KR19990073551A - Toy robot for speech recognition and control method thereof - Google Patents

Abstract

The present invention relates to a voice recognition robot toy that listens to a user's voice and performs an action required by the user, and a control method thereof. The robot toy includes a memory means for storing a plurality of voice patterns for recognizing the voice signal of the user, at least one compressed melody sound having a predetermined pitch and beat, and a voice of the user. A system control unit for recognizing a voice pattern and outputting a robot control signal having a predetermined operation pattern according to the voice command, and a robot mechanism unit operating in at least one pattern based on the robot control signal of the system control unit. . Accordingly, the user's voice command is speech-independent and the user performs the operation to generate the appropriate dog motion and dog sound according to the corresponding robot mechanism (doggy mechanism), and the user has a price competitiveness and operation reliability. Maximize the interest and interest of, and get tired of use can improve the life as a toy.

Description

Robot toy for voice recognition and its control method {TOY ROBOT FOR SPEECH RECOGNITION AND CONTROL METHOD THEREOF}

The present invention relates to a voice recognition robot toy and a method of controlling the same, and more particularly, to a voice recognition robot toy and a method of controlling the same, for receiving a user's voice and performing an action required by the user.

In general, robot toys are classified into high-end type of remote control method using electric power and simple type using mainspring, and are commercialized in various forms. In addition, there are various forms of toys that cause movement in response to sound.

Therefore, such a robot toy has a problem that it is easily demonstrated when the user executes all of its functions two or three times, and the user does not have the interest, so the lifespan is short.

The present invention has been made in order to solve the above problems, perform the operation to generate the appropriate dog action and dog sound in accordance with the corresponding robot mechanism (dog mechanism) by speech recognition of the user's voice command to the speaker-independent type. The present invention aims to provide a voice recognition robot toy and its control method which have price competitiveness and operation reliability.

1 is a block diagram showing a robot toy for speech recognition according to the present invention,

Figure 2 is a block diagram showing in detail the system control unit of the voice recognition robot toy according to the present invention,

3 is a block diagram of an ASIC speech recognition unit according to the present invention;

Figure 4 is an exploded perspective view showing a robot mechanism part of the voice recognition robot toy (TOY ROBOT) according to the present invention,

5 is an operation cross-sectional view showing only the front and rear legs corresponding to the standing posture of the voice recognition robot toy according to the present invention,

6 is an operation cross-sectional view showing only the front and rear legs corresponding to the walking posture of the voice recognition robot toy according to the present invention,

Figure 7 is a cross-sectional view showing only the front and rear legs corresponding to the sitting position of the voice recognition robot toy according to the present invention,

8 is a front sectional view showing the front leg of the voice recognition robot toy according to the present invention,

9 is a side cross-sectional view of FIG. 5,

10 is a side cross-sectional view showing a state in which the front leg of the voice recognition robot toy in accordance with the present invention,

11 to 12 is a side cross-sectional view showing the operation of the neck portion of the voice recognition robot toy according to the present invention,

13 to 14 is a cross-sectional view showing the tail operation of the voice recognition robot toy according to the present invention,

15 is a schematic system diagram showing an operation flow of a voice recognition robot toy according to the present invention;

16 is a system flow diagram showing the operation flow from the sensing of the robot mechanism unit to the motor drive unit according to the present invention,

17 is an operation flowchart showing a calibration operation of the voice recognition robot toy according to the present invention,

18 is a flow chart showing the overall operation of the voice recognition robot toy according to the present invention.

