CN101692261A

CN101692261A - Individualized emotion model applied to child user playmate robot and application method thereof

Info

Publication number: CN101692261A
Application number: CN200910169792A
Authority: CN
Inventors: 王巍; 王志良; 解仑; 郑思仪; 谷学静
Original assignee: University of Science and Technology Beijing USTB
Current assignee: University of Science and Technology Beijing USTB
Priority date: 2009-09-02
Filing date: 2009-09-02
Publication date: 2010-04-07

Abstract

The invention discloses an individualized emotion model applied to a child user playmate robot, which mainly comprises a random process model of an emotional state stimulation transfer process, a markov chain model of an emotional state spontaneous transfer process, a cybernetic model of a mood state stimulation transfer, and a dynamic balance model of a mood state spontaneous transfer. On the basis of the prior emotion model, the invention discloses the integration of a mood model fused with an individualized interactive technique and the four models, and also discloses an application method of the models for software realization.

Description

A personalized emotion model and its application method applied to children's playmate robot

技术领域 technical field

本发明属于一种人机交互技术领域，具体来说本发明涉及针对儿童用户的玩伴机器人个性化交互技术，特别是儿童用户玩伴机器人的个性化情感模型的建立及其应用方法。The invention belongs to the field of human-computer interaction technology. Specifically, the invention relates to a personalized interaction technology for a playmate robot for children, especially the establishment and application method of a personalized emotional model for a playmate robot for children.

背景技术 Background technique

为了真正实现自然和谐的人机交互，迫切要求机器人在进行交互服务过程中具有更强的人性化：能够根据用户的特定需求进行有针对性的交互服务，即：符合“以人为中心”的、具备个性化色彩的多模态信息协同实时交互服务，实现人类用户与机器人的和谐共存，进行如同人与人之间那样自然和谐的交流。而目前，这方面相关的工作开展得还很少。机器人个性化交互服务技术是一个亟待解决的问题。In order to truly realize natural and harmonious human-computer interaction, it is urgently required that robots be more humanized in the process of interactive services: they can provide targeted interactive services according to the specific needs of users, that is: in line with the "human-centered" and The multi-modal information collaborative real-time interactive service with personalized color can realize the harmonious coexistence of human users and robots, and carry out natural and harmonious communication like that between human beings. At present, there is little work in this area. Robot personalized interactive service technology is an urgent problem to be solved.

美国麻省理工学院所研制的Kismet机器人，通过所建立的环境、内部刺激和行为动作的认知心理模型，对外界输入的刺激和内部需要进行综合判断，从而引起表现行为的各种变化。具有与儿童用户相似的行为方式和能力，比如模仿父母与孩子之间表示情绪状态的反馈方式，小孩向父母表达需求和愿望的方式，以及儿童自我学习与人和环境交流的方式。日本早稻田大学开发的儿童玩伴机器人有视觉、听觉、触觉和嗅觉传感器来感知外界刺激信号。根据所建立的儿童大脑结构化模型，通过外部和内部的刺激相应地改变其情绪状态，而后由多种方式表达出来，并通过非线性计算所产生的内部钟来表述心境矢量的激活成分。The Kismet robot developed by the Massachusetts Institute of Technology, through the established cognitive mental model of the environment, internal stimuli, and behavioral actions, makes comprehensive judgments on external input stimuli and internal needs, thereby causing various changes in performance behavior. Have behaviors and abilities similar to those of child users, such as imitating the way of feedback between parents and children to express emotional states, the way children express their needs and wishes to parents, and the way children learn to communicate with people and the environment. The children's playmate robot developed by Waseda University in Japan has visual, auditory, tactile and olfactory sensors to perceive external stimulus signals. According to the established structural model of the children's brain, external and internal stimuli change their emotional state accordingly, and then express it in a variety of ways, and express the activation components of the mood vector through the internal clock generated by nonlinear calculation.

为了在人机交互过程中，体现出个性化的因素，解决途径之一是在机器人上建立情感模型，通过外部刺激的变化，引起模型输出的变化，进而驱动机器人的动作表达的变化。完成个性化交互。In order to reflect personalized factors in the process of human-computer interaction, one of the solutions is to establish an emotional model on the robot, and through changes in external stimuli, changes in the output of the model are caused to drive changes in the robot's action expression. Complete personalized interactions.

根据Parkinson(1996)的研究指出，情感可以划分为心境和情绪这两个广泛的分类。他们是有一定差别的，1)心境持续时间相对较长，情绪相对较短；2)心境是逐渐发生、连续性的。抑扬的，情绪则快速发生、情节性；3)心境强度相对较弱，情绪则较强；4)心境是常常伴随着较小事件的发生，以及持久稳固的环境条件逐步形成时引起的连锁反应而发生，或者是认知过程和新陈代谢的结果，情绪则一般是由特定的事件引起；5)心境提供自我的当前状态信息，情绪提供情境的当前状态信息；6)心境不是指向特定对象的，情绪则指向特定对象。According to Parkinson's (1996) research, emotion can be divided into two broad categories of mood and emotion. There are certain differences between them, 1) the duration of mood is relatively long, and the mood is relatively short; 2) mood occurs gradually and continuously. Circadian, the emotion is rapid and episodic; 3) the intensity of the mood is relatively weak, and the emotion is strong; 4) the mood is often accompanied by the occurrence of small events, and the chain reaction caused by the gradual formation of persistent and stable environmental conditions However, when it occurs, or is the result of cognitive process and metabolism, emotion is generally caused by a specific event; 5) Mood provides information about the current state of the self, and emotion provides information about the current state of the situation; 6) Mood does not point to a specific object, Emotions are directed at specific objects.

情感计算的概念由美国麻省理工大学媒体实验室的R.Picard教授于1995年提出，并于1997年正式出版专著“Affective Computing(情感计算)”。在该书中她定义“情感计算是与情感相关、来源于情感或能够对情感施加影响的计算”。一经提出，就引起了很大反响，她领导的情感计算课题组以对人类情绪的生理信号处理为基本出发点，研究取得了很多进展，其应用领域日益扩大，Picard教授在美国麻省理工大学媒体实验室的技术报告中已经涉及到约50种应用。1996年日本文部省就以国家重点基金的方式开始支持“情感信息的信息学、心理学研究”的重大研究课题，参加该项目的有十几个大学和研究单位，主要目的是把情感信息的研究从心理学角度向心理学、信息科学两个交叉学科领域过渡。The concept of affective computing was proposed by Professor R. Picard of the MIT Media Laboratory in 1995, and the monograph "Affective Computing (Affective Computing)" was officially published in 1997. In the book, she defines "affective computing is computing related to emotion, derived from emotion or able to exert influence on emotion". Once it was proposed, it aroused great repercussions. The affective computing research group led by her took the physiological signal processing of human emotions as the basic starting point. The research has made a lot of progress, and its application fields are expanding day by day. About 50 applications have been covered in the laboratory's technical reports. In 1996, the Japanese Ministry of Education began to support the major research project of "Informatics and Psychology Research on Emotional Information" in the form of national key funds. There are more than a dozen universities and research units participating in the project. The main purpose is to integrate the research on emotional information. Transition from the perspective of psychology to two interdisciplinary fields of psychology and information science.

国内北京科技大学提出了人工心理的概念，认为人工心理就是利用信息科学的手段，对人的心理活动(着重是人的情感、意志、性格、创造)的更全面再一次人工机器(计算机、模型算法等)模拟，其目的在于从心理学广义层次上研究情感计算、情绪与认知、动机与情绪的人工机器实现问题。The domestic University of Science and Technology Beijing put forward the concept of artificial psychology, and believes that artificial psychology is to use the means of information science to comprehensively analyze human psychological activities (emphasis is on human emotion, will, character, and creation). Artificial machines (computers, models) Algorithm, etc.) simulation, the purpose of which is to study the artificial machine realization of affective computing, emotion and cognition, motivation and emotion from the broad level of psychology.

