Accelerometer-based hand gesture recognition using feature weighted naïve bayesian classifiers and dynamic time warping

Accelerometer-based gesture recognition is a major area of interest in human-computer interaction. In this paper, we compare two approaches: naïve Bayesian classification with feature separability weighting [1] and dynamic time warping [2]. Algorithms based on these two approaches are introduced and the results are compared. We evaluate both algorithms with four gesture types and five samples from five different people. The gesture identification accuracy for Bayesian classification and dynamic time warping are 97% and 95%, respectively.