Explanation of symbols on the main parts of the drawing

One ; System control unit 3; Memory

5; Robot mechanism part 6; First microphone

7; Second microphone 8; speaker

9; Sound generator 10; Power amplifier

11; A / D D / A converter 12; Volume control unit

13; First memory section 14; Clicker

15; Second memory section 16; Timer

17; Circular buffer 18; Power supply

19; Speech recognition unit 21; Robot control unit

23; Sound decoder 25; List Controller

27; Memory controller 29; Voice recognition calculation unit

31; Zero crossing rate 33; Energy output department

35; Unit speech detector 37; Preprocessor

39; Preamplification 40; Robot body

43; Pastoral ministry 45; Walking drive part

47; Posterior joint drive 49; Tail pivot

51; Anterior joint drive 53; Rotating cam

55; Support bar 57; Pivot pin

59; Rotating plate 60; Shoulder

63; Lifting cam 65; Lift guide part

67; Rotary induction pin 69; Power transmission bar

70; Shoulder joint 73; Branch office

75; Front contact 77; Anterior knee joint

79; Fore ankle joint 80; motor

81; Front joint gear 83; rope

85; Elastic restoring member 87; Homing Cam

89; Tail member 91; Pivotal support

93; Power transmission gear 95; Sitting cam

97; Hip support 99; Posterior knee joint

100; Hind limb branch 101; Back ground contact

103; Posterior grazing joint 105; Departure prevention part

M1; 1st motor drive part M2; 2nd motor drive part

M3; Third motor drive unit M4; 4th motor drive part

M5; 5th Motor Drive

According to the present invention, in the voice recognition robot toy that follows the user's voice, a plurality of voice patterns for recognizing the voice signal of the user, at least one piece having a predetermined pitch and time signature A memory controller for compressing and storing a melody sound, a system controller for recognizing the voice of the user from the voice pattern of the memory means, and outputting a robot control signal to have a predetermined operation pattern according to the voice command; It is achieved by a voice recognition robot toy comprising a robot mechanism part operating in at least one pattern based on the robot control signal of the system control part.

In addition, the first microphone for converting the voice of the user and the outside noise into an electrical signal and output to the system control unit, the second microphone for converting the outside noise into an electrical signal and output to the system control unit, Preferably, the system controller further includes a sound generator that listens to the user based on a melody sound signal.

Here, the system controller preferably extends the compressed melody sound of the memory means based on the recognized voice of the user and outputs the melody sound signal.

It is preferable that an A / D D / A converter is provided between the system controller and the microphone to convert either one of an analog audio signal and a digital audio signal to the other side.

The memory unit may include a first memory unit configured to store a plurality of melody sound compression data in which a plurality of melody sounds are compressed at a predetermined compression rate, and a plurality of voice patterns for speech recognition, and the user input from the outside. Preferably, a calculation space is provided for recognizing a voice signal and the robot mechanism unit includes a second memory unit for storing the robot algorithm to perform the doggy operation.

The system controller may further include a circular buffer for temporarily storing a digital voice signal of a user in units of frames output from the first microphone, and a digital voice signal stored in the circular buffer. A voice recognition unit for recognizing the user's voice with a Viterbi algorithm according to a voice recognition word according to a recognition constant, and a robot that controls the robot mechanism unit according to the robot's command of the user's command recognized by the voice recognition unit. Preferably, the control unit includes a sound decoder that extends and restores the melody sound compression data having the dog sound and outputs the sound generator.

In addition, the robot mechanism unit based on the robot control signal from the system control unit walking driving means for generating a walking driving force to at least one pair of four legs to walk the four legs, based on the robot control signal from the system control unit A front joint driving means for driving the at least one front leg joint of the two forelimbs, a rear joint driving means for driving the rear leg joints at the same time based on a robot control signal from the system control unit Preferably, it includes a robot body that accommodates the rear joint driving means, the front joint driving means and the working joint driving means and is provided with four legs. In addition, it is preferable to further include a neck rotating means for rotating the head up and down based on the robot control signal from the system control unit, and a tail rotating means for rotating the tail to the right and left based on the robot control signal from the system control unit.

Here, the pastoral driving means is a first motor driving unit for generating a rotational motion having a predetermined ratio by the robot control signal, a rotary cam for converting the rotational motion of the first motor driving unit into a linear motion, and the rotary cam It is preferable to include a support bar for linear movement, and a rotation plate connected to the support bar and pivoting about a pivot pin installed to be fixed to one side of the robot body according to the linear movement.

