AN ARTIFICIAL INTELLIGENT VIDEO ASSISTANT INVIGILATOR TO CURB EXAMINATION MALPRACTICE

ABSTRACT

The Concept of motion detection is to use the previously observed motion sequence to determine movements in a local region of a video frame. In real applications, this problem can be regarded as a spatiotemporal sequence forecasting problem. The spatiotemporal sequence forecast problem is different from the one-step time series forecasting problem. The prediction target of the problem are in sequence which contains both spatial and temporal structures. Examination Malpractice has been an issues to any examination body. Students are always involved in Examination Malpractice which has led to problems such as discouragement for hard work, low productivity, poor job performances, bribery, corruption and certificate racketeering. The manual method of invigilating an examination is not efficient any longer and require an extra efforts from invigilators to curb exam malpractices. In this study Artificial Intelligent Video Assistant Invigilator was introduced to determine movements in a local region of a video frame during examination. A deep neural network sequence model, which is a bidirectional long short-term memory with conditional random fields (Bi-LSTM-CRF) to extract motion detection prediction from each of the layer stacked was developed. Then, one-dimensional convolutional neural networks (1d-CNNs) was trained to run through all the extracted video frame. Data extracted by the Bi-LSTM with weighted sums was analyzed. The motion detection prediction is made from the weighted summation of the 1d-CNNs.The objective of the research is to develop a smart video assistant invigilator using Artificial Intelligent to curb or if not eradicate examination malpractices in educational sector. The Key Performance Indicators (KPIs) for true positive and true negatives are the observations that are correctly predicted. The information to minimize false positives and false negatives was observed in the evaluation .The results of the Average Precision, Average F-Score and Average Recall was 95.0%, 95.2% and 95.3% respectively.

Keywords: Machine Learning, Spatiotemporal Sequence, Cyberwarfare, Artificial Intelligent