Sammendrag
To bring quantitative infant movement analysis into clinical use, an easy-to-use, portable unit for motion capture is needed. In this work, a measurement system based on a single camera is analyzed and the significant parameters with respect to measurement precision are identified.
Movement data from infants at high risk of developing neurologial dysfunctions has earlier been collected with a magnetic tracker. A classifier that is able to distinguish between normal and abnormal movement patterns was developed. Seen as the data collecting component of such a classification system, the camera parameters´ impact on correct classification is investigated by simulating the effect a camera would have on existing 3D infant movement data using a pinhole model of the camera.
The results indicate that the data collected by a single-camera measurement system still enables the classifier to distinguish between normal and abnormal infants if a camera with reasonable resolution (>=640x480) and frame rate (>=25 fps) is chosen.
The influence of the tracking algorithm has been partially investigated, as noise such as jitter and backlash has been simulated. The effects of temporary marker occlusions must be further tested, but are believed to be small because of the ergodic nature of the signal we are measuring and the features used being insensitive to time series length. Band markers reduce the occlusion problem, but the data produced will be slightly different than by using point markers. This difference is small compared to overall movements, and is most likely negligible.
Vis fullstendig beskrivelse