An image-processing based technique to obtain instantaneous horizontal walking and running speed.

Walking and running speed is a fundamental parameter studied in a wide range of areas such as sport biomechanics, rehabilitation, health promotion of the elderly, etc. Given that walking or running speed is not constant even within a stride, instantaneous changes in the body motion need to be evaluated to better understand one's performance. In this study, a new cost- and time- efficient methodology to determine instantaneous horizontal walking and running speed was developed. The newly developed method processes the movies taken with a (high-speed) camera. It consists of five sub-steps, which are performed in a serial order: (1) Subtraction of the background image, (2) filtering, (3) binarization and centroid determination, (4) transformation to the laboratory coordinate system and (5) differentiation. To test the accuracy of the newly developed method, the output (position and speed) was compared with the data obtained using motion capture. The average root mean squared (RMS) error (difference between the outputs of the newly developed method and motion capture) of position-time curves was 0.011m-0.033m. The average RMS error of speed-time curves was 0.054m/s-0.076m/s. It was shown that this new method produces accurate outputs of instantaneous walking and running speed.