基于机器学习的智能化教室学生行为异常增强识别算法

Enhanced Identification Algorithm Based on Machine Learning for Students'Abnormal Behavior in Intelligent Classrooms

为降低光照和摄像头视角的影响,提高智能化教室内学生异常行为的识别效果,提出了基于机器学习的智能化教室学生行为异常增强识别算法.首先构建出人体姿态表达形式,提取学生行为的人体骨骼节点信息,根据人体骨骼节点信息时序相关性特征,结合线性变换方式,进行典型行为特征增强;然后定义人体姿态质心加速度,设定学生行为异常判定标准,结合CNN-10网络实现学生行为识别.实验结果表明,增强后的典型学生行为可以反映出行为特征差异,且误检占比较低.

To reduce the influence of lighting and camera perspective,and improve the recognition effect of abnormal student behavior in intelligent classrooms,a machine learning based algorithm for enhancing the recognition of abnormal student behavior in intelligent classrooms was proposed.Firstly,a human posture expression form was constructed,and the human skeleton node information of students'behavior was extracted.Based on the temporal correlation characteristics of the human skeleton node information and combined with linear transformation,typical behavior features were enhanced.Then the acceleration of the human body posture center of mass was defined and the criteria for judging abnormal student behavior was set to achieve student behavior recognition combining with CNN-10 network.The experimental results indicated that the enhanced typical student behavior traits could reflect differences in behavioral characteristics,and the false detection rate is relatively low.