ATCam: 공공편의시설 접근성 및 사용성의 실제적 향상을 위한 심층신경망 기반 장애인 고충 탐지 모델 및 중재 제공기법

Abstract

We have developed an assistive technology, called ATCam(Assistive Technology Camera), which is based on deep neural network computer vision technology, to automatically detects and provides context-sensitive appropriate intervention to people with special needs in the context of using a variety of public facilities. The developed ATCam system consists of four inter-related modular components: sensing module, identifying module, decision-making module, and feedback & communication module. The sensing module serves as a system input. The identifying module, which is a deep neural network computer vision model, detects a person with disability based on the assistive device/means he/she is currently using. The decision-making module recognizes the intent and special need of the person and determines context-sensitive appropriate intervention. Based on the decision, the feedback & communication module meets the recognized special need by providing necessary adaptive information/interface or connecting to the facility manager or a staff in charge without delay. In the lab trials, by showing high levels of accuracy and precision, It was demonstrated that ATCam holds good potentials to substantially enhance the accessibility and usability of the existing as well as new public facilities for people with a variety of disabilities.