Topic: Learning emotion recognition through auditory cues and language (ESR 2)
Supervisors: Prof. Dr. Stefan Wermter
Born in South Africa and grew up in Sweden. Currently a PhD Student at the University of Hamburg in Germany. My research interests revolve around machine learning and neural networks.
ESR 2 has the topic of learning emotion recognition through auditory cues and language. Our research on this topic has led us to focus on understanding the relationship between language syntax and affective labels provided by neural networks from language in Human-Robot interactions (HRI). We do this in order to better understand the intended semantics of an affective statement and how this can be applied to decision making for social robot agents. This involves integrating this understanding of affective language into a user study with a robot, wherein the goal is to determine if the assumptions made by the robot based on the affective and syntactic structure of used language aligns with the user’s intention.
Without testing developed affective language understanding systems in an embodied environment with human participants, we are unable to learn if the systems we design are applicable for real-world usage and are robust towards factors such as background noise and grammatical incorrectness. As such, we are required to conduct experiments where users are allowed to express themselves in an intuitive manner and as naturally as possible. In doing so, we are able to design better systems that are developed with real-world applications in mind.
Example application video:
van-Maris, A., Sutherland, A., Mazel, A., Dogramadzi, S., Zook, N., Studley, M., Winfield, A. and Caleb-Solly, P. (2020).
The Impact of Affective Verbal Expressions in Social Robots.
Late-breaking Reports on Real-World Human-Robot Interaction at the 15th Annual ACM/IEEE International Conference on Human-Robot Interaction (HRI-20), pages 1-1, Cambridge, UK.
Sutherland, A., Magg, S. and Wermter, S. (2019).
Leveraging Recursive Processing for Neural-Symbolic Affect-Target Associations.
Proceedings of the International Joint Conference on Neural Networks (IJCNN), pages 1–1, Budapest, Hungary.
Barros, P., Churamani, N., Lakomkin, E., Siqueira, H., Sutherland, A. and Wermter, S. (2018).
The OMG-Emotion Behavior Dataset.
Proceedings of the International Joint Conference on Neural Networks (IJCNN), pages 1–7, Rio de Janeiro, Brazil.
Churamani, N., Sutherland, A. and Barros, P. (2018).
An Affective Robot Companion for Assisting the Elderly in a Cognitive Game Scenario.
Workshop on Intelligent Assistive Computing at the IEEE World Congress on Computational Intelligence (WCCI), pages 1–1, Rio de Janeiro, Brazil.
Griffiths, S., Alpay, T., Sutherland, A., Kerzel, M., Eppe, M., Strahl, E. and Wermter, S. (2018).
Exercise with Social Robots: Companion or Coach?
Workshop on Personal Robots for Exercising and Coaching at the Human-Robot Interaction Conference (HRI), pages 1–1, Chicago, USA.
Siqueira, H., Sutherland, A., Barros, P., Kerzel, M., Magg, S. and Wermter, S. (2018).
Disambiguating Affective Stimulus Associations for Robot Perception and Dialogue.
Proceedings of the IEEE-RAS International Conference on Humanoid Robots (Humanoids), pages 433–440, Beijing, China.
Sutherland, A. (2018).
Toward Emotion Recognition From Early Fused Acoustic and Language Features Using Recursive Neural Networks.
Proceedings of the International Ph.D. Conference on Safe and Social Robotics (SSR), pages 55–56, Madrid, Spain.
Sutherland, A., Magg, S., Weber, C. and Wermter, S. (2018).
Analyzing the Influence of Dataset Composition for Emotion Recognition.
Workshop on Language and Robotics at the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pages 1–1, Madrid, Spain.