What Is Affective Computing?

The concept of affective computing was put forward by Professor Picard of the MIT Media Lab in 1997. She pointed out that affective computing is related to emotions and originates from or can affect emotions. Hu Baogang and others from the Institute of Automation of the Chinese Academy of Sciences also proposed the definition of affective computing through their own research: "The purpose of affective computing is to build a harmonious person by giving computers the ability to recognize, understand, express and adapt to human emotions. Computer environment and make your computer more intelligent and comprehensive. "

For a long time, emotion has been out of the sight of researchers in cognitive sciences. It wasn't until the end of the 20th century that the identity of emotion as an important part of the cognitive process was widely recognized by the academic community. Contemporary cognitive scientists have compared emotions with classical cognitive processes such as perception, learning, memory, and speech. Studies on emotion itself and the interaction between emotions and other cognitive processes have become research hotspots in contemporary cognitive science. Affective computing ( affective computing) has also become an emerging research area.

As we all know, people can have emotional ups and downs of emotions such as anger, sorrow, and joy anytime, anywhere. Then, in the process of human-computer interaction, can the computer experience the joy and sorrow of the person and act with the machine? Affective computing research is to try to create a computing system that can perceive, recognize and understand human emotions, and can respond intelligently, sensitively and friendly to human emotions. Emotional abilities.

At present, there are still many challenges in the research of affective computing, for example, the acquisition and modeling of emotional information, the problem of emotional recognition and understanding, the problem of emotional expression, and the realization of natural and harmonious humanization and intelligent human-computer interaction. Obviously, in order to solve the above problems, we need to know how people perceive the environment, what emotions and intentions they have, and how they respond appropriately. Human emotion communication is a very complicated process, which is not only affected by time, place, environment, people, objects and experiences, but also has expression, language, movement or physical contact. Therefore, in the process of human-computer interaction, the computer needs to capture key information, identify the user's emotional state, detect human emotion changes, and use effective clues to select the appropriate user model (based on the user's operation mode and facial expression characteristics) , Attitude preferences, cognitive styles, knowledge backgrounds, etc.), and form expectations for the intentions behind changes in users emotions, and then activate the corresponding database to provide new information that users need in a timely and proactive manner.

The development of affective computing research depends to a large extent on the new advances in the research of human intelligence and emotion in psychological science and cognitive science. Psychological research shows that emotion is the maintenance or change of a relationship between people and the environment. When objective things or situations are consistent with people's needs and desires, it will cause people to positively affirmative emotions, and when they do not meet, it will cause people to be negative Negative emotions. Emotion has three components: subjective experience (individual self-perception of different emotional states), external performance (quantified form of action of various parts of the body when the emotional state occurs, that is, facial expression), and physiological arousal (physiological response produced by emotion). From

Traditional human-computer interaction is mainly carried out through keyboard, mouse, screen and other methods. It only seeks convenience and accuracy, and cannot understand and adapt to human emotions or moods. Without this ability to understand and express emotions, it is difficult to expect computers to be as intelligent as humans, and it is difficult to expect human-computer interaction to be truly harmonious and natural. Because communication and exchanges between humans are natural and affectionate, in the process of human-computer interaction, people naturally expect computers to have emotional capabilities. Affective Computting is to give computers the ability to observe, understand and generate a variety of emotional characteristics similar to humans, and finally to enable computers to interact naturally, lovingly and vividly like humans. ^[1]

At present, the research and development of artificial intelligence has reached a high level, and its research content is gradually expanding and extending. The study of human emotion and cognition is an advanced stage of artificial intelligence, and its research will greatly promote personification

At the recent ISO / IECJTC1 / SC35 conference in Berlin, Germany, the "Information Technology-Affective Computing User Interface-Framework" was jointly proposed by three Chinese scientific research institutions and enterprises, the Software Research Institute of the Chinese Academy of Sciences, the China Electronics Standardization Institute, and Xiaoi Robot. "The proposal was formally approved. This is the first user interface subcommittee standard on sentiment computing, and it is also China's first international standard in the user interface field. ^[3]

People look forward to having and using more humane and intelligent computers. In human-computer interaction, people change from operating a computer to a computer-assisted person; from people moving around a computer, to computers turning around a person; computers change from cognitive to intuitive. Obviously, in order to achieve these transformations, computers in human-computer interaction should have emotional capabilities. Affective computing research is trying to create a computing system that can perceive, recognize and understand human emotions, and can respond intelligently, sensitively and friendly to human emotions.

