METHOD AND DEVICE FOR MARKING SIGN LANGUAGE GESTURES Russian patent published in 2024 - IPC A61B5/05 G06F17/40 G06F3/01 G06F3/4883 

FIELD: data processing.

SUBSTANCE: group of inventions relates to information technology, in particular to a device and a method for marking sign language gestures. Disclosed is a device for implementing the method, which comprises steps of: displaying a video fragment in a user interface, comprising one gesture component, a gesture or a sign language phrase, wherein in the frame there is one full-face speaker performing at least one sign language gesture during the video process, wherein the frame completely covers the region of space in which the gesture is performed, and parts of the speaker’s body involved in the gesture, providing an interface element providing an indication of a right hand and a left hand, displaying in the user interface a set of visual representations characterizing gesture components included in a predetermined sign language notation system, wherein the set of visual representations comprises at least the following units: hand configuration visual representations unit, hand orientation visual representations unit, wherein the visual representation is an image, an animated image or a short video, which simply characterizes the corresponding gesture component, wherein at least one visual representation unit comprises visual representation sub-units for each hand, wherein the sub-units are formed according to the gesture component parameters, wherein the gesture component parameters include the hand configuration type or the hand orientation type, is received from a user for a gesture contained in a video fragment, for each hand participating in the gesture, an indication of the hand and indication of one or more visual representations from each unit of visual representations performed using the onscreen pointer, for each hand participating in the gesture, forming a gesture identifier corresponding to the notation of this gesture in the notation system, according to received instructions from user corresponding to this hand, and forming an entry in the markup file, containing an indication of the video fragment and for each hand participating in the gesture, a gesture identifier, wherein the data set comprising the markup file and the video fragments specified in the markup file is used for machine learning of the gesture recognition model, video fragments are used as training data, and gesture identifiers from a markup file are used as markup.

EFFECT: group of inventions provides high rate of marking by simplifying the interpretation of gesture components.

11 cl, 10 dwg