November 10, 2019

Joel V. Joseph

Topic: Emotion recognition and expression based on Human Motion (ESR 3)

Email: josephjo@post.bgu.ac.il

Supervisors: Dr. Raziel Riemer, Dr. Hila Riemer


Joel is from Delhi, India and has a Bachelors in Applied Psychology and a Masters in Cognitive Science. His research interests lie in the area of Artificial Intelligence, Biomechanics and bi-direction co-learning of humans and machines.

Research project: The effect of dynamic interaction on user: Evaluation of politeness in HRI

Summary of Research

This research is focused on emotional recognition and expression through body movement and posture. Experiments were carried out in a lab setting where participants emotional state was manipulated using movies. During the experiment the participants body motion was recorded using a motion capture system and force plate. Data generated from the experiment was cleaned and processed to analyse posture, body angles and walking parameters. The results reveal significant differences between male and female participants in expression of emotions across various different parameters such as variability of shoulder angle, mean step height etc. Even though we see such differences in emotional expression across gender no underlying differences in their interpretation or chronic tendencies was found. Further research is required to understand the underlying mechanism for such differences.

Ongoing research is focused on using motion capture data to build machine learning models that utilise the information form the experiments to build emotion recognitions systems. Focusing on some methods that could incarnate transfer learning from motion capture to video input data to achieve the task along with incorporating facial features to the developed model.

Video Link: https://youtu.be/rp0Wq34iWP0

Github: https://github.com/JoJoseph25