Affective Computing and its Potential Application in the Life of Elderly People   

1. Introduction

My presentation consists of three parts: first, basic concepts of emotions and affective computing; second, affective computing research in our USTC; third, affective computing’s potential application in the life of elderly people.

Till now, there are no common accepted definitions of emotions provided by psychologists. There are many emotion models. Previous researchers thought logical thought processes are unaffected by emotions. Now researches recognize that emotion is fundamental to human experience, such as cognition, perception, learning, communication, and rational decision-making. Emotion intelligence is very important to intelligence. Prof. Picard in MIT gave the definition of affective computing: Computing that relates to, arises from, or deliberately influences emotions, whose research content mainly consists of emotion recognition, emotion expression and having emotions.

Our research mainly focuses on emotion recognition from visual information and emotion expression by speech.

3.1 emotion recognition

Facial expression is a convenient way for humans to communicate emotion. As a result, research on expression recognition has become a key focus area of personalized human-computer interaction. Most current research focuses on visible images or videos and good performance has been achieved in this regard. Whereas varying light exposure can hinder visible expression recognition, infrared thermal images, recording the temperature distribution formed by face vein branches, are not sensitive to imaging conditions. Thus, thermal expression recognition is a useful and necessary complement to visible expression recognition. Besides, a change in facial temperature is a clue that can prove helpful for emotion inference. Furthermore, most existing research has been based on posed expression databases, which are elicited by asking subjects to perform a series of emotional expressions in front of a camera. These artificial poses are usually exaggerated. Spontaneous expressions, on the other hand, may be subtle and differ from posed ones both in appearance and timing. It is therefore, most important to study natural expression recognition, ultimately leading to more practical applications thereof. Thus, we propose the research of spontaneous facial expression recognition and emotion inference by using fusion of visible and thermal infrared images, which includes three parts: 1)USTC-NVIE, a natural visible and infrared facial expression database; 2)Spontaneous Facial Expression Recognition by Using Head Motion and facial information; 3)Emotion inference by thermal information; 4)Spontaneous facial expression recognition by using fusion of visible and thermal infrared images; 5)attitude recognition and its application in service robots.

3.2 emotion expression

As a technique that can let computer speak, speech synthesis is drawing more and more attention. Today, much speech synthesis software can synthesize neutral speech naturally and flowingly. However, it is hard to make computers speak with “emotion” as that in our daily life, because of the complexity of emotion model. Interactive Genetic Algorithms which can be acted self-organizingly, adaptively and self-learningly can just resolve the problem of difficulty in modeling emotional speech synthesis. As a result, we have designed an emotional speech synthesis process, which adjusts the parameters (XML tags) used to synthesize emotional speech dynamically, using interactive Genetic Algorithms, to optimize the quality of emotional speech. Also, this paper includes an evaluation experiment, which proves the feasibility of the algorithms.

Not only China, but also western countries are facing the emergence of the old society. It means, we will face a tremendous shortage on staff and qualified healthcare personnel in the near future. Furthermore, people prefer more and more to live in their own homes as long as possible instead of nursery homes. In this case, we need high-tech devices that are assistive social robots, which can be divided into two categories, service type robots and companion type robots. The function of the former is to help the elder live independently, or monitor their activity etc. The latter is mainly used to enhance health and psychological wellbeing of the old. Present research has shown that companion type robots really increase positive mood in elderly. However there is a big gap between the elderly’s requirement and the robots’ abilities. We propose a framework of human-robot assistance for elderly. It consists of both emotion recognition and emotion assistance. It means we should close the loop between robots and users. POMDP (Partial Observable Markova Decision Process) may be a good algorithm for this.

-> Download presentation