JP2022006610A - Social capacity generation device, social capacity generation method, and communication robot - Google Patents

Abstract

To provide a social capacity generation device, a social capacity generation method, and a communication robot capable of forming emotional connection between a robot and a human.SOLUTION: A social capacity generation device includes: recognition means for acquiring human information on a human, extracting feature information on the human from the acquired human information, recognizing an approach that occurs between a communication device for executing communication and a human, and recognizing an approach that occurs between a human and a human; learning means for learning emotional interactions of humans by a multimodal using the extracted feature information on the human; and motion generation means for generating an action on the basis of the information on the learned emotional interactions of humans.SELECTED DRAWING: Figure 1

Description

The present invention relates to a social ability generation device, a social ability generation method, and a communication robot.

Today, smart speakers and communication robots are being developed. Such systems focus on features such as turning lights on or off, accessing calendars, reading emails, and setting appointments, depending on instructions. In such a system, the input of instructions is limited to, for example, selection by a touch panel, commands defined by voice, and the like, and it is difficult to build a relationship with a person.

Therefore, a system that can have a relationship with people is desired. For example, Patent Document 1 proposes a system in which a person is involved in a dialogue or operation with a device in contrast to a dialogue between a companion device and a person. In the technology described in Patent Document 1, a companion device detects utterances and actions with a user and expresses them through movement, graphics, sound, light, and fragrance to provide an intimate existence.

Special Table 2019-521449 Gazette

However, with the technique described in Patent Document 1, it is difficult to form an emotional connection between a robot and a human.

The present invention has been made in view of the above problems, and is a social ability generation device, a social ability generation method, and a communication robot capable of forming an emotional connection between a robot and a human. The purpose is to provide.

(1) In order to achieve the above object, the social ability generation device according to one aspect of the present invention acquires human information about a person, extracts characteristic information about a person from the acquired personal information, and communicates. Multimodal human emotional interaction using cognitive means that recognize the action that occurs between the device and the person and the action that occurs between the person and the extracted characteristic information about the person. It is provided with a learning means learned by the above and a motion generating means for generating an action based on the learned emotional interaction information of the person.

(2) Further, in the social ability generation device according to one aspect of the present invention, the learning means performs learning using an implicit reward and an explicit reward, and the implicit reward is the above-mentioned. It is a reward learned by multimodal using characteristic information about a person, and the explicit reward is a reward based on the result of evaluating the behavior of the communication device generated by the motion generating means with respect to the person. You may do so.

(3) Further, in the social ability generation device according to one aspect of the present invention, the cognitive means includes a sound collecting unit that collects an acoustic signal and a photographing unit that captures an image including a user. The picked-up acoustic signal is subjected to voice recognition processing to extract characteristic information related to voice, and the captured image is subjected to image processing to extract characteristic information related to human behavior contained in the image. The characteristic information related to a person includes the characteristic information related to the voice and the characteristic information related to the human behavior, and the characteristic information related to the voice includes a voice signal, information on the volume of the voice, information on the intonation of the voice, and the meaning of the speech. The characteristic information related to the person's behavior is at least one, and the characteristic information regarding the person's behavior includes the person's facial expression information, the person's gesture information, the person's head posture information, the person's face orientation information, the person's line of sight information, and the person and person. It may be at least one of the distances between them.

(4) Further, in the social ability generation device according to one aspect of the present invention, the learning means is to learn using social norms, social components, psychological findings, and humanistic findings. You may.

(5) In order to achieve the above object, in the social ability generation method according to one aspect of the present invention, the cognitive means acquires human information about a person, extracts characteristic information about a person from the acquired human information, and then extracts the characteristic information about the person. The communication device that communicates and the action that occurs between people are recognized, the action that occurs between people is recognized, and the learning means uses the extracted characteristic information about the person to make the person emotional. The interaction is learned in a multimodal manner, and the action generation means generates an action based on the learned emotional interaction information of the person.

(6) In order to achieve the above object, the communication robot according to one aspect of the present invention acquires human information about a person, extracts characteristic information about a person from the acquired human information, and communicates with a communication device and a person. Multimodal learning of human emotional interactions using cognitive means that recognize the actions that occur between people and the actions that occur between people and the extracted characteristic information about the person. It includes a learning means and a motion generating means for generating an action based on the learned emotional interaction information of the person.

(7) Further, the communication robot according to one aspect of the present invention is provided with a display unit, and the motion generating means has a good relationship with a person by causing the person to behave in a way that maximizes positive emotions. An image that maintains the state may be generated, and the generated image may be displayed on the display unit.

According to (1) to (7), an emotional connection can be formed between a robot and a human.
According to (2), learning can be performed without using a lot of teaching data.
According to (3), a lot of information based on human reaction can be obtained.
According to (4), it is socially intelligent and can navigate social scenarios.
According to (7), it is possible to make a person behave in a way that maximizes positive emotions, and it is possible to maintain a good relationship with the person.