由于情感可划分为心境(Mood)和情绪(Emotion)两个广泛的分类，并且是可以用计算机模拟的，因此其转移过程在不同条件下可分为4种，如图1所示：Since emotion can be divided into two broad categories of mood (Mood) and emotion (Emotion), and can be simulated by computer, its transfer process can be divided into four types under different conditions, as shown in Figure 1:

(1)情绪状态刺激转移过程：在外部事件的刺激下，情感状态由心境的动态平衡状态移到某种激发水平的情绪激发状态，由A线表示；(1) Emotional state stimulation transfer process: under the stimulation of external events, the emotional state moves from the state of dynamic balance of mood to the state of emotional stimulation at a certain level of stimulation, represented by line A;

(2)情绪状态自发转移过程：当外界刺激作用结束后，某种情绪状态将在一定的时间内由激发状态自发地转移到心境的动态平衡状态，由B线表示；(2) Emotional state spontaneous transfer process: when the external stimulus is over, a certain emotional state will spontaneously transfer from the excited state to the dynamic equilibrium state of mood within a certain period of time, represented by the B line;

(3)心境状态刺激转移过程：在某种特异性的外部事件的刺激下，心境状态将在以心境的动态平衡状态为中心的一定的范围内发生转移，由C线表示；(3) Mood state stimulus transfer process: Under the stimulation of a specific external event, the state of mind will shift within a certain range centered on the dynamic equilibrium state of the state of mind, represented by the C line;

(4)心境状态自发转移过程：在外界刺激消失后，某种心境激发状态将在一定的时间内自发地向心境动态平衡状态转移，由D线表示。(4) Spontaneous transfer process of mood state: After the external stimulus disappears, a certain state of emotional stimulation will spontaneously transfer to the dynamic equilibrium state of mood within a certain period of time, represented by the D line.

综合上述的4个过程，针对儿童用户的玩伴机器人个性化情感模型包含4个，分别是：Combining the above four processes, the personalized emotional model of the playmate robot for children includes four, namely:

1)情绪状态刺激转移过程的随机过程模型；1) Stochastic process model of emotional state stimulus transfer process;

2)情绪状态自发转移过程的马尔可夫链模型；2) Markov chain model of the spontaneous transfer process of emotional state;

3)心境状态刺激转移的控制论模型；3) A cybernetic model for the transfer of mood state stimuli;

4)心境状态自发转移的动态平衡模型。4) A dynamic equilibrium model of spontaneous transfer of mood states.

在滕少冬的博士论文《应用于个人机器人的人工心理模型的研究》)(北京科技大学博士论文，2006)中的第三章及第四章，对于情绪状态刺激转移过程的随机过程模型及情绪状态自发转移过程的马尔可夫链模型已经做出了精准的描述。In the third and fourth chapters of Teng Shaodong's doctoral thesis "Research on Artificial Mental Models Applied to Personal Robots" (Doctoral Thesis of Beijing University of Science and Technology, 2006), the stochastic process model of emotional state stimulus transfer process and emotional state The Markov chain model of the spontaneous transfer process has made an accurate description.

情绪状态刺激转移过程的随机过程模型：此过程的随机过程模型，是建立在情感状态概率空间的基础上，用λ＝(N，M，π，A，B)描述。其中，N是情绪种类；M是刺激种类；π是初始状态概率分布向量为；状态转移概率矩阵为观察值概率矩阵为 $B = [\begin{matrix} b_{11} & b_{12} & b_{13} \\ b_{21} & b_{22} & b_{23} \\ b_{31} & b_{32} & b_{33} \end{matrix}] .$ The stochastic process model of emotional state stimulus transfer process: the stochastic process model of this process is based on the emotional state probability space, described by λ=(N, M, π, A, B). Among them, N is the emotion type; M is the stimulus type; π is the initial state probability distribution vector; the state transition probability matrix is The observation probability matrix is $B = [\begin{matrix} b_{11} & b_{12} & b_{13} \\ b_{twenty one} & b_{twenty two} & b_{twenty three} \\ b_{31} & b_{32} & b_{33} \end{matrix}] .$

情绪状态自发转移过程的马尔可夫链模型：自发转移过程的马尔可夫链模型是：P^(t+1)＝P^(t)A。The Markov chain model of the emotional state spontaneous transfer process: The Markov chain model of the spontaneous transfer process is: P ^(t+1) = P ^(t) A.

其中，是状态转移矩阵；P^(t)是前一时刻的情感概率向量；P^(t+1)是后一时刻的情感概率向量。in, is the state transition matrix; P ^(t) is the emotion probability vector at the previous moment; P ^(t+1) is the emotion probability vector at the next moment.

但是，目前开展的研究工作主要限于情绪状态刺激转移过程的随机过程模型及情绪状态自发转移过程的马尔可夫链模型这两个模型的建立和讨论上，而对于心境状态刺激转移的控制论模型、心境状态自发转移的动态平衡模型以及上述四个模型的整合乃至软件化应用等，尚未有人做出完整而系统的阐述，特别是将人类用户的个性化因素及综合性、通用性的评估标准考虑到实际系统中。因此，融合个性化交互技术将是一个有前景的研究方向。However, the current research work is mainly limited to the establishment and discussion of the stochastic process model of the emotional state stimulus transfer process and the Markov chain model of the emotional state spontaneous transfer process, while the cybernetic model of the emotional state stimulus transfer process , the dynamic balance model of the spontaneous transfer of mood states, the integration of the above four models, and even software applications, etc., have not yet been fully and systematically elaborated, especially the individual factors of human users and comprehensive and universal evaluation criteria into the actual system. Therefore, integrating personalized interaction technology will be a promising research direction.

发明内容 Contents of the invention

本发明一种运用于儿童用户玩伴机器人的个性化情感模型，除了利用已有的情绪状态自发转移过程的马尔科夫链模型和状态刺激转移过程的HMM情绪模型情绪模型外，特别是对于心境模型的建立，综合了心境的昼夜波动、心境的周变化、心境的月变化和心境的季节变化等个体的生理性周期变化因素和人格特征水平等个体内源性因素建立了心境状态自发转移的动态平衡模型；基于控制论的思想，利用惯性环节对心境状态刺激转移过程建模，在某种特异性的外部事件的刺激下，心境状态将在以心境的动态平衡状态为中心的一定的范围内发生转移，建立了心境状态刺激转移的控制论模型；并对上述四个模型进行了整合，建立了t时刻的情感强度值模型。并针对儿童用户，完善儿童玩伴机器人的完整情感变化。The present invention is a personalized emotional model applied to children's playmate robots. In addition to utilizing the existing Markov chain model of the emotional state spontaneous transfer process and the HMM emotional model emotional model of the state stimulus transfer process, it is especially for the state of mind The establishment of the model integrated the diurnal fluctuations of mood, the weekly changes of mood, the monthly changes of mood and the seasonal changes of mood, and other individual endogenous factors such as the individual's physiological cycle change and the level of personality traits to establish the mechanism of spontaneous transfer of mood states. Dynamic balance model: Based on the idea of cybernetics, the inertial link is used to model the transfer process of mood state stimuli. Under the stimulation of a specific external event, the state of mind will be in a certain range centered on the dynamic balance state of the mood A cybernetic model of emotional state stimulus transfer is established; and the above four models are integrated to establish an emotional intensity value model at time t. And for children users, improve the complete emotional changes of children's playmate robots.

一、本发明的一些概念。1. Some concepts of the present invention.

1、情感能量与情感强度向量1. Emotional energy and emotional intensity vector

4个模型相辅相成，在4个过程中共同作用。它们的理论基础却是一致的。在心理学中心理能量观点的基础上，滕少冬提出了情感能量的概念。它是本发明中个性化情感模型的建立的出发点和基础，同时也是针对儿童用户的玩伴机器人个性化交互技术的基础。The four models complement each other and work together in the four processes. Their theoretical basis is the same. On the basis of the concept of psychological energy in psychology, Teng Shaodong proposed the concept of emotional energy. It is the starting point and foundation of the establishment of the personalized emotion model in the present invention, and is also the basis of the personalized interaction technology of the playmate robot for children users.