The tail rotation means may include a second motor driver for generating a rotational motion having a predetermined ratio by the robot control signal, and a rotational cam for converting the rotational motion of the second motor driver to a rotational motion having a predetermined circumferential angle. A means, a tail member pivoting by the circumferential angle by the pivoting movement, a guide bar connecting the pivoting cam means and the tail member to guide the pivoting force, and a support member supporting the pivoting force of the guide bar. It is desirable to.

The walking driving means may include a third motor driving part for generating a rotational motion having a predetermined ratio by the robot control signal, and a pair of lifting and lowering shoulders of the front leg based on the rotational motion of the third motor driving part. An elevating cam, an elevating guide means installed on the front leg to guide the elevating movement, and one side of the elevating guide means and the shaft of the elevating cam are connected to induce the forward leg to rotate forward by a predetermined circumferential angle. It is preferable to include a rotation guide pin, and a power transmission bar installed at one side of the front leg and one side of the rear leg to allow the front leg and the rear leg to access each other.

In addition, the front joint driving means is connected to the walking driving means, and the rectangular shoulder portion formed integrally with the lifting guide means of the walking driving means;

The front leg portion provided at the end of the shoulder portion and the front leg portion provided at the end of the front leg portion are in contact with the ground, the front contact portion, and between the front surface contact portion and the front leg portion, or between the front leg portion and the shoulder portion, respectively A front joint gear part for rotating by the circumferential angle of the fourth motor driving part for generating rotational force to the front joint gear part, and a joint part for bending each of the front leg support part and the front contact part according to the rotational force generated from the front joint gear part; It is preferable to include a rope connected to the front joint gear to bend the front surface contact portion.

The rear joint driving means may further include a fifth motor driver for driving a predetermined rotation angle in either of the pre-purification and reverse rotation directions based on the robot control signal, and the pair of rotational force from the fifth motor driver. A power transmission gear unit for transmitting to a rear leg of the seating shaft, and a seating cam provided at an end portion of the shaft provided on the rotation shaft of the power transmission gear unit to guide the rotational force in a direction in which the rear half of the robot sits downward, and a rotation shaft of the seating cam; A hip support part coupled to the upper end portion of the rear leg at a predetermined distance from each other so as to be relatively rotatable with each other, a rear leg support connected to a lower end of the hip support part and rotated by a predetermined circumferential angle by the rear joint; A rear surface contact portion connected to a lower end of the rear leg portion and in contact with the ground; One side of the elbow joint and the robot body located above the power transmission gear portion are mutually coupled to prevent the separation and separation of the rear elbow portion, the rear limb and the rear ground contact portion from the robot body and bending of the rear elbow joint It is preferable to include a departure prevention portion to assist the operation.

According to the present invention, in the method of controlling a robot toy for speech recognition having a robot mechanism, determining the initial posture of the robot from a sensor informing the initial state of the robot mechanism, and determining the initial posture Afterwards correcting with at least one basic posture, recognizing a voice command of the user, and controlling the robot mechanism to follow the operation of the voice command based on the recognition. It is achieved by the control method of robot toy.

Hereinafter, with reference to the accompanying drawings will be described in detail a voice recognition robot toy and its control method as follows. 1 is a block diagram showing a voice recognition robot toy according to the present invention, Figure 2 is a block diagram of the ASIC speech recognition unit according to the present invention, Figure 3 is a flow of operation of the voice recognition robot toy according to the present invention Is a schematic diagram.

As shown in Figure 1, the robot toy is a system control unit (ASIC-application specific IC / custom-type semiconductor) 1 for controlling the whole, and the melody sound data having a predetermined voice recognition and dog sounds are stored in the system A memory unit (ROM, RAM) 3 provided to the control unit 1, and a robot mechanism unit 5 that performs an arbitrary operation by the command of the system control unit 1 is included. In addition, a first microphone 6 for converting a user's voice and outside noise into an electrical signal and outputting the electrical signal to the system controller 1, and converting the outside noise into an electrical signal, the system controller 1 And a second microphone 7 which is outputted as a second sound generator, and a sound generator 9 which audibly sounds to the user based on the melody sound data from the system controller 1. An A / D D / A converter 11 for converting one side of an analog audio signal and a digital audio signal to the other side is provided between the system controller 1 and the microphones 6 and 7. Here, the gear box in the plurality of motor driving part of the robot mechanism part 5 is divided into legs, tails according to the sorting operation, such as standing, sitting, lying, lying, walking, forefoot, tail swinging, head bending A plurality of sensors (not shown) for sensing the current position of the head is installed, the sensor generates a sensing signal to recognize the current operation of the robot mechanism unit 5 to the system control unit 1, In accordance with the processing algorithm of the system control unit 1, it provides important information for the sensing signal to be properly harmonized with the voice of the user.

