A method for determining whether the person performing an action is a robot without waking up the robot

Through visual and sound source angle information, we can determine whether the action performer is a robot, which solves the problem that the awakening robot cannot distinguish the user's communication object, and realizes the intelligent communication response between the robot and the user.

CN113688711BActive Publication Date: 2024-11-26ZHEJIANG LAB
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202110946066.X
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2021-08-17
Publication Date
2024-11-26
Estimated Expiration
2041-08-17

AI Technical Summary

Technical Problem

The unwake-free robot cannot determine whether the user communicates with the robot during voice input, resulting in the inability to give a response.

Method used

By combining visual and sound source angle information, we can determine whether the user's action performer is a robot after voice input. The specific steps include collecting visual information and voice source information, building a coordinate system, determining the voice speaker, and judging the face orientation of the voice speaker's face. , and combine semantic understanding information to identify micro-actions to determine whether the executor of the voice speaker is a robot.

Benefits of technology

Effectively determine whether the action performer is a robot, thereby responding and replying, improving the intelligence level of communication between the awakenless robot and the user.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN113688711B_ABST
    Figure CN113688711B_ABST
Patent Text Reader

Abstract

The present invention relates to the field of artificial intelligence robots, and in particular to a method for a non-wake-up robot to determine whether an action performer is a robot, comprising the following steps: step 1, the non-wake-up robot collects visual information and voice sound source information through a visual module and a voice module, the visual information includes character information and micro-motion information, and the voice sound source information includes sound source positioning information and semantic understanding information; step 2, judging the voice speaker according to the sound source positioning information; step 3, after determining the voice speaker, judging the face orientation of the voice speaker; step 4, extracting the action performer and intention according to the semantic understanding source information, and at the same time combining the speaker's face orientation, identifying micro-motion information, and judging whether the action performer in the speaker's voice is a robot. Based on vision, sound source positioning angle and semantic understanding information, the present invention can effectively judge whether the action performer is a robot, so as to respond and reply.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence robots, and in particular to a method for a non-awakening robot to determine whether a person performing an action is a robot. Background Art

[0002] Service robots are mainly used to communicate with users and then realize specific functions. In general, users can talk to robots only after they wake up the robots by voice inputting wake-up words, such as "hello XX" and other words that are less frequently used in real life. However, each voice input of wake-up words is not in line with real life and is cumbersome to use. Therefore, robots without wake-up are more intelligent than robots with wake-up. However, robots without wake-up cannot judge whether users are communicating with robots or people around them during voice conversations, so they cannot give a responsive reply. Summary of the invention

[0003] In order to solve the above technical problems existing in the prior art, the present invention proposes a method for a non-wake-up robot to determine whether the action performer is a robot. By combining information such as vision and sound source angle, the method determines whether the action performer is a robot after the user's voice input dialogue. The specific technical solution is as follows:

[0004] A method for a non-awakening robot to determine whether a person performing an action is a robot comprises the following steps:

[0005] Step 1: The wake-up-free robot collects image recognition character information and user voice input information through the visual module and the voice module, namely, visual information and voice sound source information. The visual information includes character information and micro-motion information, and the voice sound source information includes sound source positioning information and semantic understanding information.

[0006] Step 2, determining the speaker of the voice according to the sound source localization information;

[0007] Step 3, after determining the voice speaker, determine the face direction of the voice speaker;

[0008] Step 4: Based on the semantic understanding of the source information, the action performer and intention are extracted. At the same time, the micro-movement information is identified based on the speaker's facial orientation to determine whether the action performer in the speaker's voice is a robot.

[0009] Furthermore, the step 1 specifically includes:

[0010] Step 1.1, construct the coordinate system: the height of the robot is heightR, the ground where the robot is located is taken as the origin, the horizontal direction of the robot has x-axis and y-axis, the right direction of the robot is the positive direction of the x-axis, the front of the robot is the positive direction of the y-axis, the vertical direction of the robot is the z-axis, and the top is the positive direction of the z-axis;

[0011] Step 1.2, obtaining the horizontal position status information and height of the person through the robot vision module, and constructing a set S representing the person information;

[0012] Step 1.3, construct the robot exclusive intention set Y = {exclusive intention Figure 1 , exclusive Figure 2 , exclusive Figure 3 ,…, exclusive intent N};

[0013] Step 1.4, construct a micro-action set X = {micro-action 1, micro-action 2, micro-action 3, ..., micro-action M} that prompts the other party to talk to the other party, and the robot vision module recognizes the micro-action during the user's voice input process and clarifies the person to be executed by the micro-action;

[0014] Step 1.5: The robot speech module performs sound source localization and speech semantic understanding based on the collected speech source information.

