JP2010094799A - Humanoid robot - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a humanoid robot for transmitting information to a specific person by using similar face and voice, which is adapted to produce love or the like that a specific speaker intends to transmit by expressing the speaker's personality in the action taken during voice output. <P>SOLUTION: Phoneme data such as fifty sounds necessary for recording and speaking the specific person's voice, and action data as the specific person's habits are previously registered in a storage (3) mounted inside a robot body (1) having a face similar to the specific person. During speak, the voice is synthesized by a main CPU unit (211) within a controller (2) and then output through a speaker (26). At the same time actuators (9) to (25) are activated to cause quadruped actions. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

The present invention relates to a humanoid robot that performs communication using gestures, facial expressions, and the like based on images, sounds, and input information from a plurality of sensors.

Various robots are widely used in consumer fields such as toys and games, and industrial fields such as manufacturing and distribution, and these robots are very similar in shape to humans and communicate with users. So-called humanoid robots have also been developed.
Some conventional humanoid robots are intended for specific individuals with respect to any one of the face, utterance mechanism, and limb movement, but generally the face, utterance mechanism, and limb movement are specified uniformly. Yes.

For example, in the invention disclosed in Patent Document 1, information related to emotion is extracted by analyzing the content of a sentence, and an e-mail that performs utterance or gesture (gesture / hand gesture) corresponding to the extracted emotion information is read out A robot is shown.

JP 2006-142407 A

As in the conventional example described above, the content of the text to be transmitted by the robot is analyzed, information related to emotion is extracted from the text, and a gesture corresponding to the extracted emotion information is expressed together with the voice. It is possible to express emotions by movement, that is, by language and non-language.
However, the robot gestures made here are only gestures corresponding to specific emotion expressions, and the movements express gestures and emotions stored in the robot in a gesture. is there.

In general, the gestures that occur when a person utters a voice in conversation or the like are various and have various habits. It is preferable that a gesture when a specific person utters a voice is performed using a gesture including the trap of the specific person. If such a gesture is made with a face or voice similar to a specific person, only the information can be transmitted. It can express emotions such as love.

The present invention is a humanoid robot that conveys information to a specific individual with a similar face and voice, and the speaker intends to convey the personality of the specific speaker by expressing the personality of the specific speaker in the movement of the voice. The purpose is to provide a humanoid robot that can produce affection.

Another object of the present invention is to provide a humanoid robot that can easily change the voice and habit of a specific individual and can realize the habit of the specific individual.

The humanoid robot of the present invention made to solve the above-described problems is a movement of each part of the human body corresponding to each of the parts constituting the outer shape of the humanoid robot having a humanoid outer shape that is a face of a specific individual. A humanoid robot having a group of actuators that cause the same movement and a speaker that emits a voice, and a model information database that stores information on a specific person as a model of the robot, and a phoneme of the voice of the specific person Accumulated speech synthesis phoneme database, 癖 motion database that accumulates movements of specific individuals as chronological data, partner information database that accumulates face image data and feature information of conversation partner, and robot action history Action history database and knowledge information database that accumulates information obtained from dialogue and information on the Internet Voice synthesizing means that synthesizes and utters a phoneme of a specific person based on the accumulated data of the speech synthesis phoneme database, operation means that produces a habit of the specific person based on the accumulated data of the haze motion database, and the partner An emotion recognition means for recognizing emotions to a conversation partner based on data stored in the information database, the model information database, and the action history database; a speech recognition means for recognizing the content of voice data of the conversation partner; and the speech recognition means A learning means for accumulating in the knowledge information database as information on the conversation with the conversation partner and the information on the Internet according to the recognition result of the recognition, the recognition result in the emotion recognition means, and the conversation partner according to the accumulated knowledge in the learning means Producing emotions, and the voice synthesizing means and the operating means according to the production results Characterized in that a control emotion presentation control means.

The humanoid robot interacts with the conversation partner from the model information database in which the scenario of a specific individual is described, and recognizes the emotion of the robot with respect to the conversation partner from the conversation. In order to produce a feeling with feelings for that emotion, control is performed to generate a voice synthesized by voice synthesis from the speaker, and the movement of each part of the human body stored in the heel data corresponding to the voice is displayed. The humanoid robot is operated so that the voice and movement are synchronized.

According to the present invention, when a robot with a similar face to a specific individual produces a voice that is similar to the voice, the specific individual's habit is produced by an action, so that emotion and affection can be produced.

BEST MODE FOR CARRYING OUT THE INVENTION

The embodiment of the present invention will be described with reference to FIGS.
The present invention realizes a humanoid robot that resembles a specific individual that has not existed in the past by controlling the voice and habit of the specific individual with a robot similar to the specific individual. If the outline is explained based on FIGS. 1 and 2, a voice of a specific individual is recorded in advance in a storage (220) mounted on the control unit (2) included in the robot body (1) similar to the specific individual. Register phoneme data such as the Japanese syllabary required for utterance and motion data that is a habit of a specific individual, and synthesize the voice by the main CPU unit (211) in the control unit (2) at the time of utterance. The actuator (9) to (25) are operated on the extremities at the same time.

