A Review of Emotions Recognition via Facial Expressions for Human-Robot Interaction

One of the most intriguing challenges in human-robot interaction is whether it can reinforce robots with emotional intelligence to make the interaction more intuitive, realistic, and natural. Emotion recognition is widely researched in broader areas such as human-computer interaction and affective computing. Here we report on recent advances in emotion recognition, particularly in the case of human-computer interaction. Our goal is to analyze the current situation regarding accepted emotion paradigms, interaction styles, and classification strategies, as well as provide insights to focus on facial expressions and how the form of emotions are classified, then provide a list of available datasets containing data for these modalities and the challenge faced by facial expression procedures. Moreover, show a Summarization of Details about the commonly used “Recognition mode, Recognized emotions, Robot Type, Sensor, Algorithm, Methodology, Dataset, Performance metrics, and Results.” This review describes recent research on emotional facial expressions during human-robot interactions. Offers different perspectives for future research.

I. Introduction

Emotions are essential to being human and influence our decisions and actions. Therefore, they play an essential role in communication and emotional intelligence. Precise ability to understand, use, and manage emotions is critical to successful interaction. Affective computing aims to empower machines with emotional intelligence to enhance natural human-machine interaction (HMI) [1]. In human-robot interaction (HRI), robots must have human-like capabilities to monitor, interpret, and express emotion. In complex interaction scenarios such as assistive, educational, and social robots, the ability of a robot to recognize emotion has a significant impact on the resulting social interactions. Several studies have examined the modalities of emotion recognition depending on (facial expressions, body movements, movements, sounds, electrical sensors, etc.) which can transmit emotional information from humans to robots, as well as how robots perceive emotional states [2]. Robots inferring and interpreting human emotions will communicate more effectively with humans. Recent studies have focused on developing algorithms to classify emotional states based on various inputs such as facial expressions, body language, voice, and physiological cues. We report recent advances in emotion recognition (ER), particularly in the context of HRI. The ER is challenging, primarily when performed in a real HRI. The setting can differ significantly from the controlled environment where most detection experiments are usually performed.