The sensor is installed at the position shown in Table 1 below.

Symbol Mounting position Sensing status Measurement content S1 Left forelimb shop Forefoot S2 Bottom Paw down S3 Sitting Lower leg joint bent S4 clerk Lower foreleg joint straight S5 Right forelimb shop Forefoot S6 Bottom Paw down S7 Sitting Lower leg joint bent S8 clerk Lower foreleg joint straight S9 Hind legs shop Forefoot S10 Bottom Paw down S11 Sitting Lower leg joint bent S12 clerk Lower foreleg joint straight

The memory unit 3 compresses and stores a plurality of voice patterns for recognizing a voice signal of a user and at least one melody sound having a predetermined pitch and beat, and also compresses and stores a dog sound according to the melody sound. First memory unit (ROM) 13, in which a plurality of digital sound signal streams of a melody sound and a dog sound are composed of a plurality of melody sound compression data compressed at a predetermined compression rate and a plurality of speech patterns for speech recognition. And a second memory unit (RAM) 15 which provides a computing space for recognizing the voice signal of the user input from the outside. Here, the second memory unit 15 stores a robot algorithm for causing the robot mechanism unit 5 to perform a doggy operation, and outputs it to the system controller 1.

The sound generator 9 power amplifies the voice signal extended and restored from the system controller 1 so as to be audibly heard by the user through the speaker 8. Have Between the system controller 1 and the first and second microphones 6 and 7 converts the output signals of the first and second microphones 6 and 7 into digital, and converts the system controller 1 and the system into digital. The A / DD / A converter 11 is connected between the power amplifiers 10. Here, the loudspeaker 8 is configured so that the melody sound compressed data stored in the first memory unit 13 is processed in a predetermined process and is presented to the user.

On the other hand, the volume controller (VOLUME CONTROLLER) 12 for adjusting the output strength of the power amplifier 10 to actually increase the voice generated from the speaker 8 and the A / DD / A converter 11 and It is connected between the power amplifier 10. For example, the volume control unit 12 is a volume control command (eg, "SPEAK LOUDER" and "sound" through the first microphone 6 itself to adjust the volume of the voice desired by the child. SPEAK SOFTER "} is input via the A / DD / A converter 11, so that the power amplifier 10 is controlled so that voice is generated at the loudness according to the command from the speaker 8. To control. As a result, the power amplifier 10 determines its magnitude, that is, gain, based on an unMute signal of the system controller 30 and an output signal of the volume controller 12.

The power regulator 18 maintains an arbitrary voltage in the range of 3 to 24 V at a constant voltage of 3.3 V. It basically uses the voltage of 3 series batteries (4.5 V), but it is optional. It is also possible.

As shown in FIG. 2, the system controller 1 is an integrated chip composed of about one million gates, and the user's voice signal input to the first and second microphones 6 and 7, namely, A / A circular buffer 17 for temporarily storing the digitized voice sampling signal converted in units of frames converted by the DD / A converter 11 and the digital voice signal stored in the circular buffer 17 are stored in the first memory unit ( 13) has a speech recognizer 19 for recognizing the meaning of the child's voice with a Viterbi algorithm according to the speech recognition constant of the compressed data stored in the compressed data. The robot controller converts a user's command recognized by the voice recognition unit 19 into a predetermined robot control signal based on the robot algorithm of the memory unit 15 and outputs it to the robot mechanism unit 5. 21 contains more And there. In addition, the system controller 1 has a sound decoder 23 which expands and restores the melody sound compression data having the dog sound and outputs the sound to the sound generating unit.

