Hand Gesture Selection and Recognition for Visual-Based Human-Machine Interface

Abstract

A new paradigm has been proposed for gesture selection and recognition. The paradigm is based on statistical classification, which has applications in telemedicine, virtual reality, computer games, and sign language studies. The aims of this paper are (1) how to select an appropriate set of gestures having a satisfactory level of discrimination power, and (2) comparison of invariant moments (conventional and Zernike) and geometric properties in recognizing hand gestures. Two-dimensional structures, namely cluster-property and cluster-features matrices, have been employed for gesture selection and to evaluate different gesture characteristics. Moment invariants, Zernike moments, and geometric features are employed for classification and recognition rates are compared. Comparative results confirm better performance of the geometric features.

Surveillance and Activity Recognition with Depth Information

“…4, lower left). Thus the outer curve or edge function of a head will appear as an ellipsis, which can be detected by a classical correlation with predefined head patterns or invariant moments [6]. To enable a robust detection of heads from persons with different body heights, this sampling procedure starts at the top of the door ( 2m) and ends at the height of a children ( 0.8m).…”

Natural hand posture recognition based on Zernike moments and hierarchical classifier

“…Most existing classification methods for hand posture belong to distance-based classifier [4] [5][10] [14]. The distance-based method calculates the distances between the input hand posture and the templates according to a certain criteria.…”

“…As gesture is a natural and intuitive communication channel between human and robot, hand gesture recognition has been studied extensively [3][4] [5]. Analysis of hand posture contributes to further dynamic gesture recognition.…”

“…Different kinds of hand postures might cause various recognition rates since incomplete features set are utilized to represent hand postures. Chalechale et al studies how to select appropriate features and hand postures, which have relatively high recognition rates based on seven hu moments [5]. However, considering user preference for hand postures in natural human-robot interaction, various kinds of postures should be recognized.…”

Secondary Diagonal FLD for Fingerspelling Recognition

“…Fang et al (2004) reported the use of cyber gloves to capture features and classify large vocabulary of signs using Fuzzy Decision trees. Recently, Chalechale and Safaei (2005) presented a Bayesian classifier using geometric and moment based properties of fingerspelling hand postures. They report a very good success rate, but they consider gestures of only single user.…”