Abstract

Motion capture systems are frequently used in biomechanics analysis. Doubtless, the most important class for this use is which one that is composed of optical motion capture systems. The optical systems have some significant advantages such as independence of number of channels (each channel represents the trajectory of one mark in time), better precision in data input and less noise due to magnetic fields and other types of distortion noises. In spite of their advantages, motion capture systems require an important effort in post-processing, because one has to clean four basic error types; discrete channel error, interchange of channels, erroneous trajectory and precision mistakes. The correction of these captures is automatically completed by the use of cubic splines or another type of interpolation curve,like cardinal splines. Nevertheless, the detection of this variety of errors is frequently made with human intervention, so that it develops into the essentially speed problem of this type of systems. The authors' work presents a general system that is able to identify and repair automatically all types of errors in optical motion capture data. The authors have made an expert knowledge model, characteristically qualitative, imprecise and inexact, which we use to detect the errors through fuzzy rule systems. Finally, when the necessary reconstruction parameters are available, this process is finished by means of classic techniques of curve examination such as interpolation curves (Splines) and numerical methods. Einführung (geändert)