JP2007221300A - Robot and control method of robot - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a robot that has a function of making communication with humans and reacts to a call with a simple configuration and to provide a control method of the robot. <P>SOLUTION: The robot is configured such that its head is turnably fitted to a shell, a directivity microphone directed in a front direction of a face of the head is provided to the head, and an array microphone comprising a plurality of microphones arranged in the front part of the surrounding of its neck at a prescribed interval and a control section for controlling the robot are provided to the shell. The control section is configured to include: a voice detection means for detecting voice from the array microphone; a means for detecting a call from a human including a sound source direction from the detected voice; a head turning control means for turning the head in the detected sound source direction; a directivity microphone voice acquisition means for acquiring a voice from the directivity microphone directed in the sound source direction; and an output control means for carrying out output control on the basis of the acquired voice. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a robot having a function of communicating with a person such as an office robot and a method for controlling the robot.

In recent years, interactive humanoid robots with a head and a torso have been widely used in various companies. When a human speaks to such a robot, the direction from which the voice is generated is determined by the robot. There are known techniques that can be detected. One of them is an autonomous behavior robot that has an imaging unit at the eye position of the head and a sound detection unit with left and right microphones at the ear position, and detects the sound generated from the sound source. The localization process identifies the direction of sound and drives the head in the specified direction so that the imager captures images around the direction of the sound source and extracts a target image of a specific shape from the image. This is a technique (see Patent Document 1). As a similar technology, a robot is equipped with a camera at the eye position of the robot's head and left and right microphones at the ear position. The auditory module, visual module, and motor control module extract auditory events, visual events, and motor events, respectively. There is a technique for displaying auditory information, visual information, and motor information (see Patent Document 2). In addition, a microphone array consisting of a plurality of microphones is provided on either the body or head of the robot to receive an acoustic signal from the sound source, calculate the sound source direction from the sum of delays of the input signals for each microphone, and There is a technique (refer to Patent Document 3) in which the direction of the sound source and the direction of the microphone array coincide with each other toward the direction of interest of the robot.
JP 2003-62777 A JP 2002-264052 A JP 2002-366191 A

  In the techniques of Patent Documents 1 and 2 described above, left and right microphones are provided at the position of the ear of the robot head, and visual information is provided at the position of the eye. The direction of the sound source is extracted with a simple configuration. , For the purpose of extracting voice from that direction, there is a problem that the configuration is complicated and expensive. In the technique of Patent Document 3, a microphone array is provided somewhere on the robot, an acoustic signal is received, a sound source direction is calculated from a delay sum for each microphone, and the attention direction of the robot is matched with the pointing direction of the microphone array. However, it does not feature a specific configuration of how to provide a microphone array and detect sound from a sound source in the direction of interest.

  An object of the present invention is to provide a robot having a function of communicating with a person and reacting to a call with a simple configuration and a control method of the robot.

  FIG. 1 is a diagram showing a principle configuration of the present invention. In the figure, 1 is a robot head rotatable with respect to the body, 10 is a directional microphone having directivity in the front direction of the head 1, 2 is a robot body, and 20 is a plurality of microphones arranged. The array microphone, 21 is a control unit, 22 is a head rotation mechanism, and 23 is an output means for performing output (display output, audio output, mechanical output, etc.) according to the caller's voice. 21a in the control unit 21 is a voice detecting means for the array microphone, 21b is a call detecting means for analyzing the acquired voice signal and detecting that the call is made by a person from a certain direction based on the sound source direction, volume and voice structure, 21c Is a head rotation control means, 21d is a sound acquisition means of a directional microphone, and 21e is an output control means for performing output control corresponding to the sound.

  A feature of the present invention is that an array microphone 20 composed of a plurality of microphones arranged at a predetermined interval is provided on the upper part of the robot body 2 (around the neck, etc.) to detect the direction of the calling person (sound source). The directional microphone 10 provided in the head 1 is provided for acquiring the caller's voice.

