Velocity analyses are supportive for coaches in order to improve swimmers' technique and have been widely studied in order to improve the athletes' performance. This work presents a low- cost prototype development to measure swim velocity encompassing noise reduction by using a microcontroller associated with an incremental encoder. Swim velocity profile identifications have been performed by using a Radial Basis Function Neural Network (RBF-NN) improved by the stochastic Generalised Extremal Optimisation (GEO) method to provide a fast convergence. The proposed RBF- NN training is aimed at adjusting Gaussian basic function centres using GEO, which has just one free parameter to be se. lt does not make use of derivatives and can be applied to non-convex or disjoint problems. Finally, the velocity data from a Brazilian elite male swimmer performing the crawl stroke have been obtained in a 25 meters test by using the prototype presented in this work. In this experiment, the pseudo-inverse was employed in the RBF-NN output layer. The proposed RBF-NN provided a multiple correlation coefficient R2 equal to 0.84.
© Copyright 2014 XIIth International Symposium for Biomechanics and Medicine in Swimming. Published by Australian Institute of Sport. All rights reserved.