Unencumbered Gestural Interaction
IEEE MultiMedia
Vision-Based Gesture Recognition: A Review
GW '99 Proceedings of the International Gesture Workshop on Gesture-Based Communication in Human-Computer Interaction
The catchment feature model: a device for multimodal fusion and a bridge between signal and sense
EURASIP Journal on Applied Signal Processing
Visual affect recognition
Hi-index | 0.00 |
Hand pose recognition for gestural systems must be able to handle variations in user hand anatomy, perspective effects, and the idiosyncrasies of individual gesture presentation. We present an inductive learning system that is able to derive a rulebase of disjunctive normal form formulae in which each DNF describes a hand pose, and each conjunct within the DNF constitutes a single rule. The system applies a suite of feature detectors during learning and is able to determine a salient subset of features for a given set of hand poses. Only the reduced feature set needs to be computed at recognition time. The rule-based induction system is capable of both exact and flexible matching. The latter produces a list of hand poses ordered by goodness of match when presented with a new hand pose image. We present the results of our experiments in which we trained our system with 931 instances of 20 different hand poses. Our system produced compact rule sets and had a recognition rate of 94%.