  First, in the control unit 21, the voice from the calling person is detected by the voice detecting means 21a of the array microphone. Next, when the call detection means 21b analyzes each signal acquired by a plurality of microphones constituting the array microphone 20, and detects that it is the direction of the sound source, the volume and the voice of the person, a signal indicating the direction of the detected sound source is obtained. This is supplied to the head rotation control means 21c. The head rotation control means 21c drives the head rotation mechanism 22 according to the signal indicating the received direction. As a result, the front of the head 1 attached to the body 2 faces the calling person straight. In this state, the caller's voice is acquired from the directional microphone 10 having directivity with respect to the front direction by the directional microphone voice acquisition means 21d, and the output control means 21e is controlled according to the acquired voice, and output control is performed. The output means 23 is driven by the means 21e. The output means 23 includes voice output, display output, robot movement, and the like. The voice acquired by the voice acquisition means 21d is recorded as a “voice message” by the output means 23, or the voice recognition means is added to recognize the voice and control the corresponding output control means 21e. can do. A plurality of array microphones 20 can be arranged in front of or around the neck (around the entire circumference), or a plurality of arrays can be arranged in front of or around the trunk 2 (chest or abdomen).

  According to the present invention, when a voice is spoken from the surroundings of the robot, the robot reacts and the head (front of the face) is directed toward the person who made the voice. I feel that I have noticed that I am motivated to communicate with the robot. Also, if an output operation is performed according to the content of the voice of the person who made the voice, it will be understood that the robot has recognized the voice, and this will trigger further deeper communication.

  In addition, only sound from the direction in which the voice is spoken can be extracted by the directional microphone, the sound in the other direction can be suppressed, and an audio signal with a high S / N ratio can be input. The recognition rate can be improved.

  FIG. 2 shows an external configuration of an embodiment of the robot, and FIG. 3 shows a system configuration of the robot.

  In FIG. 2, reference numerals 1 and 2 are the same as the same reference numerals in FIG. 1, 1 is the head, and 10 is the directivity provided at the center position of the face of the head 1 (in the example shown, the position of the nose). 11 is a camera provided at the position of the eye of the head, 2 is a torso (body), 20 is a semicircle in front of the robot body 2 around the neck and connecting both shoulders with the head 1 as an axis. An array microphone composed of a plurality of microphones arranged at regular intervals along the line, 24 is a monitor unit for displaying characters and images to an external person, 25 is a speaker for outputting sound to the external person, and 26 is a robot. , 27 is a moving device that supports the body 2 of the robot and moves the robot. In the embodiment of FIG. 2, the array microphone 20 is arranged at regular intervals in front of the neck. However, the arrangement type of the array microphone 20 is as follows depending on the arrangement position and the arrangement configuration. It can be provided in various formats as shown in (4).

(1) Form to be placed in front of neck
(2) Place around the neck, for example, place 6 microphones every 60 degrees around the circumference.
(3) Form to be placed in front of the fuselage
(4) Form to be placed around the fuselage

  FIG. 3 shows a system configuration of the robot, which is a system configuration for controlling the entire robot shown in FIG. In the figure, reference numeral 30 denotes a RAM that stores data, a ROM that stores programs and data, and a voice process that performs processing such as analysis (volume, frequency band, etc.), voice recognition, and voice synthesis on voice (digital signal) input from a microphone. A CPU board equipped with a circuit and a CPU for processing by a program, 31 is a head rotating mechanism, 32 is a directional microphone (10 in FIG. 2) provided on the head, 33 is an array microphone (20 in FIG. 2), 34 is a camera (11 in FIG. 2), 35 is a monitor unit (24 in FIG. 2), 36 is a moving device drive mechanism, 37 is a speaker (25 in FIG. 2), and 38 is a power supply unit that supplies power to each part of the robot. It is.

  FIG. 4 is a flowchart of the first embodiment, which is executed in the CPU board 30 of FIG.

