2026-03-14T02:53:33Z https://ipsj.ixsq.nii.ac.jp/oai

oai:ipsj.ixsq.nii.ac.jp:00241809 2025-01-19T07:32:49Z 1164:8666:11564:11871

Development of a Reaction Time Measurement Tool Using AI-Based Skeleton Estimation for Collision Avoidance in the Visually Impaired Development of a Reaction Time Measurement Tool Using AI-Based Skeleton Estimation for Collision Avoidance in the Visually Impaired Akira, Kimura Akira, Kimura セッション4 This study developed an AI-based skeleton estimation program to help visually impaired individuals avoid collisions by providing adequate reaction time. To evaluate the device's effectiveness, baseline reaction times were measured via stimulus-response tasks. Ten visually impaired participants from Gunma Prefecture responded to a recruitment notice and participated in the study. Key measurements included the time from electrical stimulation to collision avoidance reaction and the time to return to an initial posture. A Python-based program was used to process video data, extract coordinates, and record response times. Results showed that, upon detecting an approaching person at a 2-meter distance, the average reaction time was 1.6(0.8) seconds, and the time for posture recovery was 2.8(2.7) seconds. These findings suggest that AI-based skeleton estimation can support future development of safety devices, enhancing independence for visually impaired individuals in real-world settings. This study developed an AI-based skeleton estimation program to help visually impaired individuals avoid collisions by providing adequate reaction time. To evaluate the device's effectiveness, baseline reaction times were measured via stimulus-response tasks. Ten visually impaired participants from Gunma Prefecture responded to a recruitment notice and participated in the study. Key measurements included the time from electrical stimulation to collision avoidance reaction and the time to return to an initial posture. A Python-based program was used to process video data, extract coordinates, and record response times. Results showed that, upon detecting an approaching person at a 2-meter distance, the average reaction time was 1.6(0.8) seconds, and the time for posture recovery was 2.8(2.7) seconds. These findings suggest that AI-based skeleton estimation can support future development of safety devices, enhancing independence for visually impaired individuals in real-world settings. technical report 情報処理学会 2024-12-16 application/pdf 研究報告アクセシビリティ（AAC） 10 2024-AAC-26 1 5 2432-2431 AA12752949 https://ipsj.ixsq.nii.ac.jp/record/241809/files/IPSJ-AAC24026010.pdf eng