JP4468777B2 - Control device for legged walking robot - Google Patents

Description

  The present invention relates to a control device for a mobile robot provided with a voice identification unit for identifying a target voice from an input to a voice input unit.

  The development of humanoid robots that move autonomously by bipedal walking while recognizing surrounding conditions based on visual information obtained from video cameras and auditory information obtained from microphones has been active recently. In order to improve the compatibility between humans and humanoid robots in a general living environment and to improve workability, it is desired to enhance the dialogue function by voice.

As a technology related to speech recognition applicable to such a speech dialogue function, when comparing a speech pattern input to a speech input device with a preset standard pattern to determine whether it is a target speech, In order to avoid the influence, Patent Document 1 proposes a method for setting a standard pattern in consideration of environmental noise. Patent Document 2 proposes a method for estimating a noise spectrum and removing a noise component from a voice input.
JP-A-5-134694 JP 2000-163098 A

  However, when performing sound recognition by collecting sound with a microphone provided on the head of the robot, for example, rotation sound of a servo motor provided on a joint of a limb or a cooling fan of an electronic control unit, operation of each movable part The problem is that it is affected by noise generated by the robot itself, such as sounds or footsteps while walking. Incidentally, in the case of the robot developed by the present applicant, the landing sound (footsteps) emitted by the legs during walking is the same volume as the call sound of “Oi” from a position 1 meter away, and the waveform is also similar. (See FIG. 3).

  In other words, since the conventional noise countermeasure methods described in Patent Documents 1 and 2 are difficult to deal with irregularly entering noise and noise with a pattern similar to the speech of interest, even if this is applied to the hearing of a mobile robot, The certainty of distinguishing human voice from ambient noise could not always be improved.

  The present invention has been devised with the object of eliminating the disadvantages of the prior art, and the main purpose of the present invention is to make it possible to perform speech recognition without being disturbed by the operation sound generated by itself. The object is to provide a robot control device.

In order to solve such a problem, a control device for a legged walking robot according to the present invention includes behavior control means for controlling movement of a movable part, movement state quantity detection means including at least a moving speed, voice input means, and voice input. In a control apparatus for a mobile robot provided with voice recognition means for identifying speech of interest from the input to the means, the operation sound of the robot is identified from the input to the voice input means so that the operation sound is reduced when talking to a human. that controls the operation of the movable portion.

Particularly, if the robot is a legged walking robot, rather good when you change the gait based on the identification result of the operation sound, furthermore, to and close to the sound source turning the sound source direction based on the identification result of the operation sound Ri and, if provided with a cooling fan, Ri was to change the air volume of the cooling fan based on the identification result of the operation sound, also to ask the utterance again to speaker based on the identification result of the operation sound but it may also be or to.

  According to the present invention, it is possible to predict the timing at which the operation sound is generated from the signal of the motion state quantity detection means, and thereby it is possible to grasp the operation sound generated by the robot itself. And, when interacting with a human being, it is possible to avoid the interference of the operating sound with the identification of the target voice by capturing the voice while still as much as possible.

  In particular, changing the gait to avoid the landing sound from interfering with the target speech, turning in the direction of the sound source and approaching the sound source to make it easier to hear, or reducing the cooling fan rotation It is effective to increase the certainty of voice recognition by reducing the noise level generated by itself or by requesting the speaker to speak again.

  Hereinafter, the present invention will be described in detail with reference to the accompanying drawings.

  FIG. 1 is a block diagram showing a main configuration of the present invention applied to a mobile robot. In this robot, a pair of microphones 1 is provided on the head, and a sound signal from the microphone 1 is input to the sound recognition unit 2, and a sound source is generated based on a sound pressure difference between the pair of microphones 1 and a sound arrival time difference. The position is specified.

