Classification of Emotions Based on Electrodermal Activity and Transfer Learning - a Pilot Study

Fredrik A Jacobsen; Ellen W Hafli; Christian Tronstad; Ørjan G Martinsen

doi:10.2478/joeb-2021-0021

Classification of Emotions Based on Electrodermal Activity and Transfer Learning - a Pilot Study

J Electr Bioimpedance. 2021 Dec 30;12(1):178-183. doi: 10.2478/joeb-2021-0021. eCollection 2021 Jan.

Authors

Fredrik A Jacobsen¹, Ellen W Hafli¹, Christian Tronstad², Ørjan G Martinsen^{1

2}

Affiliations

¹ Department of Physics, University of Oslo, Oslo Norway.
² Department of Clinical and Biomedical Engineering, Oslo University Hospital, Oslo Norway.

Abstract

This paper describes the development, execution and results of an experiment assessing emotions with electrodermal response measurements and machine learning. With ten participants, the study was carried out by eliciting emotions through film clips. The data was gathered with the Sudologger 3 and processed with continuous wavelet transformation. A machine learning algorithm was used to classify the data with the use of transfer learning and random forest classification. The results showed that the experiment lays a foundation for further exploration in the field. The addition of augmented data strengthened the classification and proved that more data would benefit the machine learning algorithm. The pilot study brought to light several areas to help with the expansion of the study for larger scale assessment of emotions with electrodermal response measurements and machine learning for the benefit of fields like psychology.

Keywords: Machine learning; assessing emotions; skin conductance; transfer learning.