WLCSSLearn: learning algorithm for template matching-based gesture recognition systems

Template matching algorithms are well suited for gesture recognition, but unlike other machine learning approaches there are no established methods to optimize their parameters. We present WLCSSLearn: an optimization approach for the WarpingLCSS algorithm based on a genetic algorithms. We demonstrate that WLCSSLearn makes the optimization procedure automatic, fast and suitable for new recognition problems even when there is no a-priori knowledge about suitable range of parameter values. We evaluate WLCSSLearn on three different datasets of gestures. We demonstrated that our method increased the accuracy and F1 score up to 20% compared to previous literature.