Description
Understanding and predicting human emotional states and physiological responses is a complex challenge with significant implications for affective computing, neuroscience, and human-computer interaction. The Emotion Arousal Pattern (EMAP) dataset offers a rich, multimodal resource that includes neuro- and peripheral physiological signals alongside emotional ratings. This dataset enables a more detailed exploration of dynamic emotional and physiological processes.
Dataset Description
The Emotional Arousal Pattern (EMAP) dataset contains neurophysiological, peripheral physiological, and self-reported emotional data from 145 individuals recorded while watching various short video clips. Extracted features include:
- 256 EEG features
- 4 peripheral physiological features: galvanic skin response (GSR), respiration, heart rate (HR), and blood volume
Resulting in a total of 260 features combined with a moment-by-moment arousal rating. Participants can request access to the train and validation set of the extracted features through the following link: EMAP Dataset Request Form.
Note: The test set is not provided and will be used to rank participants on the scoreboard.
Competition Tasks
Primary Task (Regression)
- Implement regression algorithms to predict arousal, heart rate, and galvanic skin response separately using different feature selection approaches.
- Assess the performance of the algorithms using root mean squared error (RMSE) as the evaluation metric.
- Visualize the results by plotting graphs that compare the predicted and true time courses for arousal, heart rate, and galvanic skin response.
Bonus Task (Classification)
- Implement classification models with feature selection to categorize arousal ratings.
- Convert labels into binary classes: Low Arousal (0.00–0.5) and High Arousal (0.51–1.0).
- Evaluate classification performance using the F1-score as the primary metric.
Baseline Code: You can find baseline code for predicting arousal, galvanic skin response, and heart rate on our GitHub page.
Submission Guidelines
- Submit a PPT slide describing your approach and achieved accuracy.
- Include your best model(s), a prediction file (prediction.py) with all preprocessing steps used, and a CSV file of selected features.
- Provide open-source code for reproducibility.
- Submit all materials to: afcrinlab@vuw.ac.nz
Eligibility
The challenge is open to everyone worldwide. A team may consist of up to 4 participants and up to 2 mentors.
Prizes
IEEE CEC 2025 conference certificates will be awarded to the 1st, 2nd, and 3rd place winners of this competition.
Important Dates
Submission Deadline: June 15th, 2025
Competition Organizers
-
Harisu Abdullahi Shehu
School of Engineering and Computer Science, Victoria University of Wellington, New Zealand.
Email: harisu.shehu@ecs.vuw.ac.nz -
Hedwig Eisenbarth
School of Psychology, Victoria University of Wellington, New Zealand.
Email: hedwig.eisenbarth@vuw.ac.nz -
Bing Xue
School of Engineering and Computer Science, Victoria University of Wellington, New Zealand.
Email: bing.xue@ecs.vuw.ac.nz -
Will Browne
School of Electrical Electronics and Robotics, Queensland University of Technology, Australia.
Email: will.browne@qut.edu.au