Emotions are used to represent various inner experiences (such as emotions, moods, and preferences), emotions are used to represent very short but strong inner experiences, and moods or states are used to describe low-intensity but long-lasting inner experiences. Emotion is the maintenance or change of a certain relationship between people and the environment. When objective things or situations are consistent with people's needs and desires, they will cause people to positively affirmative emotions, and when they do not meet, it will cause people to negatively affect emotions.

Emotion has three components: subjective experience, that is, the individual's self-perception of different emotional states; external expressions, that is, expressions, the quantified form of the actions of various parts of the body when the emotional state occurs. Expressions include facial expressions (patterns of changes in facial muscles), gesture expressions (expression movements in other parts of the body), and intonation expressions (changes in speech tone, rhythm, speed, etc.); physiological arousal, that is, the physiology of emotional production Response is a physiological activation level with different response patterns.

In summary, the important role of emotion is mainly manifested in four aspects: emotion is a psychological tool for people to adapt to survival, can stimulate the motivation of psychological activities and behaviors, is the organizer of psychological activities, and is an important means of interpersonal communication. From the perspective of biological evolution, we can divide human emotions into basic emotions and complex emotions. Basic emotions are congenital, with independent neurophysiological mechanisms, internal experience and external performance, and different adaptive functions. There are five basic emotions in people: happiness when the current goal is making progress, anxiety when the self-protected goal is threatened, sadness when the current goal is not achieved, anger when the current goal is frustrated or blocked, and taste ( (Taste) disgust against the goal. Complex emotions are derived from different combinations of basic emotions.

Emotional measurement includes the measurement of three components of emotion dimension, expression and physiological index. For example, to determine a person's anxiety level, we can use questionnaires to measure their subjective feelings, measure their facial expressions by recording and analyzing facial muscle activity, and measure blood pressure with a sphygmomanometer, test blood samples, and measure adrenaline levels in blood .

Determining the emotional dimension is important for sentiment measurement, because only when the emotional dimension is determined can a more accurate assessment of the emotional experience be made. The emotional dimension has two polarities. For example, the agitation of emotion can be divided into agitation and calmness. Agitation refers to a strong and explicit emotional state, while calm refers to a stable and quiet emotional state. Psychological emotion

Summary of Affective Computing Content

The generation and operation of emotions is a very complicated process. Emotional computing must be based on scientific emotional theories. It is realistic. There is no brand-new scientific emotional theory to guide the development of real emotional robots. Is impossible. This new theory of emotion must break through the limitations of psychology and the limitations of social sciences, and become an independent scientific theory that crosses the natural sciences and social sciences. Its fundamental purpose is to digitize emotions. This brand-new sentiment theory is "mathematical sentimentology". It takes "unified value theory" as the theoretical premise, and uses a mathematical logic method to analyze the science of sentiment phenomena and rules.

To sum up, "the unified theory of value" and "mathematical sentiment" mainly complete the theoretical framework of emotional robots through the following steps. ^[4]

Affective Computing for Unified Metrics

Transform the "dissipative structure theory" of physics, and define the concept of "value" (that is, generalized ordered energy) from a physics perspective, so that the value theory is based on natural science; realize the unity of the value of living materials Measure; then, achieve a unified measure of labor value; achieve a unified measure of the value of production materials. In this way, all values have a unified, clear and stable metric or metric: the energy scale (joules). ^[4]

Generalized Value Law of Affective Computing

Derive the "maximum value rate rule" from the "maximum ordering rule" (value rate is the ratio of output value to input value per unit time, and in the economic field, value rate is profit rate); from "maximum value rate" The "rules of choice" or "median value rate rules" are derived from the "rules"; The "generalized value law" is derived from the "median value rate rules". The basic connotation of the general law of value is that the value rate of things is constantly approaching the median value rate of the subject. In other words, the difference between the value rate of things (that is, the difference between the value rate of things and the median value rate of the subject) is constantly going to zero. ^[4]

Affective Computing Interrelation

Knowledge (cognition), emotion (emotion), and will (will) are the three basic forms of human psychological activity, which are a subjective reflection of factual relationship, value relationship, and self-behavior relationship; Value relationship is a kind of Special fact relationship and self-behavior relationship is a special value relationship. Therefore, emotion is a special kind of cognition, and will is a special kind of emotion. Cognition is mainly about how to understand, emotion It is mainly about the "how to", and the will is mainly about "how to do"; Emotion is based on cognition, cognition is based on emotion, will is based on emotion, and emotion is based on will. ^[4]