It is a figure which shows the communication example of the communication robot which concerns on embodiment. It is a block diagram which shows the configuration example of the communication robot which concerns on embodiment. It is a figure which shows the outline example of the communication robot which concerns on embodiment. It is a figure which shows the flow of the cognition, learning and the social ability performed by the social ability generation device of an embodiment. It is a figure which shows the example of the data which the cognitive part which concerns on embodiment recognizes. It is a figure which shows the system example which performs deep reinforcement learning using raw data in a comparative example. It is a figure which shows the example of the agent creation method used by the operation generation part which concerns on embodiment. It is a flowchart which shows the procedure example of the social capacity generation processing which concerns on embodiment. It is a figure which shows the communication example of a communication robot and a user which concerns on embodiment. It is a figure which shows the image example which is displayed on the display part at the time of communication of a communication robot and a user which concerns on embodiment. It is a figure which shows the image example which is displayed on the display part at the time of communication of a communication robot and a user which concerns on embodiment. It is a figure which shows the example of communication by a communication robot at the time of communication with a friend of a user which concerns on embodiment. It is a figure which shows the example which applied the communication robot of embodiment to the car navigation system in a vehicle. It is a figure which shows the connection example with various devices in a home when applied to the car navigation which concerns on embodiment.

Hereinafter, embodiments of the present invention will be described with reference to the drawings. In the drawings used in the following description, the scale of each member is appropriately changed in order to make each member recognizable.

<Overview>
FIG. 1 is a diagram showing a communication example of the communication robot 1 according to the present embodiment. As shown in FIG. 1, the communication robot 1 communicates with an individual or a plurality of people 2. Communication mainly consists of dialogue g11 and gesture g12 (movement). The operation is expressed by an image displayed on the display unit in addition to the actual operation. Further, when the e-mail is transmitted to the user via the Internet line or the like, the communication robot 1 receives the e-mail and notifies the user that the e-mail has arrived (g14). Further, the communication robot 1 makes a proposal g14 by communicating with the user whether advice is necessary, for example, when a reply to an e-mail is required. The communication robot 1 transmits a response (g15). Further, the communication robot 1 presents the weather forecast of the place according to the scheduled date and time and the place according to the schedule of the user, for example.

In the present embodiment, the social ability of the robot is generated so that an emotional connection can be formed between the robot and the human, and communication with the human is performed according to, for example, the reaction or behavior of the human. Then, in the present embodiment, the human and the robot sympathize with each other at the emotional level and communicate with each other. In this embodiment, so to speak, communication between a person and a pet is realized by learning social norms and the like. In the present embodiment, the above is realized by learning the social background (background) of the user, the interaction between people, and the like in communication.

<Configuration example of communication robot 1>
Next, a configuration example of the communication robot 1 will be described.
FIG. 2 is a block diagram showing a configuration example of the communication robot 1 according to the present embodiment. As shown in FIG. 2, the communication robot 1 includes a receiving unit 101, a photographing unit 102, a sound collecting unit 103, a sensor 104, a social ability generation device 100, a storage unit 106, a first database 107, a second database 109, and a display unit. It includes a 111, a speaker 112, an actuator 113, and a transmission unit 114.

The social ability generation device 100 includes a cognitive unit 105 (cognitive means), a learning unit 108 (learning means), and a motion generation unit 110 (motion generation means).
The motion generation unit 110 includes an image generation unit 1101, a voice generation unit 1102, a drive unit 1103, and a transmission information generation unit 1104.

<Functions and operations of communication robot 1>
Next, the functions and operations of each functional unit of the communication robot 1 will be described with reference to FIG.

The receiving unit 101 acquires information (for example, e-mail, blog information, news, weather forecast, etc.) from the Internet via a network, and outputs the acquired information to the cognitive unit 105 and the motion generation unit 110. Alternatively, for example, when the first database 107 is on the cloud, the receiving unit 101 acquires information from the first database 107 on the cloud and outputs the acquired information to the cognitive unit 105.

The photographing unit 102 is, for example, a CMOS (Complementary Metal Oxide Semiconductor) photographing element, a CCD (Charge Coupled Device) photographing element, or the like. The photographing unit 102 outputs the captured image (human information which is information about a person; still image, continuous still image, moving image) to the cognitive unit 105 and the motion generation unit 110. The communication robot 1 may include a plurality of photographing units 102. In this case, the photographing unit 102 may be attached to the front and the rear of the housing of the communication robot 1, for example.

The sound collecting unit 103 is, for example, a microphone array composed of a plurality of microphones. The sound collecting unit 103 outputs an acoustic signal (human information) collected by a plurality of microphones to the recognition unit 105 and the motion generation unit 110. The sound collecting unit 103 may sample each of the acoustic signals collected by the microphone with the same sampling signal, convert the analog signal into a digital signal, and then output the sound to the recognition unit 105.

The sensor 104 is, for example, a temperature sensor that detects the temperature of the environment, an illuminance sensor that detects the illuminance of the environment, a gyro sensor that detects the inclination of the housing of the communication robot 1, and an acceleration sensor that detects the movement of the housing of the communication robot 1. , A pressure sensor that detects pressure. The sensor 104 outputs the detected detection value to the recognition unit 105 and the motion generation unit 110.

The storage unit 106 stores, for example, items to be recognized by the recognition unit 105, various values (threshold values, constants) used at the time of recognition, an algorithm for performing recognition, and the like.

The first database 107 stores, for example, a language model database, an acoustic model database, a dialogue corpus database, and acoustic features used in speech recognition, a comparative image database and image features used in image recognition, and the like. do. Each data and feature amount will be described later. The first database 107 may be placed on the cloud or may be connected via a network.

The second database 109 stores data on the relationship between people, such as social components, social norms, social customs, psychology, and humanities, which are used during learning. The second database 109 may be placed on the cloud or may be connected via a network.

