The method of basis vectors for recognition sign language by using sensor KINECT

Abstract

For signs recognition different methods, such as the method of reference vectors, hidden Markov models, fuzzy models, the model of artificial neuron, image difference method are used. The tasks of recognition of gestural speech of any language are characterized by many parameters. First of all this is characteristics of the transmission channel of gestural speech, recognition vocabulary scope, the variability of gestures, etc. Besides technical and economic problems Kazakh speech technologies development, among which there is Sign language recognition, are firstly influenced by peculiarities of the Kazakh Language and speech (many rules of word formation, seven cases of nouns, territorial varieties of the Kazakh language and speech) causing difficulties in automatic processing. The development of Kazakh Sign Language database consisting of 42 gestures is the first step in designing the system of automatic recognition of separate hand gestures. For gestures recognition Kinect sensor, hand skeleton coordinates obtained via it and key characteristics processed through XML files via tools and mathematic calculations in MATLAB are used. Gestures recognition objects in this article are letters, digits and some gestures in static form.