Mathematical Model of Affective Computing

The philosophical nature of emotion and values is "a subjective reflection of the human brain on the value characteristics of things". Among them, emotion is an indirect and relative reflection of the value characteristics of things, and values are the direct and Absolute reflection; The objective purpose of values is to identify the value rate of things. A mathematical matrix composed of the value rates of all different things can be used to describe a person's value system (ie W = {i × j} m × n ); The objective objective of emotion is to identify "the difference in value rate of things". A mathematical matrix composed of the difference in value rate of all things can be used to describe a person's emotional system (ie M = {i × j} m × n); the emotion matrix and the corresponding coefficient matrix can perform intersection and union operations; each emotional element in the emotional system can be composed of several emotional sub-elements to form an emotional matrix, Form two-dimensional and multi-dimensional emotion matrices. ^[4]

Affective computing builds will

The philosophical nature of the will is "a subjective reflection of the human brain on the value characteristics of its own behavior"; The objective purpose of the will is to identify and deal with the "value rate of its own behavior", and the value rate difference of all different behavior modes can be used A mathematical matrix is formed to describe a person's will system (ie X = {xi × j} m × n); The will matrix and the corresponding action coefficient matrix can perform intersection and union operations; in the will system Each will element of can be composed of a will matrix composed of several will sub-elements, so as to form a two-dimensional and multi-dimensional will matrix, that is, the will of ultra-complex behavior is composed of a will matrix of several complex behaviors. Will is composed of a matrix of wills for simple behavior. Will of simple behavior is composed of a matrix of wills for instinctive behavior. ^[4]

Emotional logic

The logic program of emotional expression is roughly: receiving the stimulus signal of a thing, extracting the value of the thing, comparing the median value rate of the person, determining the intensity and direction of the emotion, selecting the mode of emotional expression, and receiving and processing the emotional expression Feedback signals; The logic program of emotion recognition is roughly: receiving the stimulus signals of things, detecting and extracting feature parameters, comparing emotion modes, generating emotion reflections, and storing emotion information into the emotion matrix; the logic program of emotion calculation is roughly: "Logarithmic conversion" of the difference in value rate of things into emotional intensity, performing intersection and union operations on emotional intensity to generate new emotion intensity, and "exponential reduction" of emotional intensity into the difference in value rate of new things; emotion The revised logic procedure is roughly: a person understands the difference in the actual value rate of something by feeling and thinking, and there is a certain difference between the subjective value rate difference of the thing stored in the emotional matrix in the brain, and the person will modify the The emotional intensity of the thing. ^[4]

Affective Computing Explains Will Run

The logic program of will expression (that is, behavioral implementation) is roughly: receiving the stimulus signal of a certain value target, extracting the subjective value rate (that is, behavior value) of various related behaviors that can most effectively achieve the value target, and comparing its own The median value rate determines the will strength and will direction of various related behaviors, selects the related behaviors with the largest will intensity, and organizes the implementation of the related behaviors. The logic program of will calculation (that is, behavior design) is roughly: establishing the value goal (according to the needs of its own survival and development), designing the overall plan (according to the rule of maximum value rate, decomposing the ultra-complex behavior into several complex behaviors, and passing The union operation and the intersection operation of the will to calculate the will strength of the ultra-complex behavior), formulate the implementation rules (according to the rule of maximum value rate, decompose the complex behavior into several simple behaviors, and perform the union and intersection operation of the will through the union operation of the will , Calculate the strength of will of the complex behavior), implement the specific behavior (decompose the simple behavior into several instinctual behaviors according to the rule of maximum value rate, and calculate the strength of will of the simple behavior through the union operation and the intersection operation of the will ), "Exponential reduction" of the above-mentioned levels of willpower to the corresponding behavioral value rate difference. The logic procedure of the will correction is roughly: through the feeling and thinking, a person understands the difference between the actual value rate of a certain behavior and the difference in the subjective value rate of the behavior memorized by the will matrix in the brain. Fix the strength of the will for the thing. ^[3]

Affective Computing Regulation Mechanism

Establish the corresponding relationship between various emotional modes and value changes, and set and regulate various emotional modes according to actual needs; Derive the "first law of emotional intensity" (that is, the logarithm of the difference between emotional intensity and the value rate of things) Proportionally, = Kmlog (1 + P)), and set and adjust the intensity coefficient of emotion according to actual needs; "Eight Dynamic Characteristics" of emotion (strength, stability, meticulousness, hierarchy, effectiveness, Periodicity, timing, and difference) are precisely defined, and the dynamic characteristics of emotion are set and adjusted according to actual needs; The "eight dynamic characteristics" of will are precisely defined, and the dynamics of will are set and adjusted according to actual needs characteristic. The objective purpose of emotion regulation is to adapt the dynamic characteristics of emotion to the changing characteristics of the value relationship in which the subject is located. The objective purpose of will regulation is to adapt the dynamic characteristics of the will to the changing characteristics of the various behavioral value relationships of the subject. ^{[2] [4]}