所谓心理能量就是推动个体进行各种心理活动以及行为的能力，我们用E表示，它有两种基本表现形式：1)自由的心理能量E_η；2)受约束的心理能量E_λ。The so-called psychological energy is the ability to drive individuals to carry out various psychological activities and behaviors. We use E to represent it. It has two basic forms: 1) free psychological energy E _η ; 2) restricted psychological energy E _λ .

定义：称λ＝E_λ/E为心理能量约束度，η＝(1-λ)为心理能量的自由度。这两个参数表达了情感状态的状态的松弛-紧张的维度。上述两种形式的心理能量表达式为：Definition: λ=E _λ /E is called the degree of constraint of psychological energy, and η=(1-λ) is the degree of freedom of psychological energy. These two parameters express the state's relaxation-tension dimension of the affective state. The expressions of the above two forms of mental energy are:

$\{\begin{matrix} {E E.}_{λ λ} = = λE λE \\ {E E.}_{η η} = = ηE ηE \\ η η + + λ λ = = 11 \end{matrix} - - - - - - ((11))$

心理能量对外部的表现形式主要有两个方面：一方面体现为驱动个体进行感兴趣或与当前需要相联系的各种活动；另一方面伴随着各种活动的进行以及根据活动目标是否达到、需要是否得到满足，而以各种情感状态和情绪表达的方式体现出来。There are two main forms of external manifestations of psychological energy: on the one hand, it is reflected in various activities that drive individuals to be interested or related to current needs; Whether needs are met or not is manifested in various affective states and emotional expressions.

从这个意义上来说，我们把以后一种形式表现出来的心理能量称为情感能量。由于E_η是自由运动的，它可全部用于表达各种情绪，而E_λ是被约束的，它只能在一定程度上表现为各种情绪，所以情感能量可以用下式表示：In this sense, we call the psychological energy expressed in the latter form emotional energy. Since E _η is free to move, it can be used to express all kinds of emotions, while E _λ is restricted, and it can only express various emotions to a certain extent, so the emotional energy can be expressed by the following formula:

E_p＝E_η+γR_λ＝(1-λ)E+γλE＝(1-λ+γλ)E (2)E _p =E _η +γR _λ =(1-λ)E+γλE=(1-λ+γλ)E (2)

同时，设

为t时刻实际表现出的情感强度绝对分布向量。此向量的求解在后面介绍。At the same time, set

is the absolute distribution vector of emotional intensity actually shown at time t. The solution to this vector is described later.

2、生理性激活与抑制的情感能量2. Physiological activation and inhibition of emotional energy

根据巴浦洛夫高级神经学说，由于生理的原因，人的大脑神经细胞在兴奋与抑制两种状态之间按一定的生理机制呈周期性的变化，伴随着这种变化，个体的意识状态也将在清醒与不清醒之间进行转化，从而使得情感能量E_p在表达的时候，其表现程度也呈现出周期性的变化。我们把由式According to Pavlov's advanced neurology theory, due to physiological reasons, the nerve cells in the human brain change periodically according to a certain physiological mechanism between the two states of excitement and inhibition. With this change, the state of consciousness of the individual also changes It will be converted between awake and unconscious, so that when the emotional energy E _p is expressed, its expression level also shows periodic changes. We put the formula

${E E.}_{p p}^{α α} = = α α {E E.}_{p p} = = α α ((11 - - λ λ + + γλ γλ)) E E. - - - - - - ((33))$

定义的情感能量称为生理性激活的情感能量，它是实际用于表现情绪的情感能量，称α(0≤α≤1)为生理性唤醒度。把由式The defined emotional energy is called physiologically activated emotional energy, which is actually used to express emotions, and α (0≤α≤1) is called physiological arousal. Putyu

${E E.}_{p p}^{β β} = = β β {E E.}_{p p} = = β β ((11 - - λ λ + + γλ γλ)) E E. - - - - - - ((44))$

定义的情感能量称为生理性抑制的情感能量，它是用于表现心境的情感能量，称β(0≤β≤1)为生理性抑制度。且有The defined emotional energy is called the emotional energy of physiological inhibition, which is the emotional energy used to express the state of mind, and β (0≤β≤1) is called the degree of physiological inhibition. and have

α+β＝1 (5)α+β=1 (5)

α和β主要由生理机制进行周期性的调节，即“生物钟”的调节。另外，α和β还会受到某些外界刺激的干扰。心理能量以及情感能量的各种形式以及转化关系如图2，图3所示。α and β are mainly cyclically regulated by physiological mechanisms, that is, the regulation of "biological clock". In addition, α and β will also be disturbed by some external stimuli. The various forms and transformation relations of psychological energy and emotional energy are shown in Figure 2 and Figure 3.

3、情绪强度3. Emotional intensity

从动力心理学的观点来看，个体产生各种不同情绪的过程，实际上就是激活的情感能量E_p ^α在不同情绪状态之间的动态分配的过程，图2反映了这样的关系。From the perspective of dynamic psychology, the process of individuals producing various emotions is actually the process of dynamic allocation of activated emotional energy E _p ^α among different emotional states. Figure 2 reflects this relationship.

称

为t时刻实际表现的情绪强度绝对分布向量，

i∈{1，2，...，N}是激活的情感能量在各维度的能量值分量。根据情感能量守恒定律，我们有下式成立：say

is the absolute distribution vector of emotional intensity actually expressed at time t,

i∈{1,2,...,N} is the energy value component of the activated emotional energy in each dimension. According to the law of conservation of emotional energy, we have the following formula established:

${Σ Σ}_{i i = = 11}^{N N} | | {E E.}_{pi p}^{αt αt} | | = = {E E.}_{p p}^{α α} - - - - - - ((66))$

4、心境强度4. Mood intensity

生理性抑制的情感能量E_p ^β在积极心境与消极心境之间变化，由其引起的心境强度变化如图3所示。The physiologically inhibited emotional energy E _p ^β varies between positive mood and negative mood, and the changes in mood intensity caused by it are shown in Figure 3.

相应的，称

为t时刻实际表现的心境强度绝对分布向量，并设与积极情绪对应的心境个数是m，与消极情绪对应的心境个数是n个，则m+n＝N。其中：Correspondingly, said

is the absolute distribution vector of mood intensity actually expressed at time t, and suppose the number of moods corresponding to positive emotions is m, and the number of moods corresponding to negative emotions is n, then m+n=N. in:

M_p ^βt为t时刻的心境强度，其计算方法在后面介绍。

M _p ^βt is the mood intensity at time t, and its calculation method will be introduced later.

i∈{1，2，...，N}是生理性抑制的情感能量在各维度的能量值分量。则根据情感能量守恒定律，我们有下式成立：

i∈{1, 2,..., N} is the energy value component of physiologically inhibited emotional energy in each dimension. Then according to the law of conservation of emotional energy, we have the following formula established:

${Σ Σ}_{i i = = 11}^{N N} | | {E E.}_{pi p}^{βt βt} | | = = {E E.}_{p p}^{β β} - - - - - - ((77))$

5、情感强度5. Emotional intensity

设

并定义三种运算：+，-，Δ。set up

And define three operations: +, -, Δ.

