วิธีการรู้จำอักขระภาษาไทยผ่านวีดิทัศน์แสดงการขยับมือแบบเรียลไทม์ ด้วยเทคนิคการเรียนรู้แบบเชิงลึกที่ใช้พลังงานต่ำ

Abstract

Persons or patients who develops speech impairment due to the effects of the disease may pose difficulties for listener who relies mainly on voice communication. Although sign language is now a standard for verbal communicate through the use of gestures or hand movements as symbols. However if the recipient has never learned sign language it may be a problem for the audience to understand. Also, the learning of sign language for patients might be impractical. It may be a new subject that takes time and is difficult to learn. The brain cannot remember or perceive new things as effectively as it should. Nowadays, there are many ways for the development of computers to be able to learn or recognize like humans. One method that is gaining popularity is computer vision method, which receives moving image data from a video camera and learns to distinguish things from video images. For this reason, the researchers are interested in solving communication problems for patients with speech impairment with algorithm that recognizes some Thai characters. The proposed algorithm detects finger movements in real time through a video camera. It them employs deep learning to recognize letters. This is to facilitate simple communication between patients and caregivers. In addition, the researcher aims to develop an algorithm which is suitable for using on low-power computing devices to widen its applicability in limited-resource domains.