中国科学院心理研究所机构知识库

目的利用动作捕捉和机器学习技术,探讨减重条件下通过关节运动的三维空间坐标统计特征测量受试者心理疲劳状态的可行性与可靠性。方法通过长时间认知任务诱发受试者的心理疲劳状态并利用量表进行评估。采用Kinect深度摄像头识别并追踪受试者2min减重跑步运动过程中25个关节点的运动信息。利用高斯过程回归算法建立心理量表与行为数据间的模型,并通过皮尔逊相关和均方根误差对模型进行验证。结果在减重条件下,基于关节运动的统计特征可以预测个体的心理疲劳状态,疲劳量表各个维度预测值与真实值间平均相关系数为0.44,均方根误差为2.94,心境状态量表模型预测值和真实值同样达到中等相关0.45,均方根误差为5.49。结论人体关节运动信息可作为有效生物特征预测受试者心理疲劳水平,且在空间或资源有限情况时,基于动作捕捉和机器学习方法建立的心理指标预测模型可为未来载人航天任务心理状态测量提供新方法。

To test the reliability of measuring the mental fatigue with joint motion characteris tics by motion capture and computer technology in the body weight support situation. Methods Mental fatigue was induced by prolonged cognitive tasks and then was evaluated using a series of psychological questionnaires. Kinect was used to identify and track 25 joint points during the 2-minute running exercise in the body weight support situation for each subject to get data acquisition. The Uaussian process regression algorithm was used to establish a model between the psychological scale and the motion capture data. Pearson correlation and root mean square error (RMSE) were applied for testing models. Results Based on the time-space characteristics of joint motion，the individual’s mental fatigue could be measured under body weight support situation. The mean correlation coefficient between the predicted value and the real scores of the fatigue was 0.44，the RMSE was 2.94，meanwhile，the mean correlation in mood states was 0.45，and the RMSE was 5.49. Conclusion The joint motion information can be used as an effective biometric to predict the mental fatigue. When the space or resources are limited，the psychological index prediction model based on the motion capture and machine learning methods can provide a new perspective in the future space missions.

载人航天领域预先研究项目(17440207)