A Study and Development of Pattern Recognition Methods for Simultaneous Translation from Sign Language

The purpose of this project is to study the existing methods of pattern recognition and to adapt and further develop them to recognize the sign language of the deaf. According to the World Health Organization, there are 360 ​​million hearing impaired and deaf people worldwide. Despite the rapid development of communication technologies, the communication between deaf people and the outside world and the social integration of deaf people into larger society continue to pose problems. Therefore, creating a product that will allow for communication devices to recognize sign language is highly pertinent. This research project examines the subject domain, the concept of machine vision, the concept of pattern recognition, recent and on-going developments in the recognition of the gestures of the deaf as well as the basic requirements in and limitations to solving the task at hand. More specifically, this work studies the methods of pattern recognition with respect to the task of recognizing sign language and describes the main steps in the system of recognizing sign language, including preliminary image processing and the system’s learning. The research conducted allows to draw conclusions about the effectiveness of the use of contour analysis especially when using the Canny operator with preliminary segmentation and background removal. To analyze the effectiveness of the proposed solution, a prototype application was created and tested. As part of the development of the prototype, the currently existing tools (methods and libraries) were researched and applied for the purpose of realizing the task. This work creates the preconditions for the further development of the recognition of sign language applications through the use of the proposed methods. This work includes an introduction, three chapters, a conclusion, and a bibliography. It also contains 1 table, 17 drawings, and 42 sources of literature. Total number of pages is 71.