The social ability generation device 100 recognizes the action that occurs between the communication robot 1 and a person, or the action that occurs between a plurality of people, and the human emotion is based on the recognized content and the data stored in the second database 109. To learn the interaction. Then, the social ability generation device 100 generates the social ability of the communication robot 1 from the learned contents. The social ability is, for example, the ability to interact between people, such as dialogue, behavior, understanding, and empathy between people.

The cognitive unit 105 recognizes the action that occurs between the communication robot 1 and a person, or the action that occurs between a plurality of people. The cognitive unit 105 acquires an image captured by the photographing unit 102, an acoustic signal collected by the sound collecting unit 103, and a detection value detected by the sensor 104. The cognitive unit 105 may acquire the information received by the receiving unit 101. Based on the acquired information and the data stored in the first database 107, the cognitive unit 105 recognizes the action that occurs between the communication robot 1 and a person, or the action that occurs between a plurality of people. The cognitive method will be described later. The cognitive unit 105 outputs the recognized cognitive result (feature amount related to sound, feature information related to human behavior) to the learning unit 108. The cognitive unit 105 performs well-known image processing (for example, binarization processing, edge detection processing, clustering processing, image feature amount extraction processing, etc.) on the image captured by the imaging unit 102. The recognition unit 105 performs well-known voice recognition processing (sound source identification processing, sound source localization processing, noise suppression processing, voice section detection processing, sound source extraction processing, sound feature calculation processing, etc.) on the acquired acoustic signal. The cognitive unit 105 extracts the audio signal (or acoustic signal) of the target person, animal, or object from the acquired acoustic signal based on the recognized result, and recognizes the extracted audio signal (or acoustic signal). As a result, it is output to the motion generation unit 110. Based on the recognized result, the cognitive unit 105 extracts an image of a target person or an object from the acquired image, and outputs the extracted image to the motion generation unit 110 as a recognition result.

The learning unit 108 learns human emotional interactions using the cognitive results output by the cognitive unit 105 and the data stored in the second database 109. The learning unit 108 stores the model generated by the learning. The learning method will be described later.

The motion generation unit 110 acquires the information received from the reception unit 101, the image taken from the photographing unit 102, the acoustic signal collected from the sound collecting unit 103, and the recognition result from the recognition unit 105. The motion generation unit 110 generates actions (utterances, gestures, images) for the user based on the learned result and the acquired information.

The image generation unit 1101 generates an output image (still image, continuous still image, or moving image) to be displayed on the display unit 111 based on the learned result and the acquired information, and generates the generated output image. It is displayed on the display unit 111. As a result, the motion generation unit 110 causes the display unit 111 to display an animation such as a facial expression, presents an image to be presented to the user, and communicates with the user. The displayed image is an image corresponding to the movement of a person's eyes, an image corresponding to the movement of a person's mouth, information such as a user's destination (map, weather map, weather forecast, information on shops and recreational areas). Etc.), images of people who have made a TV call to the user via the Internet line, etc.

The voice generation unit 1102 generates an output voice signal to be output to the speaker 112 based on the learned result and the acquired information, and outputs the generated output voice signal to the speaker 112. As a result, the motion generation unit 110 outputs an audio signal from the speaker 112 to communicate with the user. The output voice signal is a voice signal assigned to the communication robot 1, a voice signal of a person who has made a TV call to a user via an Internet line, or the like.

The drive unit 1103 generates a drive signal for driving the actuator 113 based on the learned result and the acquired information, and drives the actuator 113 with the generated drive signal. As a result, the motion generation unit 110 controls the motion of the communication robot 1 to express emotions and the like, and communicates with the user.

The transmission information generation unit 1104 is, based on the learned result and the acquired information, for example, transmission information (audio signal, image) that the user wants to transmit to another user having a conversation on the network. ) Is generated, and the generated transmission information is transmitted from the transmission unit 114.

The display unit 111 is a liquid crystal image display device, an organic EL (Electroluminescence) image display device, or the like. The display unit 111 displays an output image output by the image generation unit 1101 of the social ability generation device 100.

The speaker 112 outputs an output voice signal output by the voice generation unit 1102 of the social ability generation device 100.

The actuator 113 drives the moving unit in response to a driving signal output by the driving unit 1103 of the social ability generation device 100.

The transmission unit 114 transmits the transmission information output by the transmission information generation unit 1104 of the social ability generation device 100 to the transmission destination via the network.

<Outline example of communication robot 1>
Next, an external example of the communication robot 1 will be described.
FIG. 3 is a diagram showing an outline example of the communication robot 1 according to the present embodiment. In the example of the front view g101 and the side view g102 of FIG. 3, the communication robot 1 includes three display units 111 (111a, 111b, 111c). Further, in the example of FIG. 3, the photographing unit 102a is attached to the upper part of the display unit 111a, and the photographing unit 102b is attached to the upper part of the display unit 111b. The display units 111a and 111b correspond to the human eye and present image information. The speaker 112 is attached in the vicinity of the display unit 111c that displays an image corresponding to the human mouth of the housing 120. The sound collecting unit 103 is attached to the housing 120.