其中，in,

+，-运算和实数域R中的加法和减法运算相似，但加法不具有交换律性质。即：

时，

The +, - operations are similar to the addition and subtraction operations in the real number field R, but the addition does not have the commutative property. Right now:

hour,

Δ运算是这样一种运算：

The delta operation is such an operation:

由于情感可划分为心境(Mood)和情绪(Emotion)，则t时刻的情感强度定义为：Since emotion can be divided into mood (Mood) and emotion (Emotion), the emotional intensity at time t is defined as:

${E E.}_{p p}^{t t} = = [[{E E.}_{p p 11}^{t t},, {E E.}_{p p 22}^{t t},, . . . . . .,, {E E.}_{pN n}^{t t}]]$

$= = Δ Δ (([[{E E.}_{}^{p p 11},, {E E.}_{p p 22}^{αt αt},, . . . . . .,, {E E.}_{pN n}^{αt αt}]],, [[{E E.}_{}^{p p 11},, {E E.}_{p p 22}^{βt βt},, . . . . . .,, {E E.}_{pN n}^{βt βt}]])) - - - - - - ((88))$

$= = [[Δ Δ (({E E.}_{}^{p p 11},, {E E.}_{}^{p p 11})),, Δ Δ (({E E.}_{p p 22}^{αt αt},, {E E.}_{p p 22}^{βt βt})),, . . . . . .,, Δ Δ (({E E.}_{pN n}^{αt αt},, {E E.}_{pN n}^{βt βt}))]]$

由式(8)可知，t时刻的情感强度与心境强度和情绪强度有关，根据上述的4个转移过程，它们的求解在下面分别进行介绍。It can be seen from formula (8) that the emotional intensity at time t is related to mood intensity and emotional intensity. According to the above four transfer processes, their solutions are introduced below.

二、个性化情感模型2. Personalized emotion model

基于上述理论基础，本申请提出如下个性化情感模型：Based on the above theoretical basis, this application proposes the following personalized emotion model:

1、心境自发转移1. Spontaneous transfer of mood

Laresn(1987)认为，平均的、稳定的心境特征并没有真实的反映个体的心境特征，心境随着时间的推移，其性质(好或不好)以及强度(弱或极度)是变化的。Pakrinsno(1996)认为与心境动态性有关的理论有动态平衡理论、社会牵引理论和非线性动态理论。并把影响心境的因素可以分作三类，其中之一来自个体内源性因素，如人格和生理因素。Laresn (1987) believes that the average and stable mood characteristics do not truly reflect the individual mood characteristics, and the mood changes over time in its nature (good or bad) and intensity (weak or extreme). Pakrinsno (1996) believed that the theories related to mood dynamics include dynamic balance theory, social traction theory and nonlinear dynamic theory. And the factors that affect mood can be divided into three categories, one of which comes from individual endogenous factors, such as personality and physiological factors.

本发明中的心境自发转移模型就是从个体内源性角度建立的。The mood spontaneous transfer model in the present invention is established from the perspective of individual endogenous.

人格对心境的影响：The influence of personality on mood:

对不同的人而言，所体验到的心境可能不同，换句话说存在着个体差异，这种差异来自于相对稳定的人格。同时，人格特征又决定了心境水平和心境变化性，心境自发地在相对稳定的心境特征水平附近波动，我们用C表示由人格决定的心境特征水平，如图3下半部分中的横线描述。For different people, the mood they experience may be different, in other words, there are individual differences, which come from relatively stable personalities. At the same time, personality characteristics determine the level of mood and the variability of mood. Mood spontaneously fluctuates around a relatively stable level of mood characteristics. We use C to represent the level of mood characteristics determined by personality, as shown by the horizontal line in the lower part of Figure 3 .

生理因素对心境的影响：Effects of physiological factors on mood:

个体的生理性周期变化会引起心境在积极与消极之间随着时间t波动。The individual's physiological cycle changes will cause the mood to fluctuate between positive and negative over time t.

1)心境的昼夜波动：根据Watson(2000)的研究，积极心境在一天中的趋势是早上较低，而后在一天中的某个时间上升到最大值；接着逐渐下降，在晚上达到最低。此过程用一余弦函数σcos(ω₁·t)表示。其中，σ是心境的昼夜影响因子，2π/ω₁是心境的昼夜波动周期。1) Diurnal fluctuation of mood: According to Watson's (2000) research, the trend of positive mood in a day is lower in the morning, then rises to the maximum at a certain time of the day; then gradually decreases, and reaches the lowest in the evening. This process is represented by a cosine function σcos(ω ₁ ·t). Among them, σ is the diurnal influence factor of mood, and 2π/ω ₁ is the diurnal fluctuation cycle of mood.

2)心境的周变化：根据Larsen & Kasimatis(1990)的研究发现，具有正弦波的7天间隔解释了日常心境的变化。积极心境在周五达到顶峰，在周二处于最低。此过程用一正弦函数ζsin(ω₂·t)表示。其中，ζ是心境的周变化影响因子，2π/ω₂是心境的周变化周期。2) Weekly variation of mood: According to the research of Larsen & Kasimatis (1990), a 7-day interval with a sine wave explains the daily variation of mood. Positive moods peak Friday and are at their lowest Tuesday. This process is represented by a sine function ζsin(ω ₂ ·t). Among them, ζ is the factor affecting the weekly change of mood, and 2π/ω ₂ is the period of weekly change of mood.

3)心境的月变化：主要是针对女性的月经周期循环的心境效应，因此，此项对心境的影响是个性化的，具有性别差异。此过程用一正弦函数τsin(ω₃·t)表示。其中，τ是心境的月变化影响因子，2π/ω₃是心境的月变化周期。3) Monthly change of mood: It is mainly aimed at the mood effect of women's menstrual cycle cycle. Therefore, the impact of this item on mood is individualized and has gender differences. This process is represented by a sine function τsin(ω ₃ ·t). Among them, τ is the influence factor of the monthly change of mood, and 2π/ω ₃ is the monthly change cycle of mood.

4)心境的季节变化：Watson(2000)认为在理论上积极情感应该有一个显著的季节模式。春季的积极心境水平较高，接着在夏季和秋季逐渐下降，最终达到冬季的最低点。此过程用一正弦函数υsin(ω₄·t)表示。其中，υ是心境的季节变化影响因子，2π/ω₄是心境的季节变化周期。4) Seasonal variation of mood: Watson (2000) believed that theoretically positive emotion should have a significant seasonal pattern. Positive mood levels were higher in spring, then gradually decreased in summer and autumn, and finally reached the lowest point in winter. This process is represented by a sine function υsin(ω ₄ ·t). Among them, υ is the factor affecting the seasonal variation of mood, and 2π/ω ₄ is the seasonal variation cycle of mood.

人格和生理因素对心境的动态变化共同产生影响，设心境的人格影响因子为ψ，心境的生理因素影响因子为ξ，且有：Personality and physiological factors jointly affect the dynamic changes of mood, let the personality influencing factor of mood be ψ, and the influencing factor of physiological factors of mood be ξ, and there are:

ψ∈(0，1)，ξ∈(0，1)，ψ+ξ＝1 (9)ψ∈(0, 1), ξ∈(0, 1), ψ+ξ=1 (9)

建立的心境自发转移模型为：The established mood spontaneous transfer model is:

${M m}_{p p}^{βt βt} = = ξ ξ \cdot &Center Dot; [[σ σ cos cos (({ω ω}_{11} \cdot &Center Dot; t t)) + + ζ ζ sin sin (({ω ω}_{22} \cdot &Center Dot; t t)) + + τ τ sin sin (({ω ω}_{33} \cdot &Center Dot; t t)) + + &upsi; &upsi; sin sin (({ω ω}_{44} \cdot &Center Dot; t t))]] + + ψ ψ \cdot \cdot C C - - - - - - ((1010))$

M_p ^βt动态变化范围的确定：Determination of the dynamic range of M _p ^βt :

∵ $\{\begin{matrix} σ \cos (ω_{1} \cdot t) &Element; [- σ, + σ] \\ ζ \sin (ω_{2} \cdot t) &Element; [- ζ, + ζ] \\ τ \sin (ω_{3} \cdot t) &Element; [- τ, + τ] \\ &upsi; \sin (ω_{4} \cdot t) &Element; [- &upsi;, + &upsi;] \end{matrix}$ ∵ $\{\begin{matrix} σ \cos (ω_{1} \cdot t) &Element; [- σ, + σ] \\ ζ \sin (ω_{2} \cdot t) &Element; [- ζ, + ζ] \\ τ \sin (ω_{3} &Center Dot; t) &Element; [- τ, + τ] \\ &upsi; \sin (ω_{4} &Center Dot; t) &Element; [- &upsi;, + &upsi;] \end{matrix}$

