The detection of human emotional states has received increasing attention over the last decades for a variety of applications and systems. However, detecting the intensity of the expressed emotion has not been investigated
as much as detecting the type of the expressed emotion. To fill this gap, we investigate the utility of different facial and speech features for the detection of the intensity of the expressed emotion. To this end, we designed different Deep Neural Network architectures and applied them to RAVDESS dataset. Obtained results show speech signal features are better indicators of emotion intensity than facial features. However, in the absence of speech signal, finding emotion intensity by facial expressions, is more accurate for males in comparison to females.
The difference between the accuracy of emotion intensity detection for two genders, motivated us to use speech signal for gender detection. The obtained results confirm the proposed model is better than state-of-the-art in emotion intensity detection and more robust in gender detection.
Original languageEnglish
Title of host publicationBenelux Conference on Artificial Intelligence and Machine Learning
Publication statusAccepted/In press - 5 Oct 2020

ID: 53930288