Further, the communication robot 1 includes a boom 121. The boom 121 is movably attached to the housing 120 via the movable portion 131. A horizontal bar 122 is rotatably attached to the boom 121 via a movable portion 132.
Further, the display unit 111a is rotatably attached to the horizontal bar 122 via the movable portion 133, and the display unit 111b is rotatably attached via the movable portion 134.
The outer shape of the communication robot 1 shown in FIG. 3 is an example, and is not limited to this.

<Data stored in the first database>
Next, an example of data stored in the first database will be described.
The language model database stores the language model. The language model is a probability model that gives the probability that an arbitrary character string is a Japanese sentence or the like. The language model is, for example, an N-gram model, a hidden Markov model, a maximum entropy model, or the like.

The acoustic model database stores the sound source model. The sound source model is a model used for identifying the sound source of the collected acoustic signal.

The acoustic feature amount is a feature amount calculated after the pickled acoustic signal is converted into a signal in the frequency domain by performing a fast Fourier transform (Fast Fourier Transform). The acoustic feature amount is calculated, for example, a static Mel-Scale Log Spectram (MSLS), a delta MSLS, and one delta power at predetermined time intervals (for example, 10 ms). The MSLS is obtained by using a spectral feature as a feature for speech recognition and performing an inverse discrete cosine transform on the MFCC (Mel Frequency Cepstrum Deficient).

The dialogue corpus database stores the dialogue corpus. The dialogue corpus is a corpus used when the communication robot 1 and the user have a dialogue, and is, for example, a scenario according to the content of the dialogue.

The comparison image database stores, for example, images used for pattern matching. The image used for pattern matching includes, for example, an image of a user, an image of a user's family, an image of a user's pet, an image of a user's friend or acquaintance, and the like.

The image feature amount is, for example, a feature amount extracted from an image of a person or an object by a well-known image process.
The above-mentioned example is an example, and the first database 107 may store other data.

<Flow of cognition, learning, and social ability>
Next, the flow of cognition and learning performed by the social ability generation device 100 of the present embodiment will be described. FIG. 4 is a diagram showing the flow of cognition, learning, and social ability performed by the social ability generation device 100 of the present embodiment.

The recognition result 201 is an example of the result recognized by the cognitive unit 105. The recognition result 201 is, for example, interpersonal relationships, interpersonal relationships, and the like.

The multimodal learning and understanding 211 are examples of learning contents performed by the learning unit 108. The learning method 212 is machine learning or the like. The learning target 213 is a social component, a social model, psychology, humanities, and the like.

Social ability 221 is a social skill, such as empathy, individualization, adaptability, emotional ahodance, and the like.

<Data to be recognized>
Next, an example of data recognized by the cognitive unit 105 will be described.
FIG. 5 is a diagram showing an example of data recognized by the cognitive unit 105 according to the present embodiment. In this embodiment, personal data 301 and interpersonal relationship data 351 are recognized as shown in FIG.

Personal data is an action that occurs in one person, and is data acquired by the photographing unit 102 and the sound collecting unit 103, and data obtained by performing voice recognition processing, image recognition processing, or the like on the acquired data. .. Personal data includes, for example, voice data, semantic data that is the result of voice processing, voice volume, voice intonation, spoken words, facial expression data, gesture data, head posture data, face orientation data, and line-of-sight data. , Co-occurrence expression data, physiological information (body temperature, heart rate, pulse rate, etc.). The designer of the communication robot 1 may select, for example, what kind of data is used. In this case, for example, the designer of the communication robot 1 may set important features of personal data in communication for actual communication or demonstration of two people. Further, the cognitive unit 105 recognizes the user's emotion as personal data based on the information extracted from each of the acquired utterance and the image. In this case, the cognitive unit 105 recognizes based on, for example, the volume and intonation of the voice, the duration of utterance, the facial expression, and the like. Then, the communication robot 1 of the present embodiment controls the user's emotions to maintain good emotions and the relationship with the user to work to maintain a good relationship.

Here, an example of how to recognize the social background of the user will be described.
The cognitive unit 105 estimates the nationality, hometown, etc. of the user based on the acquired utterance, the image, and the data stored in the first database 107. The cognitive unit 105 extracts a life schedule such as a user's wake-up time, outing time, return time, and bedtime based on the acquired utterance and image and the data stored in the first database 107. The cognitive unit 105 is based on the acquired utterances, images, life schedule, and data stored in the first database 107, and is based on the user's gender, age, occupation, hobbies, career, taste, family structure, and religion. , Estimate the degree of attachment to the communication robot 1. Since the social background may change, the communication robot 1 updates the information on the social background of the user based on the conversation, the image, and the data stored in the first database 107. In order to enable emotional sharing, the social background and the degree of attachment to the communication robot 1 are not limited to the level at which input such as age, gender, and career can be input, and for example, emotional ups and downs according to the time of day. Recognize based on the loudness and intonation of the voice for the topic. In this way, the cognitive unit 105 learns about things that the user is confident and unaware of, based on daily conversation and facial expressions during conversation.

Interpersonal relationship data is data related to the relationship between a user and another person. By using interpersonal relationship data in this way, social data can be used. The interpersonal relationship data is, for example, the distance between people, whether or not the lines of sight of the people who are interacting with each other intersect, the intonation of the voice, the loudness of the voice, and the like. The distance between people depends on the interpersonal relationship, as will be described later. For example, in the case of a couple or a friend, the interpersonal relationship is L1, and the interpersonal relationship between businessmen is L2, which is larger than L1.