∴[σcos(ω₁·t)+ζsin(ω₂·t)+τsin(ω₃·t)+υsin(ω₄·t)]∴[σcos(ω ₁ t)+ζsin(ω ₂ t)+τsin(ω ₃ t)+υsin(ω ₄ t)]

∈[(-σ-ζ-τ-υ)，(+σ+ζ+τ+υ)]∈[(-σ-ζ-τ-υ), (+σ+ζ+τ+υ)]

＝[-(σ+ζ+τ+υ)，(σ+ζ+τ+υ)]=[-(σ+ζ+τ+υ), (σ+ζ+τ+υ)]

＝[-1，1]= [-1, 1]

∵C∈[-1，+1]∵C ∈ [-1, +1]

∴ $M_{p}^{βt} &Element; [(- ξ - ψ), (ξ + ψ)] = [- (ξ + ψ), (ξ + ψ)] = [- 1, + 1] - - - (11)$ ∴ $m_{p}^{βt} &Element; [(- ξ - ψ), (ξ + ψ)] = [- (ξ + ψ), (ξ + ψ)] = [- 1, + 1] - - - (11)$

以上参数中，ω_i，i∈{1，2，3，4}取值较大时，相应的生理性周期变化引起的心境随着时间t的波动周期越小，即心境变化性越强。因此，针对儿童用户来说，其值应取较大值。Among the above parameters, when the value of ω _i , i∈{1, 2, 3, 4} is larger, the fluctuation period of the mood caused by the corresponding physiological cycle changes with time t is smaller, that is, the mood variability is stronger. Therefore, for child users, the value should be a larger value.

2、心境刺激转移2. Mood stimulus transfer

在外部因素事件和情境刺激下，心境和情绪都受到影响，但其变化过程是有区别的：第一，持续时间上的差别；第二，相对强度上的差别；第三，信号功能上的差异。在本发明中，对于情绪受激后的变化过程，利用HMM这个双重随机过程来构造情绪状态刺激转移过程的情感模型，用HMM的前向和后向算法来模拟情绪在外界刺激下的变化规律，这一方法在后面将会介绍；对于心境受激后的变化过程，则采用控制论的调整策略进行研究。Under external factors, events and situational stimuli, both mood and emotion are affected, but the change process is different: first, the difference in duration; second, the difference in relative intensity; third, the difference in signal function difference. In the present invention, for the change process after the emotion is stimulated, the double stochastic process of HMM is used to construct the emotional model of the emotional state stimulation transfer process, and the forward and backward algorithms of HMM are used to simulate the change law of emotion under external stimulation , this method will be introduced later; for the change process of emotional stimulation, the adjustment strategy of cybernetics is used to study.

Larsen(2000)提出应将控制理论应用到心境调节的动态过程中。本发明根据此观点对心境的刺激转移过程建模。Larsen (2000) proposed that control theory should be applied to the dynamic process of mood regulation. The present invention models the stimulus transfer process of moods from this perspective.

根据已有研究，心境对刺激的反应强度是略微平缓的，一阶惯性环节更适合描述。According to previous studies, the intensity of mood responses to stimuli is slightly flat, and the first-order inertial link is more suitable for description.

本发明中，定义心境的刺激转移模型为：In the present invention, the stimulus transfer model of defining state of mind is:

$T T {\overset{\cdot \cdot}{M m}}_{p p}^{βt βt} + + {M m}_{p p}^{βt βt} = = Event Event ((t t)) - - - - - - ((1212))$

其中，T称为心境转移时间常数，它是表征心境发生转移惯性的一个重要参数。有研究发现，女性可能比男性更容易受情绪传染或影响，因此，T参数也是男女性别差异对心境的影响参数之一，是个性化参数。后面将会介绍此参数对心境激发子过程的影响。Event(t)是外源性因素(如工作方式、生活事件、家中变故等)对心境的影响强度。Among them, T is called the time constant of mood shift, which is an important parameter to characterize the inertia of mood shift. Studies have found that women may be more susceptible to emotional contagion or influence than men. Therefore, the T parameter is also one of the parameters that affect the mood of gender differences between men and women, and it is a personalized parameter. The effect of this parameter on the mood eliciting subprocess will be described later. Event(t) is the intensity of influence of exogenous factors (such as work style, life events, family changes, etc.) on mood.

心境的刺激转移过程分为两个子过程：The stimulus transfer process of mood is divided into two sub-processes:

1)心境激发子过程：1) mood eliciting sub-process:

该子过程与一个零状态响应过程类似。假设心境在t时刻受某一外源性因素激发时，心境强度初值为

则事件影响强度为：This sub-process is similar to a zero-status response process. Assuming that the mood is stimulated by an exogenous factor at time t, the initial value of the mood intensity is

Then the impact intensity of the event is:

在此条件下，求解式(12)表示的心境刺激转移模型。Under this condition, solve the mood stimulus transfer model represented by formula (12).

$L L [[T T {\overset{\cdot \cdot}{M m}}_{p p}^{βt βt} + + {M m}_{p p}^{βt βt} = = Event Event ((t t))]]$

$&DoubleRightArrow; &DoubleRightArrow; T T ((s the s {M m}_{p p}^{βs βs} - - {M m}_{p p}^{β β 00})) + + {M m}_{p p}^{βs βs} = = Event Event ((s the s)) &DoubleRightArrow; &DoubleRightArrow; T T ((s the s {M m}_{p p}^{βs βs} - - 00)) + + {M m}_{p p}^{βs βs} = = Event Event ((s the s))$

$&DoubleRightArrow; &DoubleRightArrow; ((Ts Ts + + 11)) {M m}_{p p}^{βs βs} = = \frac{Event Event ((t t))}{s the s} &DoubleRightArrow; &DoubleRightArrow; {M m}_{p p}^{βs βs} = = \frac{Event Event ((t t))}{s the s} \cdot &Center Dot; \frac{11}{((Ts Ts + + 11))} - - - - - - ((1414))$

∴ $M_{p}^{βt} = L^{- 1} [\frac{Event (t)}{s} \cdot \frac{1}{(Ts + 1)}] = Event (t) \cdot [1 - e^{- t / T}]$ ∴ $m_{p}^{βt} = L^{- 1} [\frac{event (t)}{the s} \cdot \frac{1}{(Ts + 1)}] = event (t) &Center Dot; [1 - e^{- t / T}]$

其中，L[·]，L^-1[·]分别是拉普拉斯变换和反变换。Among them, L[·], L ^-1 [·] are Laplace transform and inverse transform respectively.

根据心理学中情绪反应的时间动力性的基本概念，可以定义心境中相应的概念。According to the basic concept of time dynamics of emotional response in psychology, the corresponding concept in mood can be defined.

定义：称T_s为心境反应调节时间，T_s≈4T。这个参数表达了心境从初始强度

变化到-0.98或+0.98(-1或+1的±2％)的最短时间。Definition: T _s is called the mood response adjustment time, T _s ≈ 4T. This parameter expresses the mood from the initial intensity

Minimum time to change to -0.98 or +0.98 (±2% of -1 or +1).

定义：称T_d为心境反应延迟时间，T_d≈0.69T。这个参数表达了心境从初始强度第一次达到Event(t)×50％所需的时间。Definition: T _d is called the mood response delay time, T _d ≈ 0.69T. This parameter expresses the mood from the initial intensity The time required to reach Event(t) × 50% for the first time.

