DESIGN AND IMPLEMENTATION OF THE HEAD, EYE, AND EYEBROW MOVEMENTS DETECTION SYSTEM BASED ON MACHINE LEARNING

• Main Author: Fauziah, Mirna
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/52993
• Institution: Institut Teknologi Bandung

Covid-19 Pandemic in Indonesia, push all of us to get more serious about entering the digital era. All physical activities have been limited and encouraged to do online, including academics. Teaching and learning activities, including exams, even made online. The problem of liveness and cheating in exams has been a challenge in the academic field for a long time. This study discusses the facial feature movement (head, eye, eyebrow) detection system as an effort to develop liveness detection system and an online exam cheating detection system via webcam in the future. The research began by collecting tens of thousands of facial images containing the movements of the facial features needed, such as eye, eyebrows, and head movements. The collected images are then classified and created as a dataset to create a real-time facial features movement detection system. This study uses a Convolutional Neural Network (CNN) architecture to create a model for detection and classification of facial feature movement, along with liveness, with a total of 3.5 million parameters used. The dlib library is used to detect and define facial features such as eyes, eyebrows, and jaw. The system design is divided into three stages, first pre-processing as a stage for preparing the dataset before the training process, the second stage is training the model using the dataset that has been collected, and the last stage is detecting, this stage used to detect the movement of facial features in real-time using a model that has been trained before. The system can work well in bright conditions with an accuracy of 94.48%, and an accuracy of 92.65% with the room light. The use of glasses does not affect the performance of the system, but the use of the mask can affect the system in predicting head movements. Hopefully, this facial feature movement detection system can help the process of developing a cheat detection system for online exams in the future.