For example, the designer of the communication robot 1 may set important features of the interpersonal data in the communication for the actual communication or demonstration of the two people. Such personal data, interpersonal relationship data, and information on the social background of the user are stored in the first database 107 or the storage unit 106.

Further, the cognitive unit 105 collects and learns personal data for each user when there are a plurality of users, for example, the user and his / her family, and estimates the social background for each person. It should be noted that such a social background may be acquired, for example, via a network and a receiving unit 101, in which case the user inputs his / her social background or selects an item using, for example, a smartphone. May be good.

Here, an example of a method of recognizing interpersonal relationship data will be described.
The cognitive unit 105 estimates the distance (interval) between people with whom communication is being performed, based on the acquired utterances, images, and data stored in the first database 107. The cognitive unit 105 detects whether or not the lines of sight of the person communicating with each other intersect based on the acquired utterance, the image, and the data stored in the first database 107. The cognitive unit 105 sends and receives the utterance content, the volume of the voice, the inflection of the voice, the received e-mail, the transmitted e-mail, and the transmitted / received e-mail based on the acquired utterance and the data stored in the first database 107. Estimate friendships, business associates, and relatives based on the previous person.

In the initial state in which the cognitive unit 105 is used, for example, one is randomly selected from a combination of several social backgrounds and initial values of personal data stored in the first database 107 to communicate. May be started. Then, the cognitive unit 105 may reselect another combination when it is difficult to continue communication with the user due to the action generated by the randomly selected combination.

<Learning procedure>
In the present embodiment, the learning unit 108 learns using the personal data 301 recognized by the cognitive unit 105, the interpersonal relationship data 351 and the data stored in the second database 109.

Here, the social structure and social norms will be explained. In a space where people participate in social interactions, interpersonal relationships differ, for example, depending on the carry between people. For example, a relationship with a person having a distance of 0 to 50 cm is an intimate relationship, and a relationship with a person having a distance of 50 to 1 m is a personal relationship. A relationship with a distance of 1 to 4 m from a person is a social relationship, and a relationship with a distance of 4 m or more from a person is a public relationship. At the time of learning, such social norms are used as rewards (suggestive rewards) as to whether or not gestures and utterances conform to social norms.

Further, the interpersonal relationship may be set according to the environment to be used and the user by setting the feature amount of the reward at the time of learning. Specifically, even if you set multiple intimacy settings, such as setting a rule that does not talk much to people who are not good at robots and a rule that actively talks to people who like robots. good. Then, in the actual environment, the cognitive unit 105 recognizes which type the user is based on the result of processing the user's utterance and the image, and the learning unit 108 selects the rule. good.

Further, the human trainer may evaluate the behavior of the communication robot 1 and provide a reward (implicit reward) according to the social structure and norms that he / she knows.

<Data stored in the second database>
Next, an example of data stored in the second database will be described.
Social components are, for example, age, gender, occupation, relationships between multiple people (parents and children, couples, lovers, friends, acquaintances, colleagues, neighbors, teachers and students, etc.).

Social norms are rules and manners between individuals and multiple people, and are associated with utterances, gestures, etc. according to age, gender, occupation, and relationships between multiple people.

The data related to psychology are, for example, data of findings obtained by experiments and verifications so far (for example, attachment relationship between mother and infant, complex such as Oedipus complex, conditioned reflex, fetishism, etc.).

Data on the humanities are, for example, data on religious rules, customs, national character, regional character, characteristic acts, behaviors, and utterances in a country or region. For example, in the case of Japanese people, when consenting, the data is such that consent is expressed by nodding without saying it in words. In addition, the data on the humanities is, for example, data on what is emphasized and what is prioritized depending on the country or region.

FIG. 6 is a diagram showing an example of a system for performing deep reinforcement learning using raw data in a comparative example.
In the comparative example, when the raw data 902 of each of the captured image 901 and the picked-up acoustic signal 901 is used for learning, it is necessary to perform deep reinforcement learning 903. The system of this comparative example is difficult to realize. The reason is that it is necessary to collect sufficient teaching data for deep reinforcement learning, but it is difficult to collect. The reason it is difficult to collect is that the number of times the required features appear in the raw data is limited.

Therefore, in this embodiment, the raw data (voice signal, image) is not directly used for learning, but the feature amount is detected from the raw data and the feature amount is used for learning, so that reinforcement learning is required instead of deep reinforcement learning. ..

FIG. 7 is a diagram showing an example of an agent creation method used by the motion generation unit 110 according to the present embodiment.
The area indicated by reference numeral 300 is a diagram showing a flow from input to agent creation and output (agent).
The image taken by the photographing unit 102 and the information 310 collected by the sound collecting unit 103 are information about a person (user, a person related to the user, another person) and environmental information around the person. The raw data 302 acquired by the photographing unit 102 and the sound collecting unit 103 is input to the recognition unit 105.

