Hide/Show Apps

Verification and MATLAB implementation of the inverse dynamics model of the METU gait analysis system

Erer, Koray Savaş
The METU Gait Analysis System employs a computer program called Kiss-GAIT for the calculation of joint angles, moments and powers using force plate data and marker trajectories as input. Kiss-GAIT was developed using Delphi and is confined to calculations related to the standard gait protocol. Because the code lacks the flexibility required to carry out various test cases, the inverse dynamics formulation being used could not be verified and the extent of the error propagation problem could not be determined so far. The first aim of this study was to develop a code for the inverse dynamics model of the METU Gait Analysis System making use of the flexible programming environment provided by MATLAB. Verified and more reliable analysis results, obtained by reformulating the inverse dynamics algorithm in a new code, are presented. Secondly, data smoothing and differentiation techniques conventionally used in gait analysis were critically reviewed. A common tool used for filtering marker trajectories is the Butterworth digital filter. This thesis presents a modified, adaptive version of this classical tool that can handle non-stationary signals owing to its coefficients which are functions of local signal structure. The results of this thesis indicate the dominancy of ground reactions as compared to inertial effects in normal human gait. This implies that the accuracy needed in body segment inertial parameter estimation is not a critical factor. On the other hand, marker trajectories must be as accurate as possible for meaningful kinetic patterns. While any smoothing and differentiation routine that produces reasonable estimates is sufficient for joint moment calculation purposes, the estimation performance becomes a key requirement for the calculation of joint powers.