Emotions4.pdf (869.49 kB)

EmotionIMX 4 - Novel EEG Features for Consumer Emotion Prediction using Correlation-Based Subset Selection

Download (869.49 kB)
conference contribution
posted on 2022-07-12, 18:09 authored by Mayur Jartarkar, Ashish Sinha, Riddhesh Sawant, Mahak Kothari, Veeky Baths


Affective neuroscience research can help in detecting emotions when a consumer responds to an advertisement. Successful emotional elicitation is a verification of the effectiveness of an advertisement. Affective neuroscience using EEG provides a cost-effective alternative to measure advertisement effectiveness while eliminating several drawbacks of the existing market research tools, which depend on self- reporting. Affective neuroscience research has also provided several techniques to classify and predict emotions. In our study, we collected EEG data from 13 participants while commercial video advertisements were shown to them. We used the “correlation-based subset selection” and “asymmetry of hemispheric channels” to create our feature set, using pearson-coefficient based significance validation. From the correlation analysis, we found that EEG channels in the central region (C3/C4) are strong predictors of emotions when elicited from a video advertisement. We find that hemispheric asymmetry-based features improved the performance of ML-based models as compared to commonly used power spectral density-based features. This study shows that EEG activations in the central region can predict consumer emotional response to commercial video advertisements and are also consistently elicited uniformly across consumers.


Jartarkar, M., Sinha, A. Sawant, R., Kothari, M. & Baths, V. (2022). Novel EEG Features for Consumer Emotion Prediction using Correlation-Based Subset Selection. In ACM IMX Workshop EmotionIMX: Considering Emotions in Multimedia Experience - ACM International Conference on Interactive Media Experiences: IMX 2022 (pp. 135-150). Aveiro, Portugal.