How it works
Emotional activities in the brain and V-A-D model
Different regions in the human brain are involved in emotions. Some regions are associated with the intensity of emotions such as the amygdala and anterior insula. These regions are involved in positive and negative emotions. Other regions, such as the lateral prefrontal cortex, are associated with cognitive emotion regulations. Emotions in the brain have been classified using different approaches and methods. The activity of some regions has been assessed using a locationist approach.
In this study, we used the constructionist approach in classifying emotions in the brain. Using this constructionist approach, we were able able to collect brain wave activities from general domain activity, outputting it as EEG data. We classified emotions using the Valence-Arousal-Dominance constructionist model. Adopting the 3 levels that are mostly used when defining the V-A-D model, valence, arousal, and dominance were categorized at 5 levels for this project.
Valence (positivity) was categorized as highly negative at 1 and highly positive at 5. Arousal (excitement/agitation) was categorized as highly exciting/agitating at 1 and highly soothing/calming at 5. And lastly, dominance (the level of control of emotions) was also categorized from levels 1 to 5.
Participants wore the Muse EEG headset while watching a curated list of 1-2 minute videos that aimed to induce a variety of emotional responses. After watching each video, participants scored themselves on this V-A-D model. Participants' EEG data and scores were then analyzed and developed into personalized graphic art.
Colors have been associated with different emotions. Past research has shown that bright colors are associated with positive emotions while cooler dark colors are associated with negative emotions. Other studies had demonstrated colors influence emotional expressions, e.g associating the color “red” as a threat stimulus. In this project, we used these color categories, along with different patterns of rigidity, brightness, and motion to control our artistic outputs.
References:
Gannouni, A et al (2020). Adaptive Emotion Detection Using the Valence-Arousal-Dominance Model and EEG Brain Rhythmic Activity Changes in Relevant Brain Lobes. IEEE Access, vol. 8, pp. 67444-67455, 2020, doi: 10.1109/ACCESS.2020.2986504.
Sutton, T.M, et al (2016). Color associations to emotion and emotion-laden words: A collection
of norms for stimulus construction and selection. Behav Res 48, 686–728. https://doi.org/10.3758/s13428-015-0598-8
Zhu, Y. (2019). Emotion Regulation of Hippocampus Using Real-Time fMRI Neurofeedback in Healthy Human. Frontiers. https://www.frontiersin.org/articles/10.3389/fnhum.2019.00242/full
Experimental Procedures: Collecting EEG Data
Using a Muse headset to record brainwave activity, raw EEG data were gathered from six participants as they watched emotion-inducing video clips. We first gathered 7 videos that were 1-2 minutes each. Then, we presented each video to participants. After each video, participants rated their emotional states on scales of high to low valence (positivity), arousal (excitement), and dominance (control over experienced emotions). As participants watched each video, their brainwave activities were recorded on a Mind Monitor application. Participants also recorded their primary emotions (happiness, anxiety, etc.), familiarity with the videos, and their demographics. The data collected was then used by our software team for the machine learning approach.