定义：称T_r为心境反应上升时间，Y_r≈2.20T。这个参数表达了心境从强度

第一次上升达到

所需的时间(Event∈积极情绪事件)，或从强度

第一次下降达到

所需的时间(Event∈消极情绪事件)。Definition: T _r is called the mood response rise time, Y _r ≈ 2.20T. This parameter expresses mood from intensity

first ascent reached

Desired time (Event ∈ Positive Emotional Event), or from intensity

The first drop reaches

Desired time (Event ∈ Negative Emotional Event).

根据上述的三个定义，可以看到心境反应调节时间的快慢，心境反应延迟时间和心境反应上升时间的长短，可能包含了重要的个体差异信息。这三个值都与心境转移时间常数T有关。因此，T参数是男女性别差异对心境的影响参数之一，是个性化参数，其大小对心境激发子过程是有影响的。According to the above three definitions, it can be seen that the adjustment time of mood response, the delay time of mood response and the length of rise time of mood response may contain important individual difference information. All three values are related to the mood transition time constant T. Therefore, the T parameter is one of the influencing parameters of gender differences on mood, and it is a personalized parameter, and its size has an impact on the sub-process of mood stimulation.

图4是当T＝0.5和T＝0.9，Event(t)＝1时，心境激发子过程。从图中可以看出，心境反应调节时间、心境反应延迟时间和心境反应上升时间均不同，反映了男女性别差异对心境的影响。Fig. 4 is the mood excitation sub-process when T=0.5 and T=0.9, Event(t)=1. It can be seen from the figure that the adjustment time of mood response, the delay time of mood response and the rise time of mood response are all different, reflecting the impact of gender differences on mood.

心境强度M_p ^βt在某一外源性因素的持续激发下不断变大，表现出了事件的影响随时间的积累作用。但其变化率却不断变小，即：影响随时间变小，表明了心境在某一特定时间刺激下，越来越不敏感，对此事件的发生变得麻木，如图5所示。Mood intensity M _p ^βt is constantly increasing under the continuous stimulation of an exogenous factor, showing the cumulative effect of the event's influence over time. However, the rate of change keeps decreasing, that is, the influence decreases over time, indicating that the mood becomes less and less sensitive to the stimulation at a certain time, and becomes numb to the occurrence of this event, as shown in Figure 5.

2)心境衰减子过程：2) Mood decay sub-process:

该子过程与一个零输入响应过程类似，是在心境激发子过程之后的。假设在t时刻外源性因素对心境的影响消失Event(t)＝0，消失时的心境强度初值为 This sub-process is similar to a zero-input response process and follows the mood-stimulating sub-process. Assuming that the influence of exogenous factors on the mood disappears at time t Event(t)=0, the initial value of the mood intensity when it disappears is

$&DoubleRightArrow; &DoubleRightArrow; T T ((s the s {M m}_{p p}^{βs βs} - - {M m}_{p p}^{β β {t t}_{00}})) + + {M m}_{p p}^{βs βs} = = 00 &DoubleRightArrow; &DoubleRightArrow; {((Ts Ts + + 11)) M m}_{p p}^{βs βs} = = {TM tm}_{p p}^{β β {t t}_{00}}$

$&DoubleRightArrow; &DoubleRightArrow; {M m}_{p p}^{βs βs} = = {TM tm}_{p p}^{β β {t t}_{00}} \cdot &Center Dot; \frac{11}{((Ts Ts + + 11))} - - - - - - ((1515))$

∴ $M_{p}^{βt} = L^{- 1} [{TM}_{p}^{β t_{0}} \cdot \frac{1}{(Ts + 1)}] = M_{p}^{β t_{0}} \cdot e^{- t / T}$ ∴ $m_{p}^{βt} = L^{- 1} [{tm}_{p}^{β t_{0}} \cdot \frac{1}{(Ts + 1)}] = m_{p}^{β t_{0}} &Center Dot; e^{- t / T}$

3、情绪自发转移3. Spontaneous transfer of emotions

根据滕少冬的研究，在概率空间的基础上，把情绪的变化过程看成一个随机过程，并进一步用马尔科夫链来描述情绪状态自发转移过程，给出了基本方程以及计算方法。此过程的随机过程模型，是建立在情感状态概率空间的基础上，用λ＝(N，M，π，A，B)描述。其中，N是情绪种类；M是刺激种类；π是初始状态概率分布向量为；状态转移概率矩阵为

观察值概率矩阵为

According to Teng Shaodong's research, on the basis of probability space, the emotional change process is regarded as a random process, and the Markov chain is further used to describe the spontaneous transfer process of the emotional state, and the basic equation and calculation method are given. The stochastic process model of this process is based on the emotional state probability space, described by λ=(N, M, π, A, B). Among them, N is the emotion type; M is the stimulus type; π is the initial state probability distribution vector; the state transition probability matrix is

The observation probability matrix is

4、情绪刺激转移4. Emotional stimulus transfer

根据滕少冬的研究，利用HMM这个双重随机过程来构造情绪状态刺激转移过程的情感模型，用HMM的前向和后向算法来模拟情绪在外界刺激下的变化规律。自发转移过程的马尔可夫链模型是：P^(t+1)＝P^(t)A。According to Teng Shaodong's research, the double stochastic process of HMM is used to construct the emotion model of the emotional state stimulus transfer process, and the forward and backward algorithms of HMM are used to simulate the change law of emotion under external stimuli. The Markov chain model of the spontaneous transfer process is: P ^(t+1) = P ^(t) A.

其中，

是状态转移矩阵；P^(t)是前一时刻的情感概率向量；P^(t+1)是后一时刻的情感概率向量。in,

is the state transition matrix; P ^(t) is the emotion probability vector at the previous moment; P ^(t+1) is the emotion probability vector at the next moment.

5、整合后t时刻情感强度模型5. Emotional intensity model at time t after integration

由于情感可划分为心境(Mood)和情绪(Emotion)，则t时刻的情感强度定义为：Since emotion can be divided into mood (Mood) and emotion (Emotion), the emotion intensity at time t is defined as:

为了将上述模型实现软件化应用，本发明还提出了一种基于上述模型的一种应用方法，其特征在于所述方法包括如下步骤：In order to implement the software application of the above-mentioned model, the present invention also proposes an application method based on the above-mentioned model, which is characterized in that the method includes the following steps:

(1)设置情绪状态的初始值；(1) Setting the initial value of the emotional state;

(2)设置心境状态的初始值；(2) setting the initial value of the state of mind;

(3)设置定时器Timer1的初始值；(3) the initial value of timer Timer1 is set;

(4)判断定时器Timer1的计时是否到1s，如未到，则继续计时，如到1s，则执行定时器Timer1的响应函数。(4) Determine whether the timing of the timer Timer1 reaches 1s, if not, continue timing, if it reaches 1s, execute the response function of the timer Timer1.

所述步骤(4)中定时器Timer1的响应函数流程为：The response function flow process of timer Timer1 in described step (4) is:

(1)计算情绪强度；(1) Calculation of emotional intensity;

(2)计算心境强度；(2) Calculate mood intensity;

(3)计算情感强度；(3) Calculate emotional intensity;

(4)根据情感滤波器，选择t时刻的输出情感；(4) Select the output emotion at time t according to the emotion filter;

(5)绘制情感强度曲线。(5) Draw the emotional intensity curve.

其中，计算情绪强度部分的流程为：1)判断是否受到激发，如未受到激发，则进行情绪自发转移过程计算输出情绪值Emotion；2)如果受到激发，则将当前情感值设置为刺激转移初始情感；3)获取外界刺激事件类型与强度；4)进行情绪刺激转移过程计算并输出情绪值Emotion。Among them, the process of calculating the emotional intensity part is: 1) judge whether it is stimulated, if it is not stimulated, then carry out the process of emotional spontaneous transfer to calculate and output the emotional value Emotion; 2) if it is stimulated, set the current emotional value as the initial stimulus transfer Emotion; 3) Obtain the type and intensity of external stimulus events; 4) Calculate and output the emotional value Emotion in the process of transferring emotional stimuli.