[0015] Furthermore, the step 1.2 is specifically as follows: the robot collects the height height of the surrounding people through its visual module, the horizontal position state information x, y, which respectively represent the horizontal position coordinates of the person on the positive x-axis and the positive y-axis, the face orientation vector (faceX, faceY, faceZ), and the information id, where id is the unique identifier of the surrounding people, and for the same person, the id remains unchanged and is a positive number; then the person information set S = {id, x, y, height, faceX, faceY, faceZ} is constructed; the time frequency of the robot collecting visual information is f, that is, the amount of information generated within one second, the duration from the start of voice input to the end of voice input is t seconds, and the total amount of information generated within t seconds is the set E = {S1, S2, S3, ..., S tf}, there are tf person information sets S in set E.

[0016] Furthermore, the sound source localization is to obtain the horizontal angle θ of the sound source input when the speech starts, 0<=θ<2π, if the person is to the right of the robot, θ=0; if the person is to the right front of the robot, If the person is in front of the robot, If the person is in front of the robot's left side, If the person is to the left of the robot, θ = π; if the person is to the left rear of the robot, If the character is directly behind the robot, If the character is behind the robot to the right,

[0017] The speech semantic understanding is to extract the relevant action performers and their intentions.

[0018] Furthermore, the step 2 specifically includes:

[0019] Step 2.1, when the voice module detects that someone starts to speak, the voice module sends the horizontal angle θ of the sound source input when the person starts to speak;

[0020] Step 2.2, construct a unit vector that starts from the origin, points to the positive direction of the x-axis, and then rotates horizontally counterclockwise by an angle of θ Specifically:

[0021] When θ=0,

[0022] when hour,

[0023] when hour,

[0024] when hour,

[0025] When θ=π,

[0026] when hour,

[0027] when hour,

[0028] when hour,

[0029] Step 2.3, extract the information set S = {id, x, y, height, faceX, faceY, faceZ} of each person around when the speech starts, and construct a vector with the starting point as the origin, that is, the position of the robot as the origin and the position of the surrounding people as the end point Calculate the corresponding and The angle between:

[0030]

[0031] Calculate the angle α corresponding to each character in the surrounding area, and select the minimum value of α, that is, the character id corresponding to min(α) is the id of the character who speaks, and record the id at the same time.

[0032] Furthermore, the step 3 specifically includes:

[0033] Step 3.1, extract the set E of all characters around the robot from the beginning of speaking to the end of speaking. Suppose there are P characters around the robot, and let the set corresponding to the speaking character be E1, and the sets corresponding to the other characters are E2, E3, ..., E P , and establish a one-to-one mapping table Q between the subscripts of set E and the corresponding character IDs, that is, the subscripts of different sets E correspond to different IDs, and different IDs correspond to the subscripts of different sets E; suppose set E i There are N elements in total, with the same collection frequency. Each person in the surrounding area has N elements in the set E. The first element is S i1 , the second element is S i2 , ..., the kth element is S ik , ..., the Nth element is S iN , let S ik The x-axis coordinate of the character's position is x ik , the y-axis coordinate is y ik , the character's height is height ik , the face orientation vector is Then the face direction vector of the speaker is

[0034] Step 3.2, let the set E i The corresponding character's position is taken as the starting point, with set E j The vector corresponding to the location of the surrounding characters as the end point is Then take the speaker's location as the starting point and set E j The vector corresponding to the location of the surrounding characters as the end point is

[0035] Step 3.3, let the set E i The vector starting from the position of the corresponding character and ending at the position of the robot is The vector starting from the position of the speaker and ending at the position of the robot is

[0036] Step 3.4, calculate the k-th moment vector With vector The angle a between jk And vector With vector The angle b between k , k=1,2,…,N, the expression is:

[0037]

[0038]

[0039]

[0040]

