Design of a Multimodal Interface based on Psychophysiological Sensing to Identify Emotion

Válber César Cavalcanti Roza, Octavian Adrian Postolache

Abstract

This work proposes a design of a multimodal interface to classify or estimate emotion states. Thus, 7emotions are considered such as:anger, boredom, disgust, anxiety/fear, happiness, sadness and normal.A couple of sensing technologies such as: galvanic skin response (GSR), heart rate (HR), electrocardiography (ECG), oxygen saturation (SpO2) and electroencephalography (EEG)are used to collect psychophysiological signals in relation with emotion state estimation. The International Affective Picture System (IAPS) dataset is used to design the classifier system. Regarding the classification task, a comparison between artificial neural networks (ANN-MLP) and support vector machine (SVM) is presented. The tests were carried out for 20 healthy volunteers ( ) of both genders with age from 23-50 years old. The proposed classifier presents accuracies of 85.71% when using ANN-MLP and 77.14% when using SVM.

Keywords

Multimodal interface, signal analysis, emotion classification, psychophysiological signals

Download

IMEKO-TC4-2017-078.pdf

DOI

-

IMEKO TC

TC4 - Measurement of Electrical Quantities

Back to the proceedings