WIP: Toward Automated Evaluation of Student Presentation Body Language

This research WIP paper introduces a framework for automating the evaluation of students' class presentations. During oral class presentations, a human judge (mostly a faculty) will evaluate several factors in a short presentation time. This makes the judging task difficult, inconsistent, and highly subjective. Certain evaluation metrics can be assessed using an automated system rather than a human judge. The evaluation system provides objective, consistent, and fair judging of students' presentations. The presenter's body language is among the factors that can be evaluated by such a system. Body language is crucial for communicating messages, emotions, and confidence. Misused body language can convey wrong messages, and be a source of anxiety, distraction, and misunderstanding. The aim is to establish an automated, continuous, computer vision-based system for objective body language/nonverbal presentation skills evaluation. We divide the non-verbal communication skills into four categories: head pose, eye contact, facial expression and emotion, and body pose and gesture. In this paper we focus on head pose as an indicator to evaluate the presenter's eye contact with the audience. We labeled an existing head-pose image dataset into three classes: eye contact (EC), weak eye contact (WEC), and no eye contact(NEC). We designed and trained a neural network model to learn and predict the speaker's eye contact using the extracted features. Finally, we established a scoring system that combines the output of the model to generate a final presentation score. The scoring system is a weighted scheme system, where the user can specify which of the three categories may be weighted higher than the other categories. The proposed system can be used by engineering students to obtain feedback and improve non-verbal presentation and communication skills.