[0041] Get the kth moment a jk The minimum value of j = 2, 3, ..., P, and a jk The minimum value of b k Compare the size, that is, min(b k ,min(a jk )); then construct a set R, which has N elements. The first element is r1, the second element is r2, ..., and the kth element is r k , ..., the Nth element is r N , if min(b k ,min(a jk ))=min(a jk ), then according to the mapping table Q, r k is min(a jk ) corresponds to the character id corresponding to j; if min(b k ,min(a jk ))=b k , then r k =0, id is a positive number, so set 0 as the id corresponding to the robot;

[0042] Step 3.5, calculate the number of times the same element appears in the set R. If 0 appears the most times, it is judged that the face of the person who speaks is facing the robot, otherwise it is judged that the face of the person who speaks is not facing the robot.

[0043] Furthermore, the step 4 specifically includes:

[0044] If the action performer in the input voice is clearly a robot name, the action performer is judged to be a robot;

[0045] If the action performer in the input voice is clearly a non-robot name, then it is determined that the action performer is not a robot;

[0046] If the action performer in the input voice is clearly a first-person pronoun, then it is judged that the action performer is not a robot;

[0047] If the action performer in the input voice is clearly a second-person pronoun, then whether the action performer is a robot is determined based on whether the face of the person speaking is facing the robot. If the face of the person speaking is facing the robot, then the action performer is determined to be a robot; otherwise, the action performer is determined not to be a robot.

[0048] If the action performer in the input voice is clearly a third-person pronoun, then it is judged that the action performer is not a robot;

[0049] If the action performer is missing in the input voice, the action performer is judged to be a robot based on whether the face of the person speaking is facing the robot. If the face of the person speaking is facing the robot, the action performer is judged to be a robot, otherwise the action performer is judged not to be a robot.

[0050] If the action performer in the input speech is missing or is a second-person pronoun, but the visual module detects that the speaker makes at least one micro-movement in the micro-movement set that prompts the other party to talk to him, then the performer of the micro-movement set is judged whether to talk to the robot. If the performer of the micro-movement is a robot, it is judged to be a conversation with the robot; if the performer of the micro-movement is not a robot, it is judged not to be a conversation with the robot; at this time, regardless of whether the face of the person speaking is facing the robot, and the micro-movement is performed on the robot during the speech input process, even if the face of the person speaking is not facing the robot, the action performer is judged to be a robot;

[0051] If the intent of the input speech is an element in the robot-specific intent set, and the action performer in the input speech is missing or is a second-person pronoun, it can be judged as a conversation with the robot. In this case, even if the face of the person speaking is not facing the robot, the action performer is judged to be a robot;

[0052] In other cases, it is judged that the person performing the action is not a robot.

[0053] Based on vision, sound source localization angle and semantic understanding information, the present invention can effectively determine whether the person performing the action is a robot, and thus respond and reply. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 is a schematic diagram of the application scenario structure of an embodiment of the present invention;

[0055] Figure 2 It is a schematic flow chart of the method of the present invention;

[0056] Figure 3 It is a schematic diagram of the coordinate system constructed by the present invention;

[0057] Figure 4 is the unit vector of the robot of the present invention when it rotates horizontally counterclockwise by an angle of θ Coordinate system diagram;

[0058] Figure 5 is a schematic diagram of an angular coordinate system of the robot of the present invention and the positions of the surrounding people;

[0059] Figure 6 It is a schematic diagram of the facial orientation process of determining the action performer of the present invention. DETAILED DESCRIPTION

[0060] In order to make the purpose, technical scheme and technical effect of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings.

[0061] like Figure 1 As shown, a method for a non-wake-up robot of the present invention to determine whether a person who performs an action is a robot, the non-wake-up robot combines vision, sound source angle and other information to identify whether the person who performs the action is a robot after the user's voice input dialogue, specifically, includes the following steps:

[0062] Step 1, such as Figure 2 As shown, the wake-up-free robot collects image recognition character information and user voice input information through the visual module and the voice module, that is, visual information and voice sound source information. The visual information includes character information and micro-motion information, and the voice sound source information includes sound source positioning information and semantic understanding information. Specifically, it includes the following sub-steps:

[0063] Step 1.1, such as Figure 3 As shown, a coordinate system is constructed: the robot's height heightR, with the ground where the robot is located as the origin, the robot's horizontal directions have x-axis and y-axis, the robot's right direction is the positive direction of the x-axis, the robot's front direction is the positive direction of the y-axis, the robot's vertical direction is the z-axis, and the top is the positive direction of the z-axis.

