在“互联网+教育”背景下，同步课堂借助信息化手段实现了跨空间同步传输教学资源与内容，被视为是促进教育公平的关键举措之一。但是同步课堂的师生主体增多且空间分离，在实践过程中出现了教学效果不佳的问题。专注度是学生有效学习发生的前提，亦是学生重要的课堂状态之一。本研究以专注度为抓手，通过在同步课堂中有效识别学生的专注度并设计干预方案，致力于提高同步课堂中学生的专注度，进而提升整个同步课堂的开展效果。

首先，本研究结合同步课堂场景的特点，对比现有关于专注度识别的研究，选择基于视频数据的专注度评估方法来识别学生的专注度。相较于人工评价法、基于日志数据和生理数据的评估方法，基于视频数据的专注度评估方法具有可实施性强、无侵入性、价格低廉、可实现自动化、反馈及时等优点。本研究在真实同步课堂场景中采集了学生的视频数据，邀请了3位专家对学生的视频数据进行分析，总结出了同步课堂中学生常见的2种专注动作（端坐听讲、记笔记）和3种非专注动作（左顾右盼、趴桌子、低头），根据这5种动作进行视频片段的拆分、标注，构建了具有四万五千张图片的专注度数据集。之后使用YOWO模型算法对数据集进行训练，五种动作的识别准确率平均值为89.7%。本研究进一步构建了专注度计算公式，邀请了7位教师对19个学生的视频片段进行评级，将其结果与机器识别的结果进行kappa一致性检验，kappa值为0.762，位于0.6~0.8之间，证明了本研究专注度识别方法的可行性。

其次，本研究构建了适合于同步课堂场景的专注度干预策略和模型。邀请了29名有同步课堂教学经验的教师进行访谈，结合已有研究，构建了“问题-原因-策略”三级干预策略库，不同于以往研究直接针对问题设计干预策略，本研究把原因作为干预策略库设计的重要维度之一，以期望“对症下药”，设计更有针对性且能解决学生实际问题的策略。此外，本研究从系统观视角出发，把学习干预视为一个严密的系统工程，设计了同步课堂场景下的专注度干预模型，包括问题诊断、策略匹配、策略实施、效果分析四个环节，四个环节之间环环相扣，整个模型在不断动态变化中迭代完善。该模型充分利用了人类智慧和机器智能的优势，具有人机协同和双师协同两大特点。

最后，在完成专注度识别和干预设计后，为将研究成果产品化、实用化，本研究进一步设计了同步课堂场景下的专注度仪表盘。该仪表盘包括主讲教师端、助学教师端、学生端三个部分，其中，主讲教师端和助学教师端的专注度仪表盘设计了课中和课后两种模式，分别供教师课上和课后使用，学生端的仪表盘仅包含了课后使用模式。仪表盘设计过程以用户真实需求为出发点，经过多轮迭代，依次完成了仪表盘指标设计、低保真原型设计和高保真原型设计。最终，本研究邀请了49位有同步课堂教学经验的一线教师从系统架构设计、功能指标设计、系统设计有效性、系统使用意见四个方面对专注度仪表盘开展了满意度评估，接受调查的教师对于专注度仪表盘的总满意度较高，为83.9%，证明本研究设计的仪表盘系统较为合理，能基本满足用户的真实需求。

综上，本研究从学生专注度视角出发，完成了同步课堂场景下学生的专注度识别、干预策略设计、干预模型构建和仪表盘设计，对同步课堂中提升学习者的专注度具有一定的理论意义和实践价值。后续研究需要通过扩大数据集规模，增添更多的课堂行为类别来进一步提高专注度识别准确率，还需要将本研究的结果应用于实践中，进一步完善专注度干预策略、干预模型以及仪表盘。