FIG. 1 is a configuration diagram of a robot body (1) showing an outline of the present invention.
The robot body (1), which is a face of a specific individual based on a photograph or the like, is composed of a speaker (26) that is an utterance mechanism mainly based on a recording sound source, a microphone (27, 28) that is a listening mechanism, a camera ( 11), (12), actuators (9) to (25) for limb movement, network connection mechanism (6), sensor touch sensors (29), (30), temperature sensor (31), and these parts It is comprised from the control part (2) to perform.

FIG. 2 is a configuration diagram of the control unit (2) included in the robot body (1).
The control unit (2) includes a main CPU unit (210) having an artificial intelligence of a humanoid robot, a sub CPU unit (230) that controls actuators and sensors, and a storage unit (220) that stores a data file such as a database. The main CPU unit (210) and the sub CPU unit (230) always exchange information with each other in the communication module (213) in the main CPU unit (210) and the communication module (232) in the sub CPU (230). Communicating.
The main CPU unit (210) includes a CPU (211) that executes application software that is artificial intelligence, a memory (212), a camera (7) (8), and a sensor input / output module (214) that controls the wireless LAN (6). And a communication module (213) for obtaining information from the sub CPU.
The sub CPU unit (230) includes a CPU (231) for controlling each actuator and managing each sensor, an actuator control module (237) for controlling the actuators (9) to (25), and a touch for detecting a handshake. A sensor input / output module (236) for controlling the sensors (29) and (30), a temperature sensor (31) for measuring the ambient temperature, a human sensor (32) for detecting a person, and the like to obtain position information The GPS (233), an acceleration sensor (234) for detecting vibrations such as shaking of the humanoid robot, and a temperature sensor (235) for measuring the temperature in the humanoid robot.

FIG. 3 is a configuration diagram showing an outline of software.
This software operates on the operation system (S101) of the CPU on the control unit (2).
On the operation system (S101), an application program (S102) that plays the role of artificial intelligence, a camera device program (S117) that controls the camera, a microphone device program (S118) that controls the microphone, and a speaker device program that controls the speaker (S119), which is composed of an action history database (S103) in which specific personal data, action history, and the like are accumulated.
The application program (S102) includes an image recognition program (S111) for recognizing the speaker's face and object, a voice recognition program (S112) for identifying voice and a conversation partner, and a recognition partner's emotion. Emotion recognition program (S113), learning control program (S114) for accumulating in the database as information on dialogue with the conversation partner and information on the Internet, speech synthesis program (S115) for synthesizing speech, and for producing emotion Consists of emotion production program.
The database (S103) includes a knowledge information database (S121) for accumulating information obtained from dialogues and information on the Internet, a speech synthesis phoneme database (S122) for synthesizing a specific person's similar voice, and a specific person's habits Saddle motion database (S123) for storing data, model information database (S124) describing a scenario for producing a specific individual, partner information database for storing face information and features of a partner who interacts with the robot ( S125), configured from an action history database (S126) for accumulating acquired data of each sensor of this system with respect to time.

According to the present invention, in order to produce a specific individual, the robot responds to emotions to the other speaker with whom the robot interacts based on inputs from a plurality of sensors.
The control method changes the pattern of the corresponding action with respect to the emotion level for the speaker registered in the partner information database.
The partner information database has emotional levels of “like”, “normal”, and “dislike” for the other speaker. The corresponding action is, for example, speaking to a favorite speaker with a gentle wording or acting with caress. Make people dislike it and ignore it.
Also, the emotion level is updated. For example, even if the speaker is a favorite speaker, when the number of talks is reduced, it changes from “love” to “like”.
The wording in the response action is handled by changing the intonation. The corresponding motion is selected from a pre-registered eyelid motion database and executed.

The flow for producing a specific individual will be described with reference to the flowchart of FIG. (S401) At that time, the robot performs face recognition processing using an image and identifies a speaker. (S402) In parallel, speaker identification is performed by voice (S403). When (S402) and (S403) match with a certain degree of accuracy, the name of the speaker is called, confirmed, and identified (S404). If the speaker is liked by comparing with the partner information database of the identified other speaker (S405), the favorite person mode (S407) is processed, and if the speaker is disliked (S406), the disliked person The mode (S408) processing is performed, and if none of them is performed, the normal human mode (S409) processing is performed. When the processing of (S407), (S408), or (S409) ends, the routine returns to the daily conversation (S401).

For example, the process of the favorite person mode (S407) will be described with reference to the flowchart of FIG. 5. When the robot changes to the favorite person mode, the robot performs a daily conversation for a favorite person while performing speech synthesis using a charming phoneme. (S501) At that time, the robot measures the emotion of the speaker by voice (S502), throws a topic for identifying the emotion, and identifies the emotion of the speaker. (S503) When the speaker is fine (S504), the process in the energetic mode (S506) is performed. When the speaker is depressed (S505), the process in the comfort mode (S507) is performed. If it is neither, processing in the normal mode (S508) is performed.

The speech synthesis for the favorite person mode, the disliked person mode, and the normal person mode can be performed by selecting different intonations from the speech synthesis phoneme database.