[0064] Step 1.2, obtain the horizontal position state information and height of the person, and construct a set representing the person information: the robot collects the height height of the surrounding people, the horizontal position state information x, y, which respectively represent the horizontal position coordinates of the person, the face orientation vector (faceX, faceY, faceZ), and the information id through its visual module. The id is the unique identifier of the surrounding people. For the same person, the id is unchanged. For the convenience of use, the id of the embodiment of the present invention is set to a positive number; then the person information set S = {id, x, y, height, faceX, faceY, faceZ}; the time frequency of the robot collecting visual information is f, that is, the amount of information generated within one second, the duration from the start of voice input to the end of voice input is t seconds, and the total amount of information generated within t seconds is the set E = {S1, S2, S3, ..., S tf}, there are tf sets S in set E.

[0065] Step 1.3, construct the robot exclusive intention set Y = {exclusive intention Figure 1 , exclusive Figure 2 , exclusive Figure 3 ,…, exclusive intention N}, such as the restaurant service robot’s exclusive intention Y = {serve rice, take chopsticks, pour water}.

[0066] Step 1.4, construct a micro-movement set X = {micro-movement 1, micro-movement 2, micro-movement 3, ..., micro-movement M} to prompt the other party to talk to him, such as X = {pat on the shoulder, raise the elbow and tap}, and the visual module can recognize micro-movements during the user's voice input process and can clearly identify the person who executes the micro-movement.

[0067] Step 1.5, the robot voice module is used to perform sound source localization and speech semantics understanding and obtain corresponding information. The sound source localization is to obtain the horizontal angle θ of the sound source input when the speech starts, 0<=θ<2π, if the person is to the right of the robot, θ=0; if the person is to the right front of the robot, If the person is in front of the robot, If the person is in front of the robot's left side, If the person is to the left of the robot, θ = π; if the person is to the left rear of the robot, If the character is directly behind the robot, If the character is behind the robot to the right, To calculate which user in the surrounding area is inputting voice.

[0068] The speech semantic understanding is to extract the relevant action performers and their intentions, such as Xiao Ming goes to pour water, the action performer is Xiao Ming, and the intention is to pour water.

[0069] Step 2, judging the speaker of the voice according to the sound source localization information, specifically includes the following sub-steps:

[0070] Step 2.1, when the voice module detects that someone starts to speak, the voice module sends the horizontal angle θ of the sound source input when the person starts to speak.

[0071] Step 2.2, such as Figure 4 As shown, construct a unit vector that starts from the origin, points to the positive direction of the x-axis, and then rotates horizontally counterclockwise by an angle of θ Specifically:

[0072] When θ=0,

[0073] when hour,

[0074] when hour,

[0075] when hour,

[0076] When θ=π,

[0077] when hour,

[0078] when hour,

[0079] when hour,

[0080] Step 2.3, such as Figure 5 As shown, extract the information set S = {id, x, y, height, faceX, faceY, faceZ} of each person around when starting to speak, and construct a vector with the starting point as the origin, that is, the position of the robot as the origin and the position of the surrounding people as the end point Calculate the corresponding and The angle between:

[0081]

[0082] Calculate the angle α corresponding to each character in the surrounding area, and select the minimum value of α, that is, the character id corresponding to min(α) is the id of the character who speaks, and record the id at the same time.

[0083] Step 3, such as Figure 6 As shown, after the voice speaker is determined, the face direction of the voice speaker is determined, which specifically includes the following sub-steps:

[0084] Step 3.1, extract the set E of each person around the robot from the beginning of the speech to the end of the speech, assuming that there are P people around the robot. For the convenience of calculation, let the set corresponding to the person who speaks be E1, and the sets corresponding to the rest of the people be E2, E3, ..., E P , and establish a one-to-one mapping table Q between the subscripts of set E and the corresponding character IDs, that is, the subscripts of different sets E correspond to different IDs, and different IDs correspond to the subscripts of different sets E; suppose set E i There are N elements in total. Since the collection frequency is the same, each person in the surrounding area has N elements in the set E. The first element is S i1 , the second element is S i2 , ..., the kth element is S ik , ..., the Nth element is S iN , let S ik The x-axis coordinate of the character's position is x ik , the y-axis coordinate is y ik , the character's height is height ik , the face orientation vector is Then the face orientation vector of the speaker is