  First, the sound is acquired by the array microphone (S1 in FIG. 4), and the sound source direction is determined from the sound acquired by the array microphone (a plurality of sounds from the plurality of microphones) (S20 in the same), and the sound volume from the microphone is determined. (S21), and further, it is determined whether the voice is likely to be a human voice (S22). Note that “speech quality” is a degree indicating that the sound is not mechanical noise, but it is possible to determine whether the sound seems to be sound or not by analyzing the frequency characteristics and the discontinuity pattern of the input sound. The determinations of S20 to S22 can be performed at the same time, but may be performed in order.

  After this, with regard to the results of S20 to S22, the sound can be acquired with (1) a certain level of sound volume, (2) a certain level of sound quality (degree), and (3) a certain level of sound source direction certainty. (S3 in FIG. 4). Thereby, it is determined whether or not the robot is called by a person. If it is determined in this determination that even one voice that does not satisfy the condition cannot be acquired, the process returns to step S1, and if it is determined that the voice that satisfies all the conditions is acquired, the call direction (sound source direction) Using the information, the head of the robot is rotated in that direction (S4 in FIG. 4). By rotating the robot's head, the caller can see from the gesture that the robot has turned his attention to his voice. This improves communication. Next, the voice is acquired by the directional microphone (S5), and thereafter, the voice of the person who called by the directional microphone is clearly acquired, and the process using the acquired voice is performed (S6). As processing using the voice at this time, the head is moved in the voice direction to react to the user, the voice is recognized and control is performed according to the recognition, or some display on the monitor unit (24 in FIG. 2). , Recording a voice message (implemented by mounting a voice recording chip on the CPU board 30), performing some voice output from the speaker (25 in FIG. 2), and arm (26 in FIG. 2) ).

  It is possible to continue detecting the voice direction by the array microphone even while the above-described processing using the voice (the processing of S6 in FIG. 4) is performed, and to follow and rotate the head even when the user moves. it can.

  In addition, pseudo directivity can be created with an array microphone, but using a directional microphone can easily improve performance. Further, since the array microphone is used only for detecting the voice direction, it is possible to use a microphone whose performance is not so high.

  FIG. 5 is an explanatory diagram of the certainty of the sound source direction measurement. A. of FIG. Indicates the configuration of the array microphone; Indicates the correlation of the microphone signal. A. of FIG. , 20a to 20e represent individual microphones constituting the array microphone 20 provided around the neck of the robot body 2 shown in FIG. 2 (along the front semicircle connecting both shoulders). Although the number is provided, the number is not limited to five. In such a configuration, the direction of the sound source is detected as follows.

  Sound source direction detection is basically performed by the following process for correlating audio input to many microphones.

  That is, when sound is input to the microphone 20a, microphones 20b, 20c,..., 20e from the direction of θ from the front, a time difference occurs to reach the five microphones. The sound acquired from each microphone is compared with a delay according to the assumed angle θ, and the most consistent angle θ is set as the direction of arrival of the sound. B. of FIG. Represents the assumed angle on the horizontal axis and the degree of coincidence on the vertical axis, and the angle (direction) with a high degree of correlation is determined as the direction of sound arrival. Actually, because there is noise in the environment, the graph has peaks at angles other than the direction of arrival, and the waveform is distorted. This degree is the certainty of sound source direction measurement (same as reliability). Specifically, when the second peak (second height) exists, the difference in height from the first peak (first height) is the certainty (reliability). Even when there is only one peak, the sharpness (sharpness) of the peak is the certainty (reliability) of the measurement.

  FIG. 6 is a flowchart of the second embodiment. First, sound is acquired by the array microphone (S1 in FIG. 6), the sound source direction is determined from the plurality of sounds acquired by the array microphone (S20), the volume from the array microphone is determined (S21), and , The input sound is recognized (S22). Based on these processing results, whether or not a speech that satisfies the following conditions: (1) at a certain volume level, (2) a keyword included in the recognized word, and (3) certainty of the sound source direction above a certain level Determine. If it is determined in this determination that even one of the conditions is not satisfied, the process returns to step S1, and if it is determined that a sound satisfying all the conditions has been acquired, information on the called direction (sound source direction) is used. The robot head is rotated in that direction (S4 in FIG. 6). Subsequently, the voice of the person called by the directional microphone oriented in the direction of the sound source is acquired by rotating the head of the sound source (S5 in FIG. 6), and processing using the sound is performed (S6).