  The voice recognition unit 2 refers to parameters such as the input gain corresponding to the noise level of the environment stored in the noise data management unit 3, the frequency response characteristics of the noise filter, and the volume level of the speaker 4, and the motion state The movement of the leg / arm part 6 of the robot from the quantity detector 5 is input to make it easy to identify the attention speech by grasping the operation sound and footstep sound emitted by the robot, and from the sound waveform, It is estimated whether or not there is an impact sound in which some objects collide with each other, and a voice pattern emitted by a person is compared with a standard pattern to understand its contents.

  In the behavior control unit 7, an action instruction value set based on a predetermined operation command is given to an actuator or the like provided on the leg / arm unit 6 to control the movement of the robot. In particular, the leg portion reduces the step length as necessary based on the identification result in the speech recognition unit 2, lowers the center of gravity to reduce the landing impact, or lowers the walking speed to increase the landing interval. Instructed to walk in silent mode. Furthermore, an utterance signal generated by an appropriate synthesized voice generation unit is output from the speaker 4, or the noise generated by itself is reduced as much as possible by controlling the air volume of the cooling fan 8 of the electronic control unit.

  A camera 9 and an image recognition unit 10 are provided to analyze the captured image of the camera, recognize the distance to the speaker, the gesture and the movement of the mouth, and identify the speaker, and fit the information An instruction is issued from the behavior control unit 7 so as to perform the corresponding operation.

  Next, processing related to speech recognition according to the present invention will be described. The voice recognition unit 2 inputs a voice signal (step 1), detects the motion state amount of the leg 6 based on the operation command value of the behavior control unit 7, and estimates the landing timing (step 2). Then, it is estimated whether or not a landing sound is superimposed on the audio signal, and the sound source position is not detected at the time of landing, and the attention sound is identified such as whether a name unique to the robot is included (step 3). If the audio signal is independent, the voice recognition process is performed (step 4). If the landing sound is superimposed, it is determined that it is difficult to completely recognize the collected voice, and it is temporarily stopped or corrected in the direction of the sound source. Pairing is performed, and the sound source is approached while walking in the silent mode as necessary (step 5).

  Further, the ambient noise level is judged (step 6). If the noise level is high, the output of the cooling fan 8 of the electronic control unit or the servo motor provided at the joint is minimized (step 7). ), Reduce the noise generated by itself, and make it easier to hear the voice of interest.

  Further, depending on the case, a request is made by utterance so that the voice is louder (step 8).

  In this way, the obtained speech signal is processed by the speech recognition unit 2 to identify the attention speech, thereby avoiding that the accurate identification of the attention speech is hindered by the noise generated by the robot itself. Advanced interpersonal dialogue function.

It is a block diagram which shows schematic structure of this invention apparatus. It is a flowchart in connection with the speech recognition process of this invention. It is an input waveform figure of the microphone provided in the head of the robot.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Microphone 2 Voice recognition part 5 Motion state quantity detector 6 Leg part, arm part 7 Action control part 8 Cooling fan

Claims (5)

Action control means for controlling the movement of the movable portion, the motion state quantity detecting means including at least moving speed, voice input means, a control device of a legged walking robot having a及beauty speech recognition means,
The voice recognition means identifies the target voice and the landing sound of the robot from the input to the voice input means,
The action control means, the control system of a legged walking robot that performs control for changing the gait of the robot such that the landing sound is reduced on the basis of the identification result of the landing sound. The voice recognition means estimates whether or not a preset attention voice and the landing sound are superimposed,
2. The legged walking robot according to claim 1, wherein the behavior control unit performs control to change a gait of the robot so that the landing sound is reduced only when the attention sound and the landing sound are superimposed. Control device. 2. The legged walking robot according to claim 1, wherein the change of the gait of the robot is at least one of reducing a step of the robot, lowering a center of gravity of the robot, and reducing a walking speed. Control device. The control device for a legged walking robot according to any one of claims 1 to 3, wherein the control device turns around and approaches the sound source based on the identification result of the operation sound. 5. The control device for a legged walking robot according to claim 1, wherein the speaker is requested to speak again based on the identification result of the operation sound. 6.

Images