Aimi Shazwani Ghazali

Affective State Classification Using Bayesian Classifier

This paper elaborates the basic structure of a machine learning system in classifying affective state. There are several techniques in classifying the states depending on the type of input-output dataset. A proper selection of techniques is crucial in determining the success rate of the system prediction. The paper proposes a machine learning technique in classifying affective states of human subjects by using Bayesian Network (BN). A structured experimental setup is designed to induce the affective states of the subjects by using a set of audiovisual stimulants. The affective states under study are happy, sad, and nervous. Preliminary results demonstrate the ability of the BN to predict human affective state with 86% accuracy.

Non-invasive non-contact based affective state identification

This paper discusses a study on detecting affective states of human subjects from their bodys electromagnetic (EM) wave. In particular, the affective states under investigation are happy, nervous, and sad which play important roles in Human-Robot Interaction (HRI) applications. A structured experimental setup was designed to invoke the desired affective states. These states are induced by exposing the subject to a specific set of audiovisual stimulations upon which the EM waves are captured from ten different regions of the subjects body by using a handheld device called Resonant Field Imaging (RFI™). Nine subjects are randomly chosen and the collected data are then preprocessed and trained by Bayesian Network (BN) to map the EM wave to the corresponding affective states. Preliminary results demonstrate the ability of the BN to predict human affective state with 80.6% precision, and 90% accuracy.

Emotion Embodiment in Robot-Assisted Rehabilitation System Using Hybrid Automata

The embodiment of emotions in the paper is structured under hybrid automata framework. In particular, the paper focuses on the description of the automata model designed for robot-assisted rehabilitation system in term of its initialization value, modes, condition for each mode, guard conditions, and transition between modes. A structured experimental setup was designed to evaluate the performance of the hybrid automata proposed. The result demonstrates the efficacy of hybrid automata approach in the rehabilitation application where emotion of the subject is taken into consideration in deploying suitable rehabilitation tasks.

Pardon the rude robot: Social cues diminish reactance to high controlling language

In many future social interactions between robots and humans, robots may need to convince people to change their behavior. People may dislike and resist such persuasive attempts, a phenomenon known as psychological reactance. This paper examines how reactance, measured in terms of negative cognitions and feelings of anger, is affected by the persuading agents social agency cues and the level of controlling language used. Participants played a decision-making game in which a persuasive agent attempted to influence their choices exhibiting high or low controlling language, and three different levels of social agency. Results suggest that controlling language will lead to increased reactance when the persuasive agent does not exhibit social cues. Surprisingly, reactance is not affected by controlling language in the same way when the persuading agent is a social robot exhibiting social cues.

Development of Emotional State Model using Electromagnetic Signal Information for Rehabilitation Robot

AbstractThe paper presents a development of emotion recognition system which can detect human emotion in real-time leveraging information captured from human bodys electromagnetic (EM) signals. A new model of controller framework was designed to embed the emotion recognition module which was evaluated on a robot-assisted rehabilitation platform. The framework is based on hybrid automata model and used to govern the suitable trajectory to deploy by the robotic platform in assisting rehabilitation therapy. The result of the new controller design demonstrates the efficacy of the approach where emotion of the subject is taken into consideration in switching the rehabilitation tasks.

Electromagnetic based emotion recognition using ANOVA feature selection and Bayes Network

The paper discusses the development of emotion recognition system which can be applied to a wider range of human population. This is achieved by measuring the unique electromagnetic (EM) signal generated upon invoking certain emotions. A set of audio-visual stimulants is designed to invoke the desired emotions under study that are happy, sad and nervous. A set of questionnaire is developed to verify the stimulant effectiveness in invoking the emotion. The recognition of the emotion is deduced from the measured electromagnetic signals radiated from the human body by a handheld device called Resonant Field Imaging (RFI™). There are ten points of interest (POIs) on the body where the signals are measured to form the dataset which later fed into Bayes Network (BN) to classify the emotion. ANOVA test is run in selecting the best features to classify the emotions. The result after eliminating 6 from 10 POIs demonstrates the system performance is not compromised. The efficiency of ANOVA and BN in selecting the best features to model the emotion recognition system has successfully optimized the cost of the system and reduced the time to measure the signals quite significantly.