The cognitive unit 105 has a plurality of information (voice volume, voice intonation, utterance content, spoken word, user's line of sight, user's head posture, user's face orientation) from the input raw data 302. , User's ecological information, distance between people, whether or not the line of sight of people intersects, etc.) is extracted and recognized. The cognitive unit 105 uses a plurality of extracted and recognized information to perform multimodal understanding using, for example, a neural network.
The cognitive unit 105 identifies an individual based on, for example, at least one of an audio signal and an image, and assigns identification information (ID) to the identified individual. The cognitive unit 105 recognizes the identified human action based on at least one of the audio signal and the image. The cognitive unit 105 performs, for example, well-known image processing and tracking processing on an image to recognize the line of sight of the identified person. For example, the recognition unit 105 performs voice recognition processing (sound source identification, sound source localization, sound source separation, utterance section detection, noise suppression, etc.) on the voice signal to recognize the voice. The cognitive unit 105 performs, for example, well-known image processing on an image to recognize the head posture of the identified person. For example, when two people are photographed in a photographed image, the cognitive unit 105 recognizes an interpersonal relationship based on the utterance content, the distance between the two persons in the photographed image, and the like. The cognitive unit 105 recognizes (estimates) the social distance between the communication robot 1 and the user according to the result of processing each of the captured image and the picked-up audio signal, for example.

The learning unit 108 performs reinforcement learning 304 instead of deep learning. Reinforcement learning involves learning to select the most relevant features (including social composition and social norms). In this case, a plurality of information used in multimodal understanding is used as a feature for input. The input of the learning unit 108 is, for example, the raw data itself, the name ID (identification information), the influence of the face, the recognized gesture, the keyword from the voice, and the like. The output of the learning unit 108 is the behavior of the communication robot. The output action may be any action that is desired to be defined according to the purpose, such as a voice response, a robot routine, an angle of orientation for the robot to rotate, and the like. In addition, in understanding Marutsu modal, a neural network or the like may be used for detection. In this case, different modality of the body may be used to detect human activity. Further, for example, the designer of the communication robot 1 may select in advance which feature to use. Further, in the present embodiment, by using implicit rewards and explicit rewards at the time of learning, it is possible to incorporate social models and social constructs. The result of reinforcement learning is the output, which is Agent 305. As described above, in the present embodiment, the agent used by the motion generation unit 110 is created.

The area indicated by reference numeral 350 is a diagram showing how to use the reward.
The implicit reward 362 is used to learn the implicit reaction. In this case, the raw data 302 includes the user's reaction, and the raw data 302 is used for the above-mentioned multimodal understanding 303. The learning unit 108 generates the implicit reaction system 372 by using the implicit reward 362 and the social model stored in the second database 109. The implicit reward may be obtained by reinforcement learning or may be given by a human. The implicit reaction system may also be a model acquired by learning.

For learning the explicit reaction, for example, a human trainer evaluates the behavior of the communication robot 1 and gives a reward 361 according to the social composition and social norms that he / she knows. The agent adopts the action that maximizes the reward for the input. As a result, the agent adopts behavior (utterance, gesture) that maximizes positive emotions toward the user.

The learning unit 108 uses this explicit reward 361 to generate an explicit reaction system 371. The explicit reaction system may be a model acquired by learning. The explicit reward may be given by the user by evaluating the behavior of the communication robot 1, and the communication robot 1 may, for example, give an explicit reward based on the user's utterances and behaviors (gestures, facial expressions, etc.). The reward may be estimated based on whether or not the user has taken the desired action.
The learning unit 108 outputs the agent 305 using these learning models during operation.

In this embodiment, the explicit reward, which is the reaction of the user, is prioritized over the implicit reward. The reason for this is that the user's reaction is more reliable in communication.

<Processing procedure example>
Next, an example of the processing procedure will be described. FIG. 8 is a flowchart showing a procedure example of the social capacity generation process according to the present embodiment.

(Step S11) The recognition unit 105 acquires an image captured by the photographing unit 102 and an acoustic signal collected by the sound collecting unit 103.

(Step S12) The recognition unit 105 recognizes, detects, or extracts characteristic information related to voice from an acoustic signal, and recognizes, detects, or extracts characteristic information related to a person from an image. The characteristic information related to voice is at least one of a voice signal, voice loudness information, voice intonation information, and the meaning of utterance. The characteristic information about a person is at least one of a person's facial expression information, a person's gesture information, a person's head posture information, a person's face orientation information, and a person's line-of-sight information.

(Step S13) The cognitive unit 105 recognizes the action that occurs between the communication robot 1 and a person, or the action that occurs between a plurality of people, based on the acquired information and the data stored in the first database 107. ..

(Step S14) The learning unit 108 learns human emotional interactions using the cognitive results output by the cognitive unit 105 and the data stored in the second database 109.

(Step S15) The motion generation unit 110 generates actions (utterances, gestures, images) for the user based on the learned result and the acquired information.

<Communication between Communication Robot 1 and people>
Next, an example of communication between the communication robot 1 and a person will be described.
The timing of the utterance of the communication robot 1 is set as an initial value, for example, when returning home, when waking up, or the like. Then, by repeating the communication, the timing at which the communication robot 1 starts the utterance may be learned.
Alternatively, the communication robot 1 may start the utterance according to the utterance or action of the user. In this case, the communication robot 1 does not start the conversation in response to the command adopted by the smart speaker or the like, but for example, the user is talking to the communication robot 1, the user's facial expression or face (face). The conversation is started by recognizing that the face is looking at the communication robot 1) or the light in the room is lit. Further, the communication robot 1 recognizes, for example, the utterance content of the user, the facial expression of the user, and the like, and ends the conversation. Then, the communication robot 1 may switch between the start timing and the end timing of the utterance according to the social background (including age and gender) of the user, the attachment to the communication robot 1, and the like.