The conversation contents in the energetic mode, the comfort mode, and the normal mode are previously described and registered in the model information database (S124) as model information of a specific individual as a scenario in accordance with the conversation flow.

An example of a scenario registered in the model information database (S124) is shown in FIG. In the figure, a model information database (S124) serving as a robot model includes a name (S601), a date of birth (S602), a hobby (S603), a special skill (S604), a scenario for expressing an effect (S605), a scenario A speech recognition phrase (S606), a voice for production (S608), and a movement for production (S609) are described in XML format.
Since it is targeted at a specific individual, the XML format is highly flexible when producing a specific personal habit or the like.

The speech recognition word / phrase (S606) serves as a speech recognition dictionary.
The number of speech recognition phrases described in the model information database (S124) is the number of phrases that can be recognized by the robot.

On the other hand, in the speech synthesis phoneme database (S122), a basic sentence prepared in advance is read out and registered in advance.

In addition, time series data of each joint that becomes the user's heel is registered in advance in the heel motion database (S123).

As shown in FIG. 7, the partner information database (S125) includes an ID (S701), a conversation partner name (S702), a date of birth (S703), a hobby (S704), a special skill (S705), and a phoneme for speech recognition (S706). ), Eigenvalues for face recognition (S707), emotion level for the conversation partner (S708), and the like. The emotion level (S708) is obtained from the behavior history database (S126), the emotion value determined for the speaker by the robot when talking with the robot by the conversation frequency or conversation sequence, and the sub CPU in the control unit (2). It is updated by determining whether or not it is gently held from the data acquired from the acceleration sensor (234) in the unit (230), whether the head is not hit, etc., and the emotion value is determined.

As shown in the figure, the action history database (S126) has an ID (S801), a date (S802) at the time of history registration, a time (S803), a place (S804), a conversation partner (S805), and a transition at that time. Scenario (S806) which is a scenario, face recognition eigenvalue newly registered when the conversation partner is unknown (S807), name of unknown person or new word when a new word is remembered (S808), etc. It is composed of fields. The new word (S808) is registered by the learning control program (S114) when the conversation with the conversation partner or data on the Internet is acquired. The location (S804) is specified and registered by the location information and the conversation sequence acquired from the GPS (233) in the sub CPU unit (230) in the control unit (2).

The knowledge information database (S121) includes an ID (S901), an information acquisition date (S902), a time (S903), a place (S904), a conversation partner (S905), a category of the transitioned scenario (S906), a new It consists of fields such as the name of a person who has learned and a new word (S907) which is a word. The knowledge information database is registered by the learning control program (S114) when the action history database (S126) or data on the Internet is acquired.

It is explanatory drawing which shows an example of the system of this invention. It is a block diagram which shows the electrical structure of the control part, actuator, and sensor which are shown in FIG. It is a block diagram which shows the software structure of a control part. It is a flowchart of the software which expresses an emotion. It is a flowchart of the software in a favorite person mode. It is a model information database. It is a partner information database. It is an action history database. It is a knowledge information database.

Explanation of symbols

1 Robot 2 Control unit 210 Main CPU unit 220 Storage 230 Sub CPU unit 6 Wireless LAN
7, 8 Camera 9-25 Actuator 26 Speaker 27, 28 Microphone 29, 30 Touch sensor 31 Temperature sensor 32 Human sensor

Claims (3)

A robot whose face shape resembles a specific individual having a face part, a speech mechanism, a sounding mechanism, a camera mechanism, a limb movement mechanism, a network connection mechanism, an information storage mechanism, and a control mechanism that supervises these parts,
A humanoid robot provided with a corresponding action executing means for causing the robot to execute a corresponding action according to a stimulus to the robot recognized based on inputs from a plurality of sensors provided in each part of the limb of the robot. The facial part is a face of a specific individual, the utterance mechanism uses a sound source approximated to the voice of the specific individual, and the corresponding action is an action approximated to the eye of a specific individual. The humanoid robot according to claim 1. A model information database that stores information on specific individuals;
A speech synthesis phoneme database that stores phonemes of specific individuals;
癖 Motion database that accumulates movements of specific individuals as chronological data,
A partner information database that stores face image data and feature information of the conversation partner,
An action history database for accumulating robot action history;
It has a knowledge information database that accumulates information obtained from dialogue and information on the Internet,
A speech synthesizer that synthesizes and utters a specific individual phoneme based on the accumulated data of the speech synthesis phoneme database;
An operation means for producing a specific individual's wrinkle based on the accumulated data of the wrinkle motion database;
Emotion recognition means for recognizing emotions for a conversation partner based on the stored data of the partner information database, the model information database, and the action history database;
Voice recognition means for recognizing the content of the voice data of the conversation partner,
Learning means for accumulating in the knowledge information database as knowledge information on the Internet and conversations with the conversation partner according to the recognition result of the voice recognition means;
An emotion effect control means for producing an emotion for a conversation partner according to the recognition result in the emotion recognition means, the accumulated knowledge in the learning means, and controlling the voice synthesis means and the operation means according to the effect result;
The humanoid robot according to claim 1, further comprising:

Images