基于混合场景人脸数据集的学习投入度识别

Learning Engagement Detection Based on Face Dataset in the Mixed Scene

学习投入度是影响教学的重要因素,其自动识别技术可以促进学生的自我调整和教师的教学改进。学习投入度自动识别的准确率主要受训练数据集的影响,在实践中常常因目标场景训练样本不足而难以训练出理想的识别模型。基于此,文章通过三组子实验,探讨了混合场景下的人脸数据集能否提高目标场景下学习投入度识别的准确率问题,结果发现:基于混合场景人脸数据集的学习投入度识别模型平均准确率高于其他两组基于非混合场景数据集的模型。此结论表明,在学习投入度识别任务中通过混合不同场景数据来扩大训练数据集,是解决当前目标场景下标注样本不足的有效策略。本研究对于开发精准的学习投入度识别模型有重要指导作用。

Learning engagement is an important factor affecting teaching,and its automatic detection technology could promote students’self-adjustment and teachers’teaching improvement.The accuracy of learning engagement’s automatic detection is mainly affected by the training dataset.However,in practice,it is often difficult to train an ideal detection model due to the insufficient training samples in the targeted scene.Based on this,the paper explored whether the face dataset in the mixed scene could improve the accuracy of learning engagement’s detection in the targeted scene through three sets of sub-experiments.It was found that the average accuracy of learning engagement detection model based on face dataset in the mixed scene was higher than that of the other two models based on face dataset in the non-mixed scene.This conclusion indicated that expanding training datasets by mixing different scene data was an effective strategy to solve the problem of lacking sufficient data in the targeted scene.This study played an important guidance role in the development of an accurate detection model for learning engagement.