[0085] Step 3.2, let the set E i The corresponding character's position is taken as the starting point, with set E j The vector corresponding to the location of the surrounding characters as the end point is Then take the speaker's location as the starting point and set E j The vector corresponding to the location of the surrounding characters as the end point is

[0086] Step 3.3, let the set E i The vector starting from the position of the corresponding character and ending at the position of the robot is The vector starting from the position of the speaker and ending at the position of the robot is

[0087] Step 3.4, calculate the k-th moment vector With vector The angle a between jk And vector With vector The angle b between k , k=1,2,…,N, the expression is:

[0088]

[0089]

[0090]

[0091]

[0092] Get the kth moment a jk (j=2,3,…,P) and a jk The minimum value of b k Compare the size, that is, min(b k ,min(a jk )); then construct a set R, which has N elements. The first element is r1, the second element is r2, ..., and the kth element is r k , ..., the Nth element is r N , such as min(b k ,min(a jk ))=min(a jk ), then according to the mapping table Q, r k is min(a jk ) corresponds to the character id corresponding to j; such as min(b k ,min(a jk ))=b k , then rk =0, because the IDs of surrounding people are agreed to be positive numbers, so it is assumed that 0 is the ID corresponding to the robot.

[0093] Step 3.5, calculate the number of times the same element appears in the set R. If 0 appears the most times, it is judged that the face of the person who speaks is facing the robot, otherwise it is judged that the face of the person who speaks is not facing the robot.

[0094] Step 4: Based on the semantic understanding of the source information, extract the action performer and intention, and identify the micro-movement information based on the speaker's facial orientation to determine whether the action performer in the speaker's voice is a robot. Specifically, it includes:

[0095] If the action executor in the input voice is clearly the name of the robot, then it is determined that the action executor is the robot. For example, if the robot name is Xiaozhi, the voice input is "Xiaozhi, go get a glass of water."

[0096] If the action performer in the input voice is clearly a non-robot name, then it is judged that the action performer is not a robot. For example, if the robot name is Xiaozhi, the voice input is "Xiao Ming, go get a glass of water."

[0097] If the action performer in the input voice is clearly a first-person pronoun, such as "I", then it is judged that the action performer is not a robot, such as the voice input "I'm going to get a glass of water".

[0098] If the action performer in the input voice is clearly a second-person pronoun, such as "you", then whether the action performer is a robot is determined based on whether the face of the speaking person is facing the robot. If the face of the speaking person is facing the robot, then the action performer is determined to be a robot. Otherwise, the action performer is determined not to be a robot, such as the voice input "You go get a glass of water".

[0099] If the action performer in the input voice is clearly a third-person pronoun, such as "he", then it is judged that the action performer is not a robot, such as the voice input "he went to pour a glass of water".

[0100] If the action performer is missing in the input voice, whether the action performer is a robot is judged based on whether the face of the speaking person is facing the robot. If the face of the speaking person is facing the robot, the action performer is judged to be a robot. Otherwise, the action performer is judged not to be a robot, such as the voice input "go get a glass of water".

[0101] If the action performer in the input voice is missing or is a second-person pronoun, but the visual module detects that the speaker makes at least one micro-action in the micro-action set X that prompts the other party to talk to him, then the micro-action performer in the set X is judged to be talking to the robot. If the micro-action performer is a robot, it is judged to be talking to the robot; if the micro-action performer is not a robot, it is judged not to be talking to the robot. At this time, regardless of whether the face of the person speaking is facing the robot, such as the voice input "go get a glass of water" and the robot's shoulder is tapped during the voice input process, even if the face of the person speaking is not facing the robot, the action performer is judged to be a robot.

[0102] If the intent of the input voice is an element in the robot-specific intent set Y, and the action performer in the input voice is missing or is a second-person pronoun, it can be judged as a conversation with the robot. For example, the restaurant service robot-specific intent Y = {serve rice, take chopsticks, pour water}, and the voice input is "go pour a glass of water", at this time, even if the face of the person speaking is not facing the robot, the action performer is judged to be a robot.

