EXAMINER, a System that Detects Cheating in Examination
BANDUNG, itb.ac.id – Cheating is a big issue during examination. Exam supervisor has to be sharp to recognize various attempts of cheating. Sometimes the cheating is unnoticeable.
Unnoticing cheating is common since human error could happen from exhaustion or boredom during exam supervision. That problem inspires Hendra Putra, Salman Abdillah and Reza Wahyu Kumara for their final project.
All three of them are students from Electrical Engineering class of 2012. They make a digital exam supervising system called EXAMINER. “This project takes 4.5 month to make in the field of my interest, image processing. Careless exam supervision is one of our reasons to develop this project. We name it EXAMINER,” Hendra said.
This body gesture detection-based digital system once displayed on Electrical Engineering Days 2018 on 22nd-24th May 2018. EEDays is an annual event held by Electrical Engineering Study Program of ITB since 2014. Taking place at Aula TImur ITB, EEDays 2018 attracted visitors that wanted to see dozens of Electrical Engineering students’ final project.
How the Motion sensor and Image Capture works
EXAMINER depends on body gesture, in this case joints, which will be recorded by motion sensor. The sensor, called Kinect sensor, will detect body gesture. Generally this sensor is applied on games that involved player’s body movement.
“Kinect Sensor can record body gesture such as finger gesture, hand movement—especially when turning pages, reaching hands and neck movement when turning,” Salman explained. The limitation of Kinect Sensor in detecting objects is aided by a subsystem called You Only Look Once (YOLO). YOLO will identify objects that get recorded by 4 cameras in a 4x4 meter room. “YOLO will detect objects such as hands or hand movements when holding paper,” Reza said.
After Kinect Sensor and YOLO successfully identify the gesture and the object, another subsystem called RabbitMQ sends the screen capture data to exam supervisor. “Screen capture is sent to the supervisor within 1-2 seconds so that it can be considered real-time,” Salman said.
Potency, Obstacles, Evaluation, and Hope
The accuracy of EXAMINER is about 94% that comes from 200 tests. The system can conduct continuous supervision for up to 3 hours. Satisfying as it is, EXAMINER still has some obstacles such as limited reach, object-depended capability, light, sensor installation angle, camera installation angle, and more.
“Indoor light and installation angle of IP Camera affects detection process. Currently, EXAMINER record 6 objects only,” Salman said. Even with its limitations, EXAMINER is highly probable to be developed and applied in modern education system. “Our hope is that this system can be applied using common camera besides IP camera, and that other constraints can be reduced,” Hendra concluded.
Reporter: Billy Akbar Prabowo
scan for download