  In the second embodiment, the voice recognition (S22 in FIG. 6) can react when a call is made with a name of a robot (for example, a name such as “Robochan”) as a specific keyword by voice recognition (S22 in FIG. 6). And no response to sudden noise. It should be noted that a plurality of keywords are registered instead of one, and it can be configured to react when even one of them is called. For example, “Robot-kun” or “Konchi-wa”.

  FIG. 7 is an explanatory diagram of voice recognition, which is executed for the voice acquired from the array microphone in step S22 of FIG. 6, and is also executed in S22 of FIG. Note that the same principle can be used in S61 of FIG. 8 (however, the voice acquired from the directional microphone is recognized).

  In FIG. 7, 20 is an array microphone, 30 is a word segmentation unit, 31 is a comparison unit, 32 is a word bank, 33 is a recognition result output unit, and 34 is a table. The principle of this voice recognition is that each candidate word in the word bank 32 in which a voice pattern representing a word is cut out from the voice inputted from the array microphone 20 by the word cutout unit 30 and the voice pattern of each candidate word is registered in the comparison unit 31 in advance. The degree of coincidence is calculated by comparing with the voice pattern of the list, and the degree of coincidence (score) is stored for each candidate word in the table 34 by the recognition result output unit 33. The shorter the word and the unclear voice, the less the difference between the word lists. In reality, the S / N is lowered or the intonation is different from that assumed, and the recognition becomes further ambiguous. The reliability of speech recognition can be calculated from the difference between the score (score) of the first candidate word and the score of other words. Simply, it is obtained by calculating from the score ratio of the first candidate and the second candidate.

  FIG. 8 is a flowchart of the third embodiment. In the flowchart of the third embodiment, a sound is acquired by an array microphone (S1 in FIG. 8), sound source direction, volume, and voice recognition are performed (S20 to S22), and a keyword is assigned to a recognized word at a volume higher than a certain level. It is included and it is determined whether or not the sound has been acquired with a certain sound source direction certainty (S3), the head is rotated in the calling direction (S4), and the sound is acquired by the directional microphone (S5). The processing up to this is the same as the flowchart of the second embodiment shown in FIG. Thereafter, the volume of the sound acquired by the directional microphone (10 in FIG. 2) is detected (S60 in FIG. 8), and the voice acquired as a clear signal by the directional microphone is recognized (S61). It is evaluated whether the acquired sound is the voice that the person talks to the robot. That is, it is determined whether there is a sound volume above a certain level and the content of the word can be recognized (S7 in FIG. 8). If it can be recognized (evaluated as a voice that can be recognized by a person), it is determined that the call is for a robot. If it is determined that it is not a human voice, the rotated neck is returned to the surroundings. The mode returns to the mode for paying attention to the sound (S8 in FIG. 8). If it can be determined, processing using voice is performed (S9 in FIG. 8). In step S8, the rotated neck is returned. However, in step S1, the process may be shifted to the process of acquiring the sound by the array microphone in order to listen to the surrounding sounds.

  According to the third embodiment, it is possible to configure a robot that reacts to voice even in an environment where it is not possible to acquire the image with an array microphone.

  (Supplementary note 1) In a robot whose head is rotatably attached to the torso, the head is provided with a directional microphone directed in front of the face, and the torso has either a neck or a torso An array microphone composed of a plurality of microphones arranged at regular intervals either in front of or around and a control unit for controlling the robot, wherein the control unit detects voice from the array microphone. Means, a means for detecting a person's call including the direction of the sound source from the detected sound, a head rotation control means for rotating the head in the detected sound source direction, and a sound from a directional microphone directed to the sound source direction. A robot comprising voice acquisition means for a directional microphone to be acquired and output control means for performing output control based on the acquired voice.

