Management of monotonous works where human workers are indispensable has paid a lot of efforts to prevent human errors. However, many human errors still occur. Workers repeating monotonous works have unique rhythm in their body movement when they are in good conditions. The paper propose a method to calculates how strong symptoms of human errors workers have as the danger degree, From body movement of workers continuously acquired with accelerometers, the proposed method calculates the danger degree, comparing the body movement just before they commit errors with that in their good conditions. The method improves the precision of error prediction, focusing on period where the danger degree is high. The method contributes to prevention of errors, because it predicts human errors from the danger degree indicating the significance of symptoms. In an experiment to confirm the effectiveness of the method, workers repeat monotonous works tracing a circle shown on a tablet PC display many times. Focusing on the danger degree, the method can detect periods just before workers make errors with the f-measure over 0.7.

Full text: IJEIR-1385_final.pdf