Under the background of "Internet+Education", the synchronous classroom utilises information technology to achieve the synchronous transmission of teaching resources and content across different locations, which is regarded as one of the key initiatives to promote educational equity. However, the increased number of teachers and students in synchronous classrooms and their spatial separation have resulted in poor teaching effects during teaching practice. Students' concentration is a prerequisite for effective learning and one of the critical classroom states for students. Focused on students' concentration, this study hopes to improve students' concentration in synchronous classrooms by identifying students' concentration status and designing a concentration intervention program, thereby enhancing the effectiveness of the entire synchronous classroom. 
Firstly, the study used an evaluation method based on video data, taking into account the features of synchronous classroom circumstances and previous research on concentration level recognition. Compared to many other evaluation methods, such as the manual evaluation method, recognition methods based on log data and identification methods based on physiological data, concentration status identification methods based on video data have the advantages of being implementable, non-invasive, inexpensive, automated, and timely in feedback. This study collected videos of students in authentic synchronous classroom circumstances. Three experts were invited to analyse the videos and summarise two common attentive behaviours ("sitting upright" and "taking notes") and three common inattentive behaviours ("looking elsewhere", "head on the desk" and "head down") by the students in synchronous classrooms. The researcher split and labelled the video clips according to these five types of actions and constructed a dataset for students' concentration status with 45,000 images. Subsequently, the researchers used the YOWO model algorithm to train the datasets, achieving an average recognition accuracy of 89.7% for the five actions. This study further constructed a calculation formula for concentration, and invited seven teachers to rate the video clips of 19 students. The kappa consistency test was conducted between the results of the human ratings and the machine recognition, with a kappa value of 0.762, falling within the range of 0.6 to 0.8, thus demonstrating the feasibility of the concentration recognition method in this study.

Secondly, the study designed concentration intervention strategies and a model suitable for synchronous classrooms. The researcher invited 29 teachers with experience in synchronous classroom teaching for interviews. Based on the previous research and the interviews, a three-level intervention strategy library of "problem-cause-strategy" was constructed.In contrast to prior research that solely focused on devising intervention strategies for problems, the study took causes as one of the important dimensions in the design of the intervention strategy library, resulting for more targeted strategies capable of resolving students' practical issues. In addition, from a systematic perspective, this study views learning intervention as a rigorous systematic engineering and designs a concentration intervention model for synchronous classrooms. The model comprises four closely interconnected components: problem diagnosis, strategy matching, strategy implementation, and effect analysis, which is iteratively refined in a constant dynamic change. The model makes full use of the characteristics of human intelligence and machine intelligence, and is characterized by "human-machine cooperation" and "cooperation between the lead teacher and the assistant teacher".

Finally, after completing the concentration recognition and intervention design, this study further designed a dashboard for concentration status in synchronous classrooms to put the research results into products and practice. The dashboard includes three interfaces for the lead teacher, the assistant teacher, and the student, respectively. The dashboards for the lead teacher and the assistant teacher are designed with two modes, one for use during class and the other for post-class, while the student dashboard is designed for post-class use only. This study designed the dashboard based on the actual needs of users. Through multiple rounds of iterations, the designs of dashboard indicators, low-fidelity prototypes, and high-fidelity prototypes were completed progressively. Subsequently, the study invited 49 teachers with experience in synchronous classroom instruction to conduct a satisfaction assessment of the concentration dashboard. The evaluation covered four aspects: system architecture design, functional indicator design, effectiveness of the system design, and feedback on system usage. The teachers indicated a relatively high overall satisfaction for the concentration dashboard at 83.9%, which proved that the dashboard system designed in this study was relatively reasonable and could meet the real needs of users.

In summary, this study, from the perspective of students' concentration, has completed the identification of student concentration in synchronous classrooms, the design of intervention strategies, the construction of intervention strategies, and the design of a dashboard. The study adds theoretical significance and practical value in improving students' concentration in synchronous classrooms. In the future, it is necessary to further improve the accuracy of concentration recognition by expanding the scale of the datasets and adding more categories of classroom behaviors. Additionally, the results of this study need to be applied in practice to further refine the concentration intervention strategies, intervention models, and the dashboard.