【作者】 陈妍； 刘嘉欣； 魏刚林； 李政霖； 武亚强； 王茜莺； 蔡明祥； 田锋；

【Author】 Yan Chen;Jiaxin Liu;Ganglin Wei;Zhenglin Li;Yaqiang Wu;Qianying Wang;Mingxiang Cai;Feng Tian;School of Computer Science and Technology, Xi’an Jiaotong University;Shaanxi Provincial Key Laboratory of Big Data Knowledge Engineering;Lenovo (Beijing) Co., Ltd.;

【机构】 西安交通大学计算机科学与技术学院； 陕西省大数据知识工程重点实验室； 联想(北京)有限公司；

【摘要】 在线学习者的注视点估计对于掌握学习者的学习状态有很好的辅助作用。实际在线学习场景在进行注视点估计时,受限于人眼构造差异与多变的环境因素,注视点估计的准确度与泛化性不高;并且由于同时在线用户数量庞大,为每位用户训练/微调整个网络模型带来了巨大资源需求和时间开销,为此本文提出了一种融合Landmarks模型和个性化校准策略的在线学习者注视点估计模型。首先,通过引入Landmarks Grid作为补充特征的方法,提出了一种通用的注视点检测模型,该模型在降低模型复杂程度的同时保证了模型的检测效果。其次,为解决在线学习者个性化差异问题,提出了一种基于偏差估计的注视点个性化校准模型,通过用户提供的少量校准样本,有效解决了注视点个性化的问题。在自构建的在线学习者注视点数据集和开源数据集上分别进行了实验,验证了本文提出的模型在注视点估计的准确度及个性化调整方面都有一定的优势。更多还原

【Abstract】 The online learner’s gaze location estimation has a good auxlilary role in grasping the learner’s learning state. When estimating gaze loaction in actual online learning scenarios, the accuracy and generalization of gaze location estimation is not high due to differences in human eye structure and changing environmental factors. Training or fine-tuning the entire network model brings huge resource requirements and time overhead. For these reasons, this paper proposes an online learner gaze estimation model that integrates the Landmarks model and a personalized calibration strategy. First, by introducing Landmarks Grid as a supplementary feature method, a general gaze point detection model is proposed, which reduces the complexity of the model and ensures the detection effect of the model. Secondly, in order to solve the problem of personalization differences of online learners, a gaze loaction personalization calibration model based on bias estimation is proposed, which effectively solves the problem of gaze location personalization through a small number of calibration samples provided by users. Experiments are carried out on the self-built online learner fixation dataset and the open source dataset respectively. And it is verified that the model proposed in this paper has certain advantages in the accuracy of gaze estimation and personalized adjustment.更多还原