计算心境强度部分的流程为：1)进行心境自发转移过程的计算；2)输出自发转移过程的心境值SelfMood；3)判断是否受到激发，如果受到，则获取当前心境值，计算心境刺激转移中受激子过程并输出心境值TriggerMood；4)如果未受到激发，则获取当前心境值，计算心境刺激转移中衰减子过程并输出心境值TriggerMood。The process of calculating the mood strength part is: 1) Carry out the calculation of the spontaneous mood transfer process; 2) Output the mood value SelfMood of the spontaneous transfer process; 3) Judge whether it is stimulated, if so, obtain the current mood value, and calculate the mood stimulus transfer process The stimulated sub-process and output the mood value TriggerMood; 4) If not stimulated, obtain the current mood value, calculate the attenuation sub-process in the mood stimulus transfer and output the mood value TriggerMood.

附图说明 Description of drawings

图1是心境与情绪状态转移图。Figure 1 is a diagram of mood and emotional state transitions.

图2心理能量以及情感能量的各种形式以及与情绪强度之间的转化关系。Figure 2 The various forms of psychological energy and emotional energy and the transformation relationship with emotional intensity.

图3心理能量以及情感能量的各种形式以及与心境强度之间的转化关系。Figure 3 The various forms of psychological energy and emotional energy and the transformation relationship with mood intensity.

图4是心境激发子过程。Figure 4 is the mood eliciting sub-process.

图5是心境强度变化率。Figure 5 is the rate of change of mood intensity.

图6是本发明的模型应用方法主流程图。Fig. 6 is a main flow chart of the model application method of the present invention.

图7是定时器Timer1的响应函数流程图。FIG. 7 is a flow chart of the response function of the timer Timer1.

图8是针对儿童用户的玩伴机器人个性化情感模型软件。Fig. 8 is a playmate robot personalized emotion model software for children users.

具体实施方式 Detailed ways

本发明中，具体实现3个情感维度(高兴，愤怒和害怕)的计算。基于上述对四个模型的讨论，具体参数的选择以及具体实施方式如下：In the present invention, the calculation of three emotional dimensions (joy, anger and fear) is specifically realized. Based on the above discussion of the four models, the selection of specific parameters and the specific implementation methods are as follows:

心境状态自发转移的动态平衡模型中，人格决定的心境特征水平C由多种因素影响，在积极消极分界线左右一定范围内变动，本发明中假设C∈[-1，+1]。心境的昼夜影响因子、周变化影响因子、月变化影响因子以及季节变化影响因子σ，ζ，τ，υ∈[0，1]，且有σ+ζ+τ+υ＝1。ω₁＝7ω₂＝30ω₃＝365ω₄。本发明中，对于男性用户，取σ＝0.7，ζ＝0.2，τ＝0，υ＝0.1；对于女性用户，取σ＝0.5，ζ＝0.2，τ＝0.2，υ＝0.1。由于心境是变化缓慢且某一心境常常能持续一段时间，因此，在本发明中，每隔1小时计算一次心境量值，心境的昼夜波动模型中，周期取24小时，则ω₁＝2π/24。此外取心境的人格影响因子ψ＝0.5，心境的生理因素影响因子ξ＝0.5。当参数确定后，心境状态自发转移的动态平衡模型就建立完成了。In the dynamic balance model of spontaneous transfer of mood states, the mood characteristic level C determined by personality is affected by various factors, and changes within a certain range around the positive and negative dividing line. In the present invention, it is assumed that C∈[-1, +1]. The diurnal, weekly, monthly, and seasonal factors affecting mood are σ, ζ, τ, υ∈[0, 1], and σ+ζ+τ+υ=1. ω ₁ =7ω ₂ =30ω ₃ =365ω ₄ . In the present invention, for male users, σ=0.7, ζ=0.2, τ=0, υ=0.1; for female users, σ=0.5, ζ=0.2, τ=0.2, υ=0.1. Because the state of mind changes slowly and a certain state of mind can often last for a period of time, therefore, in the present invention, the state of mind value is calculated every 1 hour. In the diurnal fluctuation model of the state of mind, the cycle is 24 hours, then ω ₁ =2π/ twenty four. In addition, the personality influencing factor of mood is ψ=0.5, and the influencing factor of physiological factors of mood is ξ=0.5. When the parameters are determined, the dynamic equilibrium model of the spontaneous transfer of mood states is established.

心境状态刺激转移的控制论模型中，T参数是男女性别差异对心境的影响参数之一，本发明中，取T＝0.5。In the cybernetic model of mood state stimulus transfer, T parameter is one of the influence parameters of gender difference on mood. In the present invention, T=0.5.

为了便于计算机实现心境刺激转移模型，在本发明中，需要求解微分方程式(12)的差分方程形式。In order to facilitate the computer to realize the mood stimulus transfer model, in the present invention, it is necessary to solve the difference equation form of the differential equation (12).

根据式(12)，有According to formula (12), we have

$T T \frac{{dM D}_{p p}^{βt βt}}{dt dt} + + {M m}_{p p}^{βt βt} = = Event Event ((t t)) - - - - - - ((1616))$

$\frac{{dM D}_{p p}^{βt βt}}{dt dt} = = \frac{Event Event ((t t)) - - {M m}_{p p}^{βt βt}}{T T} - - - - - - ((1717))$

根据欧拉法，可得：According to Euler's method, we can get:

${M m}_{p p}^{β β ((k k + + 11))} = = {M m}_{p p}^{βk βk} + + h h \cdot &Center Dot; \frac{Event Event ((k k)) - - {M m}_{p p}^{βk βk}}{T T} - - - - - - ((1818))$

其中，h称为步长，是计算心境强度的间隔时间，本发明中，取h＝0.02。Wherein, h is called the step length, which is the interval time for calculating the mood intensity. In the present invention, h=0.02.

这种微分方程差分化，可以为人工心理模型在计算机上的应用带来方便，但是却会带来一定误差，如表1，2所示。This differentiation of differential equations can bring convenience to the application of artificial mental models on computers, but it will bring certain errors, as shown in Tables 1 and 2.

表1 差分化后的误差(心境激发子过程)Table 1 Error after differential differentiation (mood eliciting subprocess)

表2 差分化后的误差(心境衰减子过程)Table 2 Error after differential differentiation (mood decay subprocess)

从表1、2中看到，微分方程差分化后的误差为10^-3数量级，因此，本发明中，仍采用差分的方法计算儿童玩伴机器人在t时刻的情感强度，以利于情感计算的计算机实现。It can be seen from Tables 1 and 2 that the error after differential equation differentiation is on the order of 10 ^-3 , therefore, in the present invention, the method of difference is still used to calculate the emotional intensity of the children's playmate robot at time t, so as to facilitate the calculation of emotion computer implementation.

情绪状态自发转移过程的马尔可夫链模型中，状态转移矩阵 $\hat{A} = [\begin{matrix} {\hat{a}}_{11} & {\hat{a}}_{12} & {\hat{a}}_{13} \\ {\hat{a}}_{21} & {\hat{a}}_{22} & {\hat{a}}_{23} \\ {\hat{a}}_{31} & {\hat{a}}_{32} & {\hat{a}}_{33} \end{matrix}] = [\begin{matrix} 0.94 & 0.03 & 0.03 \\ 0.03 & 0.94 & 0.03 \\ 0.03 & 0.03 & 0.94 \end{matrix}] .$ In the Markov chain model of the emotional state spontaneous transition process, the state transition matrix $\hat{A} = [\begin{matrix} {\hat{a}}_{11} & {\hat{a}}_{12} & {\hat{a}}_{13} \\ {\hat{a}}_{twenty one} & {\hat{a}}_{twenty two} & {\hat{a}}_{twenty three} \\ {\hat{a}}_{31} & {\hat{a}}_{32} & {\hat{a}}_{33} \end{matrix}] = [\begin{matrix} 0.94 & 0.03 & 0.03 \\ 0.03 & 0.94 & 0.03 \\ 0.03 & 0.03 & 0.94 \end{matrix}] .$

情绪状态刺激转移过程的随机过程模型中，HMM模型为λ＝(N，M，π，A，B)。由于情感维度是3，所以N＝3；某种刺激确定性地只引发某一种情绪，所以M＝N＝3。初始初始状态概率分布向量为π[π₁ π₂ π₃]＝[1/3 1/3 1/3]；状态转移概率矩阵为In the stochastic process model of emotional state stimulus transfer process, the HMM model is λ=(N, M, π, A, B). Since the emotional dimension is 3, N=3; a certain stimulus can only trigger a certain emotion, so M=N=3. The initial initial state probability distribution vector is π[π ₁ π ₂ π ₃ ]=[1/3 1/3 1/3]; the state transition probability matrix is

观察值概率矩阵为：

The observation probability matrix is:

最大刺激强度T_max＝55。

The maximum stimulation intensity T _max =55.

