Emotion detection while listening to quran recitation using EEG and ECG

Abstract

Emotion modelling and identification has attracted substantial interest from several disciplines including computer science, cognitive science and psychology. Despite the fact that many qualitative studies have been carried out on emotion, quantifying physiological signals remains one of the less-investigated aspects. Therefore, the purpose of this study is to examine various human emotions exhibited by subjects while listening to recitations of Quranic verses based on their perceived meaning or the tone of the verse. This work focuses on understanding and analysing brain and heart activities for two groups: one group understands the language of Al-Quran, while the other group does not. This study attempts to identify the factors (content or intonation) that elicit subjects’ emotions while listening to Quranic recitations. The study uses two methods to measure subjects’ physiological signals: the electroencephalogram (EEG) and the electrocardiogram (ECG). The resulting data are used to analyse subjects’ emotional properties. A solution based on kernel density estimation (KDE) and mel frequency cepstral coefficients (MFCC) is proposed for recognising dynamically developing emotional patterns from EEG and ECG signals. This work uses the multilayer perceptron (MLP) classifier. This classifier’s features are based on the affective space model (ASM), which is represented by two factors: valence and arousal. The experimental setup presented in this work to elicit emotions is based on passive valence/arousal. The EEG and ECG data were collected from 20 Muslim subjects, 10 of whom understood the language of Al-Quran (Arabic), while the remainder did not. In the experiment, visual and auditory stimuli (passive stressors) were used to induce positive and negative emotions. The International Affective Picture System (IAPS) was used to elicit emotions. The results support the use of EEGs as a reliable source for evaluating four basic human emotions. While ECGs can also successfully identify these four basic emotions, the accuracy of emotion extraction from ECG signals is lower than the accuracy from EEG signals. Additionally, the MFCC algorithm with 12 extracted features resulted in higher accuracy than the KDE algorithm when extracting emotions from signals. The groups who understood Arabic reported lower valence than those who did not, indicating that they were affected by the content or meaning of the recitations.