Detection of Anomalies in Working Posture during Obstacle Avoidance Tasks using One-Class Support Vector Machine

抄録

This study was conducted to evaluate working posture during obstacle avoidance tasks using a one-class support vector machine (SVM), and to compare the efficacy of this method in relation to traditional ergonomic evaluation methods. Eleven right-handed male participants performed reaching tasks in which they were required to move their right arm toward a predefined target position while avoiding obstacles. Working position, obstacle height/width, and obstacle presence varied among the experimental conditions. Working posture and subjective difficulty were assessed for each condition. The one-class SVM was applied to the quaternion of each body segment, which was determined based on measurements of working posture. Rates of postural anomalies were calculated for each experimental condition. The rates of postural anomalies for the right upper limb and head/neck increased as the width of the obstacle increased. Similarly, the subjective difficulty of work increased as obstacle width increased. Observational methods are unable to identify postural anomalies with regard to the right shoulder abduction angle and right lateral bending angle of the neck. However, our findings indicate that the proposed one-class SVM can detect postural anomalies in the right upper limb and head/neck for different obstacle widths.