2026-05-21T19:58:27Z https://ipsj.ixsq.nii.ac.jp/oai

oai:ipsj.ixsq.nii.ac.jp:00213056 2025-01-19T17:17:25Z 6164:6165:6640:10712

A Supporting Technique for Comparative Analysis of Factory Work by Skilled and Unskilled Workers using Neural Network with Attention Mechanism A Supporting Technique for Comparative Analysis of Factory Work by Skilled and Unskilled Workers using Neural Network with Attention Mechanism Qingxin, Xia Atsushi, Wada Takanori, Yoshii Yasuo, Namioka Takuya, Maekawa Qingxin, Xia Atsushi, Wada Takanori, Yoshii Yasuo, Namioka Takuya, Maekawa 行動認識 This study presents a method for identifying significant activity differences between skilled and unskilled factory workers by a neural network with an attention mechanism using wrist-worn accelerometer sensor data collected in real manufacturing. To discover skill knowledge from skilled workers, industrial engineers manually identify activity differences between skilled and unskilled workers, which is likely to obtain skill knowledge, by watching video recordings or sensor data. However, a factory has many workers and manual comparison between pairs of workers is time-consuming. We propose an attention-based neural network to visualize the importance of input segments that contribute to the classification output, which is useful to identify activity differences between workers. This study presents a method for identifying significant activity differences between skilled and unskilled factory workers by a neural network with an attention mechanism using wrist-worn accelerometer sensor data collected in real manufacturing. To discover skill knowledge from skilled workers, industrial engineers manually identify activity differences between skilled and unskilled workers, which is likely to obtain skill knowledge, by watching video recordings or sensor data. However, a factory has many workers and manual comparison between pairs of workers is time-consuming. We propose an attention-based neural network to visualize the importance of input segments that contribute to the classification output, which is useful to identify activity differences between workers. 情報処理学会 2021-06-23 eng conference paper https://ipsj.ixsq.nii.ac.jp/records/213056 マルチメディア，分散協調とモバイルシンポジウム2021論文集 2021 1 1133 1140 https://ipsj.ixsq.nii.ac.jp/record/213056/files/IPSJ-DICOMO2021158.pdf application/pdf 3.0 MB 2023-06-23