Decision-making content of an agent affected by emotional feedback provided by capture of human's emotions through a Bimodal System

• Published in: International journal of computer science issues Vol. 12; no. 6; p. 1
• Main Author: Razuri, Javier G
• Format: Journal Article
• Language: English
• Published: Mahebourg International Journal of Computer Science Issues (IJCSI) 01.11.2015
• Subjects: Accuracy; Adversarial Risk Analysis; Affective Computing; Broaden and Build Theory; Classification; Classifiers; Computer and Systems Sciences; Computer science; data- och systemvetenskap; Detection of Emotional Information; Emotional self-regulation; Emotions; Facial; Facial Expression Recognition; Human; Human behavior; Machine Learning; Speech Emotion Recognition; Voice
• ISSN: 1694-0814; 1694-0784; 1694-0784

Affective computing allows for widening the view of the complex world in human-machine interaction through the comprehension of emotions, which allows an enriched coexistence of natural interactions between them. Corporal features such as facial expression, kinetics, structural components of the voice or vision, to mention just a few, provide us with valid information of how a human behaves. Among all the carriers of emotional information we may point out two, voice and facial gestures as holders of an ample potential for identifying emotions with a high degree of accuracy. This paper focuses on the development of a system that will track a human's affective state using facial expressions and speech signals with the purpose of modifying the actions of an autonomous agent. The system uses a fusion of two baseline unimodal classifiers based on bayes Net giving rise to a multi-classifier. The union of the three classifiers forms a bimodal scheme of emotion classification. The outputs from the baseline unimodal classifiers are combined together through a probability fusion framework applied in the general multi-classifier. The system classifies six universal basic emotions using audiovisual data extracted from the eNTERFACE05 audiovisual emotion database. The emotional information obtained could provide an agent with the basis for taking an affective decision. It is shown by experimental results that the proposed system can detect emotions with good accuracy achieving the change of the emotional behavior of the agent faced with a human.