Natural human gestures classification using multisensor data
Michał Cholewa, Przemysław Głomb · 2015
We study the two stage classification approach using Hidden Markov Models and Bayesian Network to natural hand gesture classification. We analyze 22 natural gestures with three sets of sensors (finger bend, accelerometers and pitch/roll), classifying each of them separately, and then combining the results using Bayesian classifier. This method achieves significant improvement over single stage classifier trained on the whole multisensor sequences.