Effects of robot facial characteristics and gender in persuasive human-robot interaction

The growing interest in social robotics makes it relevant to examine the potential of robots as persuasive agents and, more specifically, to examine how robot characteristics influence the way people experience such interactions and comply with the persuasive attempts by robots. The purpose of this research is to identify how the (ostensible) gender and the facial characteristics of a robot influence the extent to which people trust it and the psychological reactance they experience from its persuasive attempts. This paper reports a laboratory study where SociBot™, a robot capable of displaying different faces and dynamic social cues, delivered persuasive messages to participants while playing a game. In-game choice behavior was logged, and trust and reactance toward the advisor were measured using questionnaires. Results show that a robotic advisor with upturned eyebrows and lips (features that people tend to trust more in humans) is more persuasive, evokes more trust, and less psychological reactance compared to one displaying eyebrows pointing down and lips curled downwards at the edges (facial characteristics typically not trusted in humans). Gender of the robot did not affect trust, but participants experienced higher psychological reactance when interacting with a robot of the opposite gender. Remarkably, mediation analysis showed that liking of the robot fully mediates the influence of facial characteristics on trusting beliefs and psychological reactance. Also, psychological reactance was a strong and reliable predictor of trusting beliefs but not of trusting behavior. These results suggest robots that are intended to influence human behavior should be designed to have facial characteristics we trust in humans and could be personalized to have the same gender as the user. Furthermore, personalization and adaptation techniques designed to make people like the robot more may help ensure they will also trust the robot.

The influence of social cues in persuasive social robots on psychological reactance and compliance

Abstract People can react negatively to persuasive attempts experiencing reactance, which gives rise to negative feelings and thoughts and may reduce compliance. This research examines social responses towards persuasive social agents. We present a laboratory experiment which assessed reactance and compliance to persuasive attempts delivered by an artificial (non-robotic) social agent, a social robot with minimal social cues (human-like face with speech output and blinking eyes), and a social robot with enhanced social cues (human-like face with head movement, facial expression, affective intonation of speech output). Our results suggest that a social robot presenting more social cues will cause higher reactance and this effect is stronger when the user feels involved in the task at hand.

The Influence of Social Cues and Controlling Language on Agent's Expertise, Sociability, and Trustworthiness

For optimal human-robot interaction, understanding the determinants and components of anthropomorphism is crucial. This research assessed the influence of an agents social cues and controlling language use on users perceptions of the agents expertise, sociability, and trustworthiness. In a game context, the agent attempted to persuade users to modify their choices using high or low controlling language and using different levels of social cues (advice with text-only with no robot embodiment as the agent, a robot with elementary social cues, and a robot with advanced social cues). As expected, low controlling language lead to higher perceived anthropomorphism, while the robotic agent with the most social cues was selected as the most expert advisor and the non-social agent as the most trusted advisor.

Physiological Signal — Based Engagement Level Analysis under Fuzzy Framework

The paper presents real-time affective state detection, in particular, the engagement level detection by using physiological signals under fuzzy framework. In order to develop the fuzzy model, the engagement model is developed by using data collected from controlled design experiment. In measuring the level of engagement, the physiological signals; namely the Electrooculogram (EOG) is recorded using G-tec data acquisition system. In the experiment, the data collected are the average endogenous eye blinks and the average trajectory errors recorded from the trajectory that the subjects have to follow in completing specific tasks. For the tasks, the subjects are asked to track a set of prescribed paths within the allocated times and have to obey different speed constraints. Various shapes of trajectories are given to the subjects in order to study the level of engagement while performing the task. The data then are used to develop the fuzzy model to measure the level of engagement (LOE) of the subjects. Following the experiments, a series of questionnaires are given to the subjects to verify their engagement level when performing the experiments. Preliminary analysis on the data shows a good match between the experimental results and the questionnaire.