• A. Salo
  • P. Grimshaw
  • H. Mononen
  • J. Viitasalo


INTRODUCTION Variation is one part, which determines repeatability and reliability in biomechanical studies. Yeadon and Challis (1994) indicated that uncontrolled variation may mask the effect of the experiment. The aim of this study was to investigate the variation in 30 co-ordinates and the influence on performance variables. METHODS Hurdle clearances were videotaped with two genlocked cameras (50 Hz, at a 90 degree angle from the hurdle symmetrically on both sides of the lane). Two randomly selected trials (female and male) were digitised eight times by the same operator using APAS. DLT- and quintic spline algorithms were applied to digitised files. Consequently, raw and smoothed 30 co-ordinates and 28 kinematic variables were analysed. Standard deviation (SD) values were calculated in each case. For the co- ordinates, SD of eight repeated digitising was calculated separately for the all 18 body landmarks in every single analysed field and at each x, y, z and diagonal (combined) directions. RESULTS The mean (all landmarks, all fields) of raw 3D co-ordinates yielded SD of 0.01 0, 0.006, 0.009 and 0.016 m for the female athlete in x, y, z and diagonal directions, respectively. The respective values for the male athlete were 0.013, 0.007, 0.012 and 0.020 m. However, single field SDs varied considerably: diagonal ranged from 0.007 to 0.056 m for female and from 0.008 to 0.085 m for male. Smoothing decreased these variations significantly and also the standard deviation of zero smoothed variables were generally low. For example, 0.00 m for the height of CM at take-off and landing for both genders. Angular velocities had larger deviations, i.e. maximal angular velocity of trail leg hip had SD of 113 "Is (mean 777 "Is) for the female athlete and 140 "Is (mean 1046 "Is) for the male athlete. DISCUSSION The repeated digitising process in this study shows that generally digitising is reliable in this practical application. However, when the deviation of pointing at the landmark reaches five to nine centimetres, the accuracy is not acceptable. This excessive inaccuracy occurred, when the landmark is obstructed from the camera view by other body parts. The practical influence of such digitising error varies for different variables, and it is not recommended to obtain final variable values from manual digitising without smoothing the co-ordinates. One function of smoothing is to cut down random digitising error and thus some of the excessive inaccuracies are reduced. However, in particular situations the effect of smoothing can be detrimental. For example: the centre of mass (CM) parabola is relatively flat and for the female athlete in this study, certain smoothing values changed the peak of this parabola to different fields of filming in different digitising repe!itions. Thus the distance of CM peak to the hurdle varied from 0.36 -+ 0.01 m in the raw data (mean -+ SD) to 0.25 * 0.07 m for smoothed data. It is clear that manual digitising has problems. However, from the careful evaluation of variables, the researchers are able to give reliable feedback to athletes and coaches. REFERENCES Yeadon, M.R., & Challis, J.H. (1994). The future of performance-related sports biomechanics research. Journal of Sports Sciences, 12, 3-32.