[0103] In other cases, it is judged that the person performing the action is not a robot.

[0104] The above is only a preferred implementation case of the present invention and does not limit the present invention in any form. Although the implementation process of the present invention is described in detail above, for those familiar with the art, they can still modify the technical solutions recorded in the above examples, or replace some of the technical features therein with equivalents. All modifications, equivalent replacements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for determining whether a person who performs an action is a robot without waking up the robot, characterized in that: The following steps are involved: Step 1: The wake-up-free robot collects image recognition character information and user voice input information through the visual module and the voice module, namely, visual information and voice sound source information. The visual information includes character information and micro-motion information, and the voice sound source information includes sound source positioning information and semantic understanding information. Step 2, determining the speaker of the voice according to the sound source localization information; Step 3, after determining the voice speaker, determine the face direction of the voice speaker; Step 4: Extract the action performer and intention based on the semantic understanding of the source information, and identify the micro-movement information based on the speaker's facial orientation to determine whether the action performer in the speaker's voice is a robot; The step 3 specifically includes: Step 3.1, extract the set E of all characters around the robot from the beginning of speaking to the end of speaking. Suppose there are P characters around the robot, and let the set corresponding to the speaking character be E1, and the sets corresponding to the other characters are E2, E3, ..., E P , and establish a one-to-one mapping table Q between the subscripts of set E and the corresponding character IDs, that is, the subscripts of different sets E correspond to different IDs, and different IDs correspond to the subscripts of different sets E; suppose set E i There are N elements in total, with the same collection frequency. Each person in the surrounding area has N elements in the set E. The first element is S i1 , the second element is S i2 , ..., the kth element is S ik , ..., the Nth element is S iN , let S ik The x-axis coordinate of the character's position is x ik , the y-axis coordinate is y ik , the character's height is height ik , the face orientation vector is Then the face orientation vector of the speaker is Step 3.2, let the set E i The corresponding character's position is taken as the starting point, with set E j The vector corresponding to the location of the surrounding characters as the end point is Then take the speaker's location as the starting point and set E j The vector corresponding to the location of the surrounding characters as the end point is Step 3.3, let the set E i The vector starting from the position of the corresponding character and ending at the position of the robot is The vector starting from the position of the speaker and ending at the position of the robot is Step 3.4, calculate the k-th moment vector With vector The angle a between jk And vector With vector The angle b between k , k=1,2,…,N, the expression is: Get the kth moment a jk The minimum value of j = 2, 3, ..., P, and a jk The minimum value of b k Compare the size, that is, min(b k ,min(a jk )); then construct a set R, which has N elements. The first element is r1, the second element is r2, ..., and the kth element is r k , ..., the Nth element is r N , if min(b k ,min(a jk ))=min(a jk ), then according to the mapping table Q, r k is min(a jk ) corresponds to the character id corresponding to j; if min(b k ,min(a jk ))=b k , then r k =0, id is a positive number, so set 0 as the id corresponding to the robot; Step 3.5, calculate the number of times the same element appears in the set R. If 0 appears the most times, it is judged that the face of the person who speaks is facing the robot, otherwise it is judged that the face of the person who speaks is not facing the robot.

2. A method for determining whether a person performing an action is a robot without waking up a robot as claimed in claim 1, characterized in that: The step 1 specifically includes: Step 1.1, construct the coordinate system: the height of the robot is heightR, the ground where the robot is located is taken as the origin, the horizontal direction of the robot has x-axis and y-axis, the right direction of the robot is the positive direction of the x-axis, the front of the robot is the positive direction of the y-axis, the vertical direction of the robot is the z-axis, and the top is the positive direction of the z-axis; Step 1.2, obtaining the horizontal position status information and height of the person through the robot vision module, and constructing a set S representing the person information; Step 1.3, construct the robot exclusive intention set Y = {exclusive intention 1, exclusive intention 2, exclusive intention 3, ..., exclusive intention N}; Step 1.4, construct a micro-action set X = {micro-action 1, micro-action 2, micro-action 3, ..., micro-action M} that prompts the other party to talk to the other party, and the robot vision module recognizes the micro-action during the user's voice input process and clarifies the person to be executed by the micro-action; Step 1.5: The robot speech module performs sound source localization and speech semantic understanding based on the collected speech source information.