基于以上的情绪状态刺激转移模型、情绪状态自发转移模型、心境状态刺激转移模型、心境状态自发转移模型和情感滤波器，用VC++和SQL Server开发针对儿童用户的玩伴机器人个性化情感模型软件。Based on the above emotional state stimulus transfer model, emotional state spontaneous transfer model, mood state stimulus transfer model, mood state spontaneous transfer model and emotional filter, use VC++ and SQL Server to develop personalized emotional model software for playmate robots for children.

此个性化人工情感软件，包括情绪状态自发转移、情绪状态刺激转移、心境状态自发转移和心境状态刺激转移四部分。每一部分都包括运行参数(初始值)的设定、情绪(心境)的计算和输出几个基本环节。The personalized artificial emotion software includes four parts: emotional state spontaneous transfer, emotional state stimulus transfer, mood state spontaneous transfer and mood state stimulus transfer. Each part includes the setting of operating parameters (initial value), the calculation and output of emotions (mood).

软件的主流程为：1)设置情绪状态的初始值；2)设置心境状态的初始值；3)设置定时器Timer1的初始值；4)判断定时器Timer1的计时是否到1s，如未到，则继续计时，如到1s，则执行定时器Timer1的响应函数。The main process of the software is: 1) setting the initial value of the emotional state; 2) setting the initial value of the state of mind; 3) setting the initial value of the timer Timer1; 4) judging whether the timing of the timer Timer1 has reached 1s, if not, Then continue timing, if it reaches 1s, execute the response function of Timer1.

上述四个模型的计算过程都放置在了定时器Timer1的响应函数中。The calculation process of the above four models is placed in the response function of the timer Timer1.

定时器Timer1的响应函数流程为：1)计算情绪强度；2)计算心境强度；3)计算情感强度；4)根据情感滤波器，选择t时刻的输出情感；5)绘制情感强度曲线。The response function process of timer Timer1 is: 1) Calculate the emotional intensity; 2) Calculate the mood intensity; 3) Calculate the emotional intensity; 4) Select the output emotion at time t according to the emotional filter; 5) Draw the emotional intensity curve.

软件的用户交互界面包括两个部分：结果展示区和参数调整区。结果展示区主要用来显示情绪、心境和情感状态的。在Picture控件上绘图，通过定时器Timer1的运行，不断更新情绪、心境和情感状态。The user interface of the software includes two parts: the result display area and the parameter adjustment area. The result display area is mainly used to display emotions, moods and emotional states. Draw on the Picture control, and constantly update the mood, mood and emotional state through the operation of the timer Timer1.

参数调整区主要是进行模型的参数设定。用Slider控件、Combo Box控件和Edit Box控件完成上述模型参数的初始化工作。The parameter adjustment area is mainly for parameter setting of the model. Use the Slider control, Combo Box control and Edit Box control to complete the initialization of the above model parameters.

在上述个性化的情感模型软件化后，嵌入到儿童玩伴机器人中，使其在后台运行。接受非结构化环境与用户的信息输入后，通过个性化的情感软件模块，产生个性化的输出，实现针对儿童用户的玩伴机器人个性化交互。After the above-mentioned personalized emotional model is softwareized, it is embedded in the children's playmate robot to make it run in the background. After accepting the unstructured environment and the user's information input, through the personalized emotional software module, a personalized output is generated to realize the personalized interaction of the playmate robot for children.

Claims

1. The utility model provides an apply to individualized emotion model that children user played with robot, the model includes emotion model and mood model, is respectively:

(1) a stochastic process model of emotional state stimuli transfer process;

(2) a Markov chain model of a spontaneous emotional state transition process;

(3) a cybernetic model of mood state stimulus transfer;

(4) a dynamic equilibrium model of spontaneous shift of mood states.

The method is characterized in that: the dynamic balance model for spontaneous mood state transition is established by the following model:

the model established by the cybernetic model of the mood state stimulus transfer is as follows:

<math> <mrow> <mi>T</mi> <msubsup> <mover> <mi>M</mi> <mo>·</mo> </mover> <mi>p</mi> <mi>βt</mi> </msubsup> <mo>+</mo> <msubsup> <mi>M</mi> <mi>p</mi> <mi>βt</mi> </msubsup> <mo>=</mo> <mi>Event</mi> <mrow> <mo>(</mo> <mi>t</mi> <mo>)</mo> </mrow> <mo>;</mo> </mrow></math>

integrating the above four models, E_p ^αIs the emotional intensity, E_p ^βThe mood intensity and the emotional intensity value model at the time t are as follows:

2. the personalized emotion model applied to the child user player robot as recited in claim 1, wherein: in the dynamic equilibrium model of spontaneous shift of the mood state, the parameter sigma,

，τ，ν∈[0，1]，

ω₁＝7ω₂＝30ω₃＝365ω₄，C∈[-1，+1]Ψ ∈ (0, 1), ξ ∈ (0, 1), Ψ + ξ ═ 1, and it can be adjusted within an allowable range according to the actual situation.

3. The personalized emotion model applied to the child user player robot as recited in claim 1, wherein: the cybernetic model of the mood state stimulus transition comprises a mood-arousing cybernetic model which can be calculated by a computer and can be integrated with a mood model

Model of theory of harmony

4. A method for applying a personalized emotion model as defined in claim 1, characterised in that said method comprises the steps of:

(1) setting an initial value of an emotional state;

(2) setting an initial value of the mood state;

(3) setting an initial value of a timer Timell;

(4) and judging whether the timing of the timer Timrl reaches 1s, if not, continuing to count time, and if so, executing a response function of the timer Timrl.

The response function flow of the timer Timerl in the step (4) is as follows:

(1) calculating the emotional intensity;

(2) calculating the intensity of the mood;

(3) calculating the emotional intensity;

(4) selecting output emotion at the time t according to an emotion filter;

(5) and (6) drawing an emotional intensity curve.

5. The method for applying the personalized emotion model of claim 4, wherein: the process of calculating the emotional intensity part comprises the following steps: 1) judging whether the Emotion is excited or not, if not, calculating an Emotion value Emotion in the process of spontaneous Emotion transfer; 2) if the stimulation is triggered, setting the current emotion value as the stimulation transfer initial emotion; 3) acquiring the type and the intensity of an external stimulation event; 4) and calculating the emotional stimulus transfer process and outputting an emotional value Emotion.

6. The method for applying the personalized emotion model of claim 4, wherein: the process of calculating the mood intensity part comprises the following steps: 1) calculating the spontaneous mood transfer process; 2) outputting a mood value SelfMood in a spontaneous transfer process; 3) judging whether the stimulation is received, if so, acquiring a current mood value, calculating a stimulated sub process in mood stimulation transfer and outputting a mood value TriggerMood; 4) if not, obtaining the current mood value, calculating the attenuation subprocess in the mood stimulus transfer and outputting the mood value TriggerMood.