Abstract
The dataset contains peripheral physiological signals recorded during an emotion inducing experiment, in which subjects could express their subjectively feelings in real-time. In the experiment, there are intensity-varying trials for three different qualities. For each subject, the dataset includes:
* Metadata: gender and age;
* Physiological signals: raw galvanic skin response, pulse and respiration signals;
* Subjective feeling: real-time subjectively felt emotion intensity;
* Context information: stimuli indexes.
This dataset’s primary goal is to support the development of emotion recognition algorithms that account for emotion dynamics by overcoming limitations of typical emotion recognition datasets in which, over an extended period of time corresponding to a task, subjects provide their felt emotional state with a single label / dimensions for the entire interval. Such static labelling does not easily allow for the depicting of the dynamic nature of emotions, since they do not contain information about feelings throughout the interval. On the other hand, incorporating emotion dynamics in emotion recognition holds the potential for better recognition and interpretation capabilities.
Beyond experts in emotion recognition, this dataset is also valuable to experts in psychological, cognitive neuroscience, and related fields, since it enables exploring relationships between physiological signals and emotional patterns.