Student Engagement Detection in Classrooms through Facial expression and Sentiment Analysis
To ensure effective learning process, several ways have been proposed in the literature for monitoring and assessment of the class. In this project, an automatic system for monitoring learner?s engagement is proposed by using facial expressions and sentiment analysis. Sentiment analysis is a computa
2025-06-28 16:36:10 - Adil Khan
Student Engagement Detection in Classrooms through Facial expression and Sentiment Analysis
Project Area of Specialization Computer ScienceProject SummaryTo ensure effective learning process, several ways have been proposed in the literature for monitoring and assessment of the class. In this project, an automatic system for monitoring learner’s engagement is proposed by using facial expressions and sentiment analysis. Sentiment analysis is a computational process of classifying text-based data in order to identify positive, negative, or neutral reviews of a learner. In these ways, the tutor can analyze the engagement level of learners and improve the teaching method and strategies to enhance learning process. The facial expression of learners will be recorded after regular intervals of time in the form of images during a three-hour class, for sentiment analysis textual feedback will be collected through Moodle LMS and google survey form. We will be built two datasets for classifier training and testing one in the form of textual data and second in the form of images. A comparative analysis will be applied at the end of the semester to find difference (if any) between written and expressed sentiments of the learners.
Project Objectives- To determine learner’s level of understanding based on real-time feedback using LMS during lecture.
- To classify text-based data into positive, negative, and neutral feedback.
- To develop a reliable and robust technique for human face detection in complex background and varying lighting conditions.
- Data Collection.
In data collection, it will be proposed to use two different techniques first, it will be proposed to use Moodle LMS feedback option during a lecture of three hours. Further these three hours will be divided into three-time window of one hour. Students will submit Moodle LMS feedback in each time window. At the end of the class, an online Google form will be provided to learner to answer the questions that have been specifically designed to take feedback from learners.
Dataset will be labeled by three expert faculty members to verify the polarity of sentences as positive, negative, or neutral by using majority preference technique that suggests if two out of three faculty members assign positive, and one of them assigns negative then this sentence will be considered as positive.
2.Data Preprocessing.
In preprocessing phase, keeping with several researches parameters for pre-processing includes Term Frequency-Inverse Document Frequency (TF-IDF): provides helpful and vital information in this section. It usually evaluates the frequency of helpful words, that eventually build the sentiment detection method simple, Stemmer: it reduces all words to its equitant stream example of this a word ‘talking’ will be stream ‘talk’. Stop Words: A, The, of, end, an, related words are called stop words these are unnecessary and not play any role in classification, WorkstoKeep: for narrow downing the in a specific amount of period , and classification will be done in WEKA that is a popular and open source tool for performing data mining.
3.Classification.
The algorithm will be trained with training data and then applied it on the real data in Supervised machine learning approach. In this it will propose to give dataset collects from Moodle LMS and an online Google form, some part of this dataset will be labeled by faculty members given as training data and other data gave as test data, f-measures, precision, and recall will be measured, tables and graphs will be used to show results.
For classification SVM a well-known machine learning algorithm will be used due to its high acceptance.
4.Comparative Analysis.
One of my fellows is working on automated detection of learner’s engagement by analyzing their facial expressions. We plan to compare our results with each other at the end of the semester to find difference (if any) between written and expressed sentiments of the learners. The comparative analysis could be applied in the quality assurance program at the university to determine and improve learner-tutor relationship.
Benefits of the ProjectEngagment of leaners can be analyzed in several ways. It also increase productivity of the teaching-learning in many ways One of the most used method to get feedback through learners is by questionaire at end of class or semester. There are many issues to get feedback by the questionaire method because learners do not give correct feedback due to many reaons. Instead of the questionnaire completed by the learners during the class, the other way is that an external observers (person apart from an educator teaching) evaluate learners engagement level in the class on daily basis. However, it is not scalable and practical on daily basis.
Engagement level of the learners in a classroom also has an impact on the teaching-learning process. Active learners are more engage with tutor activities whereas passive learners are less engage with tutor in a classroom It is very difficult for tutor to observe the engagment level of every learners in the classroom. Computer Vision techniques are also used for monitor the class in robust and efficient way Sentiment analysis is widely applied on text reviews and feedback about movies, historical or most visited place, airlines reviews, software product reviews, and in education betterment In education development, tutor evaluation process plays a vital role that can help to evaluate the performance matrix and find the gap in the learner learning process. It is seen that during lectures many learners feel difficulty related to the tutor teaching method including, speed of delivery, tone of voice, or due to learner’s level of understanding. Mostly he can not express his reviews about the teaching style due to frustration, anger, or other reasons. Therefore, it is necessary to develop an evaluation system which provides the tutor with current status of their learners learning process.
Technical Details of Final DeliverableA complete realtime intelligent system including android application, desktop application and a dashboard which provides realtime information about whole class engagement level to the present class teacher.
Final Deliverable of the Project HW/SW integrated systemCore Industry EducationOther Industries IT Core Technology Artificial Intelligence(AI)Other Technologies Cloud Infrastructure, OthersSustainable Development Goals Quality EducationRequired Resources| Item Name | Type | No. of Units | Per Unit Cost (in Rs) | Total (in Rs) |
|---|---|---|---|---|
| Total in (Rs) | 70000 | |||
| Nikon D.3200 DSLR camera | Equipment | 1 | 35500 | 35500 |
| Stand for Camera | Equipment | 1 | 5200 | 5200 |
| Storage Device for camera | Equipment | 1 | 7000 | 7000 |
| Cloud platform and domain | Equipment | 1 | 13000 | 13000 |
| Stationary and Printing cost | Miscellaneous | 1 | 3300 | 3300 |
| LED Screen | Equipment | 1 | 6000 | 6000 |