基于Kinect的学习者头部姿态动态识别方法

摘要

Online learning gradually becomes a new way of learning, which is facing the challenge of how to evaluate the learning effectiveness.Nowadays, it is a popular solution to use somatosensory information to recognize the gesture and state of learner.Besides, head pose identification is the most important part of human gesture recognition.Due to the different height and sitting position of the learner, the traditional method works with a high false alarm ratio in practice.To this end, a dynamic head pose identification solution is proposed.Experiments demonstrate that the proposed solution can effectively distinguish the learner's head pose, which is more adaptive and scientific.%在线学习渐渐成为一种新型的学习方式,如何有效评价学习效果是在线学习所面临的新挑战.目前一种较为流行的方案是利用Kinect体感数据对学习者学习过程中的姿态识别进而判断学习者的状态,以实现学习过程的监测.其中,头部姿态是人体姿态识别的重要内容.然而由于学习者身高或者坐姿的不同,传统基于预定阈值进行低头、侧头等行为判别的方法在实际应用中存在较大的误差.有鉴于此,论文提出了一种学习者姿态的动态识别方案,方案利用Kinect设备获得学习者的头部位置动态估算出头部偏转判别阈值.实验结果表明,该方案可在学习者身体移动情况下对其头部姿态进行有效识别.