Joint location error in two-dimensional cinematographic analysis

Wu-Chou Chen, Purdue University

Abstract

Twenty individuals with previous experience in digitizing the segmental endpoints of the human body from film records were used as subjects (plotters) in this study. To control the digitizing image, two 35mm positive slides were used: one with marked segmental endpoints; the other one with unmarked segmental endpoints. The slides were projected to create images that were 5%, 15%, and 30% (small, medium, large) relative to the real height of the individual in the slide. Eight joints were specified for the analysis: right and left knee, hip, elbow, and shoulder joints. The joints were arranged into two paired conditions: the bent versus straight joint condition; the visible versus hidden joint condition. Data were collected for the variation of the X coordinates, variation of the Y coordinates and for accuracy, which was the deviation (D) from the actual segmental endpoint $$\rm(D = \sqrt{(x - x\sb0)\sp2+(y - y\sb0)\sp2}).$$Statistical analyses were made using SAS version 6.07. The within and between plotter variations and accuracy were analyzed for the image size effect by holding the joint condition fixed. Subsequently, the image size by joint condition interaction and the paired joint condition effects were analyzed. In studying the within plotter variations, the plotters seemed to be more consistent for the Y coordinate than for the X coordinate. The plotters appeared to be more consistent for the medium and large image sizes than they were for the small image size. In studying the between plotter variations, the image size appeared to affect the plotter's identification of the actual location of the segmental endpoints for the small-large image pairs and the small-medium image pairs. The plotters, however, were not affected by the medium-large image pairs. In studying the digitizing accuracy, D, the plotters appeared to be more accurate for the medium and large image sizes than they were for the small image size. Significant image size by joint condition interactions were found for the X and Y coordinates and for the D values. In the analysis of the paired joint conditions, the plotters seemed less affected by the joint condition for the medium and large image sizes than they were for the small image size and less affected by the visible-hidden joint conditions pairs than they were by the bent-straight joint condition pairs. As a result of the findings of this study, film images used for cinematographical analysis should be at least 15% of the actual height of the subject and special care should be given to digitizing specific joints.

Degree

Ph.D.

Advisors

Widule, Purdue University.

Subject Area

Physical education|Educational software|Anatomy & physiology

COinS