3. A method for determining whether a person performing an action is a robot without waking up a robot as claimed in claim 2, characterized in that: The step 1.2 is specifically as follows: the robot collects the height height of the surrounding people through its visual module, the horizontal position status information x, y, which respectively represent the horizontal position coordinates of the person on the positive x-axis and the positive y-axis, the face orientation vector (faceX, faceY, faceZ), and the information id, where id is the unique identifier of the surrounding people, and for the same person, the id remains unchanged and is a positive number; then the person information set S = {id, x, y, height, faceX, faceY, faceZ} is constructed; the time frequency of the robot collecting visual information is f, that is, the amount of information generated within one second, the duration from the start of voice input to the end of voice input is t seconds, and the total amount of information generated within t seconds is the set E = {S1, S2, S3, ..., S tf }, there are tf person information sets S in set E.

4. A method for determining whether a person performing an action is a robot without waking up a robot as claimed in claim 2, characterized in that: The sound source localization is to obtain the horizontal angle θ of the sound source input when the speech starts, 0<=θ<2π, if the person is to the right of the robot, θ=0; if the person is to the right front of the robot, If the person is in front of the robot, If the person is in front of the robot's left side, If the person is to the left of the robot, θ = π; if the person is to the left rear of the robot, If the character is directly behind the robot, If the character is behind the robot to the right, The speech semantic understanding is to extract the relevant action performers and their intentions.

5. The method for determining whether a person performing an action is a robot without waking up a robot as claimed in claim 1, characterized in that: The step 2 specifically includes: Step 2.1, when the voice module detects that someone starts to speak, the voice module sends the horizontal angle θ of the sound source input when the person starts to speak; Step 2.2, construct a unit vector that starts from the origin, points to the positive direction of the x-axis, and then rotates horizontally counterclockwise by an angle of θ Specifically: When θ=0, when hour, when hour, when hour, When θ=π, when hour, when hour, when hour, Step 2.3, extract the information set S = {id, x, y, height, faceX, faceY, faceZ} of each person around when the speech starts, and construct a vector with the starting point as the origin, that is, the position of the robot as the origin and the position of the surrounding people as the end point Calculate the corresponding and The angle between: Calculate the angle α corresponding to each character in the surrounding area, and select the minimum value of α, that is, the character id corresponding to min(α) is the id of the character who speaks, and record the id at the same time.

6. A method for determining whether a person performing an action is a robot without waking up a robot as claimed in claim 1, characterized in that: The step 4 specifically includes: If the action performer in the speaker's input voice is clearly a robot name, the action performer is determined to be a robot; If the action performer in the input voice is clearly a non-robot name, then it is determined that the action performer is not a robot; If the action performer in the input voice is clearly a first-person pronoun, then it is judged that the action performer is not a robot; If the action performer in the input voice is clearly a second-person pronoun, then whether the action performer is a robot is determined based on whether the face of the person speaking is facing the robot. If the face of the person speaking is facing the robot, then the action performer is determined to be a robot; otherwise, the action performer is determined not to be a robot. If the action performer in the input voice is clearly a third-person pronoun, then it is judged that the action performer is not a robot; If the action performer in the input voice is missing, the action performer is judged to be a robot based on whether the face of the person speaking is facing the robot. If the face of the person speaking is facing the robot, the action performer is judged to be a robot, otherwise the action performer is judged not to be a robot. If the action performer in the input speech is missing or is a second-person pronoun, but the visual module detects that the speaker makes at least one micro-movement in the micro-movement set that prompts the other party to talk to him, then the performer of the micro-movement set is judged whether to talk to the robot. If the performer of the micro-movement is a robot, it is judged to be a conversation with the robot; if the performer of the micro-movement is not a robot, it is judged not to be a conversation with the robot; at this time, regardless of whether the face of the person speaking is facing the robot, and the micro-movement is performed on the robot during the speech input process, even if the face of the person speaking is not facing the robot, the action performer is judged to be a robot; If the intent of the input speech is an element in the robot-specific intent set, and the action performer in the input speech is missing or is a second-person pronoun, it can be judged as a conversation with the robot. In this case, even if the face of the person speaking is not facing the robot, the action performer is judged to be a robot; In other cases, it is judged that the person performing the action is not a robot.