Usage of matrix feature vectors in recognition of tactile sign language
Abstract
The paper provides an approach to recognition of a tactile sign language. All main stages of recognition are fully implemented: search and separation of hand contour on an image, drawing of a convex hull and "defects" for the contour, designing the feature vectors, and, finally, the gesture recognition. Different variants of the solution are suggested and implemented on the standardization stage, in particular, using the formats of image conversions. An approach of designing and usage of "matrix feature vectors" for a standardized image of hand contour is provided. The mathematical apparatus is presented for operating with matrices in a Euclidean space of matrices with a fixed dimension. Such usage of matrices as feature vectors is important because, in many problems, matrices are natural representatives of objects that are being analyzed. The paper provides different variants of compliance distances for "matrix feature vectors", namely ellipsoidal and orthogonal distances.
References
Golik A.O. Recognition of tactile sign language: designing the feature vectors based on amount and structure of defects. XIX International Conference "Problems of Decision Making under Uncertainties". — April 23–27, 2012. — Mukachevo, Ukraine. — Р. 83.
Donchenko V.S. Euclidian space of vectors and matrixes: constructive methods of description of basic structures and their usage // International Journal "Information technologies & Knowledge". — 2011. — 5, № 3. — P. 203–216.
Donchenko V., Zinko T., Skotarenko F. "Feature vectors" in grouping information problem in applied mathematics: vectors and matrixes / Problems of Computer Intellectualization. — Editors Vitali Velichko, Aleksey Voloshin, Krassimir Markov. — Kyiv, Ukrain — Sofia Bulgaria: NASU, V.M. Glushkov Institute of Cybernetics, ITHEA: 2012 — P.111–124.