  (Supplementary Note 2) In Supplementary Note 1, the call detection means of the control unit determines whether the volume acquired from the array microphone is higher than a certain level, whether the direction of the sound source is detected, or whether the voice is detected. A robot characterized by detecting it as a call by detecting it.

  (Supplementary note 3) In Supplementary note 1, when the call detection means of the control unit recognizes the voice acquired from the array microphone and detects a predetermined keyword, the head rotation control means moves to the sound source direction. A robot characterized by performing head rotation control.

  (Supplementary Note 4) In any one of Supplementary Notes 1 to 3, after the head rotation control by the head rotation control unit of the control unit, the voice acquired from the directional microphone by the voice recognition unit is evaluated, and the voice is If the evaluation cannot be performed, the robot rotates the head rotation control means to rotate the head to return to the previous direction.

  (Additional remark 5) In additional remark 4, about the audio | voice recognized from the said directional microphone, the said voice recognition means evaluates as a voice by judging whether the volume was more than a fixed amount and the word was able to be recognized by voice recognition. Characteristic robot.

  (Appendix 6) The head is rotatably attached to the torso, and the head is provided with a directional microphone directed in the front direction of the face, and the torso is provided at a predetermined interval around the neck in front of it. In a method for controlling a robot having an array microphone composed of a plurality of microphones arranged with a microphone, sound is acquired by the array microphone, a sound source direction is detected from the sound acquired by the array microphone, and the detection is performed. The head is rotated in the direction of the sound source, and after the head is rotated, the voice is acquired from the directional microphone, and the voice acquired from the directional microphone is recognized and controlled. Robot control method.

It is a figure which shows the principle structure of this invention. It is a figure which shows the external appearance structure of the Example of a robot. It is a figure which shows the system configuration | structure of a robot. 3 is a flowchart of the first embodiment. It is explanatory drawing of the certainty of a sound source direction measurement. FIG. 6 is a diagram illustrating a flowchart of a second embodiment. It is explanatory drawing of speech recognition. FIG. 10 is a diagram illustrating a flowchart of a third embodiment.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Robot head 10 Directional microphone 2 Robot body 20 Array microphone 21 Control part 21a Array microphone voice detection means 21b Call detection means 21c Head rotation control means 21d Directional microphone voice acquisition means 21e Output control means 22 Head Part rotation mechanism 23 output means

Claims (4)

In a robot whose head is rotatably attached to the torso,
The head is provided with a directional microphone directed in front of the face,
The body includes an array microphone composed of a plurality of microphones arranged at regular intervals either in front of or around the neck or the body, and a controller for controlling the robot.
The control unit includes: sound detection means from the array microphone; means for detecting a call to a person including a sound source direction from the detected sound; head rotation control means for rotating the head in the detected sound source direction; Directional microphone sound acquisition means for acquiring sound from a directional microphone directed in the direction, and output control means for performing output control based on the acquired sound,
A robot characterized by that. In claim 1,
The means for detecting the call of the control unit is as a call by detecting whether the volume acquired from the array microphone is higher than a certain level, whether the direction of the sound source is detected, or whether the voice is detected. A robot characterized by detecting. In claim 1,
The means for detecting the call of the control unit performs head rotation control in the direction of the sound source by the head rotation control means when the voice obtained from the array microphone is recognized and a predetermined keyword is detected. A robot characterized by that. The head is rotatably attached to the torso, and the head is provided with a directional microphone directed in the front direction of the face, and the torso is arranged at a predetermined interval around the neck. In a control method for a robot having an array microphone composed of a plurality of microphones,
Voice is acquired by the array microphone,
The direction of the sound source is detected from the sound acquired by the array microphone,
Rotate the head in the direction of the detected sound source,
After rotating the head, obtain voice from the directional microphone,
Recognizing and controlling the voice acquired from the directional microphone,
A robot control method characterized by the above.

Images