An example of communication between the communication robot 1 and the user will be described with reference to FIGS. 9 to 12.
FIG. 9 is a diagram showing an example of communication between the communication robot 1 and the user according to the present embodiment. The example of FIG. 9 shows a state in which the communication robot 1 talks to the user and has a conversation with the user when the user returns home. During communication, the communication robot 1 acts (speech, gesture, image presentation) so as to maintain a good relationship with a person. The communication robot 1 drives a boom horizontal bar or the like by a drive unit 1103 and an actuator 113 to control gestures and gestures.

10 and 11 are diagrams showing an example of an image displayed on the display unit during communication between the communication robot 1 and the user according to the present embodiment.
The example of FIG. 10 is an example in which a heart mark indicating attachment is displayed on the display units 111a and 111b, and an image corresponding to an image of a laughing mouth is displayed on the display unit 111c.
The example of FIG. 11 is an example in which an illustration imagining the place of the scheduled date and the weather forecast of the day are presented, and the weather forecast is spoken and read aloud by the speaker 112. In the example of FIG. 11, an example of presenting an illustration imagining a place is shown, but the communication robot 1 acquires an image (photograph or moving image) of the planned place via the receiving unit 101, and the acquired image. May be presented.
In this way, the display unit 111 causes the person to behave in a manner that maximizes positive emotions, thereby displaying an image that maintains a good relationship with the person. Thereby, according to the present embodiment, it is possible to make the person behave in a way that maximizes positive emotions, and it is possible to maintain a good relationship with the person.

FIG. 12 is a diagram showing an example of communication by the communication robot 1 at the time of communication with a friend of a user according to the present embodiment. In the example of FIG. 12, the communication robot 1 sends a message to a friend of the user by a dialogue with the user. The terminal 200 owned by the user's friend receives this message and displays it on the display unit (g301). Then, the terminal 200 transmits the reply g302 to the user to the communication robot 1 according to the operation result of the user's friend. The communication robot 1 presents a reply from a user's friend in an action (utterance, gesture, image presentation) based on the information received from the terminal 200.

In the above-mentioned example, the communication robot 1 has described an example of communicating with a user by using voice, motion (gesture), and an image, but the present invention is not limited to this. In order to communicate with the user, it is preferable to use two or more output means used by the communication robot 1, and it is sufficient that two or more of voice, motion (gesture), and image are used. Alternatively, the output means may be, for example, text and action, text and voice, or the like. Further, it is preferable that the communication robot 1 has a plurality of output means so as not to make the user bored.

Further, the input from the user to the communication robot 1 is not limited to the above-mentioned voice and image. It suffices if the user's behavior can be acquired, and other information may also be acquired. The other information is, for example, contact information such as a user touching or hitting the communication robot 1.

As described above, in the present embodiment, the communication robot 1 recognizes the action that occurs between the communication robot 1 and a person, or the action that occurs between a plurality of people. Then, in this embodiment, the emotional interaction of human beings is learned from the perceived contents from machine learning and psychology, social customs, humanities, etc., and the social ability of the robot is generated from the learned contents. did. Further, in the present embodiment, in addition to the implicit reward, the explicit reward is used in the learning.

Thereby, according to the present embodiment, it is possible to generate social skills of the robot based on emotional interaction with humans. According to this embodiment, it is possible to provide a social robot and an agent of a house that fosters relationships with people. According to this embodiment, it is possible to generate sympathetic communication and interaction between machines and humans. According to this embodiment, it is possible to provide a concept of a "machine" friend such as a pet friend or a robot friend. According to this embodiment, it is possible to provide a machine that is socially intelligent and can navigate social scenarios. Thereby, according to the present embodiment, it is possible to form an emotional connection between the robot and the human.
Further, according to the present embodiment, features are extracted from each of the picked-up acoustic signal and the captured image, and reinforcement learning is performed using the extracted features. Therefore, deep machine learning using raw data is performed. It is possible to perform learning without using a lot of teaching data as in.

<Modification example>
In the embodiment, the communication robot 1 will be described as an example of a device for communicating, but the present embodiment can be applied to other devices such as an in-vehicle navigation device, a smartphone, a tablet terminal, and the like. For example, when applied to a smartphone, a still image of the communication robot 1 as shown in FIG. 3, for example, may be displayed on the display unit of the smartphone so that communication by voice is mainly performed. Alternatively, the gesture of the communication robot 1 may be displayed by animation on the display unit of the smartphone.

FIG. 13 is a diagram showing an example in which the communication robot of the present embodiment is applied to the car navigation system 300 in the vehicle. The car navigation system 300 may be a smartphone, a tablet terminal, or the like.
The car navigation system 300 causes the display unit to display an image of the communication robot. In this case, the communication robot operates as an agent. Then, the agent realizes the functions of the communication robot 1 (excluding the drive unit, the actuator, etc.) by using the photographing unit, the sound collecting unit, the display unit, the speaker, and the like provided in the car navigation system 300.

When applied to the car navigation system 300, the communication robot displayed on the display unit may be a still image or an animation. In this case, the agent responds by at least a voice dialogue. Even in this case, during communication, the agent acts (speech, gesture, image presentation) to maintain a good relationship with the person.

