SMATCH: Smart Classroom Attendance Checker Using Beacon, Smartphone and Data Mining Techniques

This research aimed to create a method and application for checking attendance, e.g. in university classes. Currently a variety of methods is available for checking attendance, including fingerprint scanning and radio frequency identification (RFID) card scanning. However, these methods are vulnerable to cheating because when a student arrives and checks in, he or she may slip out of a classroom before the end of the class. There is simply no practical way to regularly keep attendance checking throughout the class hours. Moreover, there is no existing application to periodically record positions where students sit and their sentiments before, during and after a class.  Therefore, we propose a system to address the attendance problem using smartphones and Bluetooth low energy beacons. The system uses facial recognition techniques to authenticate, the beacon’s signal strength to track the sitting positions, and image processing techniques to analyze facial sentiments. In our experiment, the facial recognition method
was based on the Active Appearance Model (68-point facial landmarks) and applied to the publicly available Cohn-Kanade face database of 10 persons (each with 5 correct face samples and 5 incorrect face samples) and converted to 128-dimension vector space. Subsequently, those data were used to classify with various data mining techniques, including Naïve Bayes, Support Vector Machine, k-Nearest Neighbors (k-NN), and Random Forest. As a result, the best classification method for the facial recognition was Random Forest when done with a 10-fold cross validation. Its accuracy was about 98% when performed with unseen data. Thereafter, we developed a web-based application programming interface (API) system for a facial recognition server using Python language and used our Android application deployed on a smartphone to connect with the API for the needed services.