Selecting a Small Set of Optimal Gestures from an Extensive Lexicon
This addresses the challenge of optimizing gesture recognition for users, though it appears incremental as it builds on existing lexicon selection methods.
The paper tackled the problem of selecting an optimal small set of gestures from a large lexicon to improve computer recognition performance, presenting an objective function (EDRM) and an efficient algorithm that combines subjective and objective measures to choose the best n gestures where n is much smaller than m.
Finding the best set of gestures to use for a given computer recognition problem is an essential part of optimizing the recognition performance while being mindful to those who may articulate the gestures. An objective function, called the ellipsoidal distance ratio metric (EDRM), for determining the best gestures from a larger lexicon library is presented, along with a numerical method for incorporating subjective preferences. In particular, we demonstrate an efficient algorithm that chooses the best $n$ gestures from a lexicon of $m$ gestures where typically $n \ll m$ using a weighting of both subjective and objective measures.