and analysis of 3D kinematic data. These algorithms have been carefully implemented in PASCAL, thoroughly tested, and shown to be useful for the processing of large 3D kinematic data sets. All software used in the studies is available from the author upon request.