FIG. 14 is a diagram showing an example of connection with various devices in the home when applied to the car navigation system according to the present embodiment. In addition, also in FIG. 14, the car navigation system 300 may be a smartphone, a tablet terminal, or the like. It is assumed that the car navigation system 300 includes a communication unit (reception unit and transmission unit), and each device at home is connected via a network. The agent applied to the car navigation system 300 responds to communication with the user, for example, opening / closing the shutter of the parking lot 401, the operation start instruction 402 of the rice cooker, the operation start instruction of the air conditioner, and the setting instruction 403 such as temperature. , The lighting start instruction 404 of the electric lamp of the room or the like, the operation start instruction 405 of the automatic lawn mower, and the like are given. In addition, the agent does not simply give an operation instruction, but sets the user's favorite temperature setting and the user's favorite room brightness setting according to the utterance with the user. Communication You may learn in advance, and based on the results of learning these, you may give the optimum timing and setting instructions so that these tasks are completed when you return home.

A program for realizing all or part of the functions of the social ability generation device 100 in the present invention is recorded on a computer-readable recording medium, and the program recorded on the recording medium is read into the computer system. , All or part of the processing performed by the social capacity generation device 100 may be performed by executing. The term "computer system" as used herein includes hardware such as an OS and peripheral devices. Further, the "computer system" shall also include a WWW system provided with a homepage providing environment (or display environment). Further, the "computer-readable recording medium" refers to a portable medium such as a flexible disk, a magneto-optical disk, a ROM, or a CD-ROM, and a storage device such as a hard disk built in a computer system. Furthermore, a "computer-readable recording medium" is a volatile memory (RAM) inside a computer system that serves as a server or client when a program is transmitted via a network such as the Internet or a communication line such as a telephone line. In addition, it shall include those that hold the program for a certain period of time.

Further, the program may be transmitted from a computer system in which this program is stored in a storage device or the like to another computer system via a transmission medium or by a transmission wave in the transmission medium. Here, the "transmission medium" for transmitting a program refers to a medium having a function of transmitting information, such as a network (communication network) such as the Internet or a communication line (communication line) such as a telephone line. Further, the above program may be for realizing a part of the above-mentioned functions. Further, a so-called difference file (difference program) may be used, which can realize the above-mentioned function in combination with a program already recorded in the computer system.

Although the embodiments for carrying out the present invention have been described above using the embodiments, the present invention is not limited to these embodiments, and various modifications and substitutions are made without departing from the gist of the present invention. Can be added.

1 ... Communication robot, 101 ... Receiver unit, 102 ... Imaging unit, 103 ... Sound pickup unit, 104 ... Sensor, 100 ... Social ability generator, 106 ... Storage unit, 107 ... First database, 109 ... Second database, 111 ... Display unit, 112 ... Speaker, 113 ... Actuator, 114 ... Transmission unit, 105 ... Cognitive unit, 108 ... Learning unit, 110 ... Motion generation unit, 1101 ... Image generation unit, 1102 ... Sound generation unit, 1103 ... Drive unit 1,104 ... Transmission information generation unit

Claims (7)

Acquires human information about a person, extracts characteristic information about a person from the acquired person information, recognizes the action that occurs between the communication device that communicates and the person, and recognizes the action that occurs between people. Cognitive means and
A learning means for learning the emotional interaction of a person by multimodal using the extracted characteristic information about the person, and
An action generation means that generates an action based on the learned emotional interaction information of the person,
A social capacity generator equipped with. The learning means learns using an implicit reward and an explicit reward.
The implicit reward is a reward learned by multimodal using the characteristic information about the person.
The explicit reward is a reward based on the result of evaluating the behavior of the communication device generated by the motion generating means with respect to the person.
The social ability generation device according to claim 1. A sound collecting part that collects acoustic signals and
Equipped with a shooting unit that shoots images including users,
The cognitive means performs voice recognition processing on the picked-up acoustic signal to extract characteristic information related to voice, and performs image processing on the captured image to perform characteristic information on human behavior included in the image. Extracted,
The characteristic information regarding the person includes the characteristic information regarding the voice and the characteristic information regarding the human behavior.
The characteristic information related to voice is at least one of a voice signal, voice loudness information, voice intonation information, and the meaning of utterance.
The characteristic information related to human behavior is at least one of a person's facial expression information, a person's gesture information, a person's head posture information, a person's face orientation information, a person's line of sight information, and a distance between people. One,
The social ability generation device according to claim 1 or 2. The learning means is
Learn using social norms, social components, psychological and humanistic findings,
The social ability generation device according to any one of claims 1 to 3. The cognitive means acquires human information about a person, extracts characteristic information about the person from the acquired human information, recognizes the action that occurs between the communication device that communicates and the person, and occurs between the person. Recognize the work,
The learning means multimodally learns the emotional interaction of a person using the extracted characteristic information about the person.
The motion generation means generates an action based on the learned emotional interaction information of the person.
How to generate social ability. Acquires human information about a person, extracts characteristic information about a person from the acquired person information, recognizes the action that occurs between the communication device that communicates and the person, and recognizes the action that occurs between people. Cognitive means and
A learning means for learning the emotional interaction of a person by multimodal using the extracted characteristic information about the person, and
An action generation means that generates an action based on the learned emotional interaction information of the person,
Communication robot equipped with. Equipped with a display
The motion generating means generates an image that maintains a good relationship with a person by causing the person to behave in a way that maximizes positive emotions, and displays the generated image on the display unit.
The communication robot according to claim 6.

Images