Multimodal JSBC UIO Postural - ECG
Description
This dataset contains synchronized multimodal recordings collected during immersive Virtual Reality (VR) emotional elicitation experiments using a Meta Quest 3 headset. The dataset combines electrocardiogram (ECG) signals acquired through a custom ESP32 + AD8232 acquisition module with posture-based features extracted using MediaPipe Pose. Participants interacted with four VR environments designed to induce different emotional states, including Neutral, Fear, Anger, and Positive Affect (Joy). Emotional elicitation was achieved through task-oriented immersive scenarios involving stress, challenge, exploration, and positive interaction mechanics. The dataset includes: - Raw and preprocessed ECG signals - Postural landmark coordinates (33 MediaPipe keypoints) - Time synchronization data - Emotion labels - Windowed multimodal sequences for Deep Learning - Metadata associated with experimental sessions Data were collected from 20 participants under controlled laboratory conditions. The dataset is intended to support research in: - Affective Computing - Emotion Recognition - Virtual Reality - Human–Computer Interaction - Serious Games - Multimodal Deep Learning - Emotion-Aware Systems This dataset may facilitate the development and evaluation of multimodal emotion recognition models under facial occlusion conditions commonly present in immersive VR environments.
Files
Steps to reproduce
All participants provided informed consent before data collection. Personal identifying information was removed from the released dataset. Researchers can directly use the processed multimodal windows for Deep Learning experiments or reconstruct the preprocessing pipeline from the raw synchronized recordings. Model evaluation was conducted using: - Internal train/validation split - Leave-One-Subject-Out (LOSO) cross-validation