The first memory unit 13 compresses and records a number of melodies and features according to a number of scenarios in which dog sounds are produced. The device used has a large capacity of 4 MBits or more and can store data in 1 word units (16 bits) for a total of 2 Mwords.

The second memory unit 15 is a block list (block) as an element for internal processing of the data signal by storing a processing program for the dog operation suitable for the user's voice, response and user voice in the system control unit 1 space for a list), space for preprocessing of voice recognition, and a data storage capacity of at least 16 MBits. At this time, the list controller 25 not only extracts the data of the second memory unit 15, but also extracts the compressed voice data of the first memory unit 13, so that the sound decoder of the system controller 1 23 is provided in the system control unit 1 to output to the control unit 23. Here, a memory controller 27 is provided between the second memory unit 15 and the system control unit 1 for data transfer therebetween. In particular, the memory controller 27 reads the compressed data and reads the compressed data. It is configured to output to the two memory section 15.

In addition, the system control unit 1 is to have a function of removing the noise input from the first and second microphones 6 and 7, for example, the first microphone 6 has a mixture of voice and noise. Is input, the pure microphone signal is input to the second microphone (7) when the toy is in contact with the user and any object or affected by the ambient noise, the system controller (ASIC) (1) of the two signals In order to reduce noise by using a correlation (function) between noises, that is, noise by correlating a voice and a noise signal through the first microphone 6 and a pure noise signal through the second microphone 7. Remove only ingredients. The first and second microphones 6 and 7 are mounted on one side and the other side of the toy based on an experimental basis, and in particular, any one of the first and second microphones 6 and 7 As a stereo microphone, the stereo microphone is installed as sensitive and directional in the voice frequency band. As elements necessary for other operations, a clock generator 14 of 24.546 MHz, a timer 16 of 32.768 kHz, and the like of the second memory unit 15 clock are configured as described above. It is essential to operate the elements, so description is omitted.

As illustrated in FIG. 3, the voice recognition unit 19 may apply a predetermined amount of noise to the digital voice signal stored in the circular buffer 17 according to the voice recognition constant of the first memory unit 13. A speech recognition calculation unit 29 for removing and calculating eigenvalues for one character as feature vector data, a zero crossing rate 31 for detecting zero points in a sampling value of the digital speech signal, and the zero crossing rate In order to improve the reliability of the detection of the zero point in the trace 31, the energy calculation unit 33 for calculating the energy for the O point, and the zero crossing plate 31 and the energy calculation unit 33 A unit speech detector 35 for detecting the end point data of any one word of the continuous digital speech signal based on the output signal, the feature vector data of the speech recognition calculator 29, and an end point of the unit speech detector 35; Based on the data By using the list controller 25, the voice processing data of the preprocessor 37 and the first memory unit 13 corresponding to the words classified by the preprocessor 37 are extracted by the list controller 25. And a second memory unit 15 for computing by the Viterbi algorithm. Here, a pre-emphasis 39 is provided between the voice recognition unit 29 and the circular buffer 17 to amplify the digital voice signal of the cooler buffer 17 to be processed faster.

In more detail, the calculation flow of the speech recognizer 19 and the configuration of the module are composed of two module groups, a Viterbi algorithm and a speech detecter algorithm. ) Consists of many server-modules that are directly converted into custom semiconductors.

First, the Viterbi algorithm is composed of one chip (CHIP) using Viterbi algorithm using HMM (Hidden Markov Model) to be used for toys of children aged 4-10 years. In addition, a block list structure that can be accommodated in the external second memory unit 15 (16 MBits) is used to process a large number of variable data generated in the process of performing the Viterbi algorithm. It is configured to operate in the area of the second memory unit 15 of 1 MBits. HMM learning method is to improve the reliability even if the users are different, that is to be speaker-independent recognition, and to recognize the phoneme unit.

The structure of the robot mechanism part 5 will be described in more detail as follows. Figure 4 is an exploded perspective view showing the robot mechanism part of the voice recognition robot toy (TOY ROBOT) according to the present invention. 8 is a front sectional view showing a front leg of the voice recognition robot toy according to the present invention, Figure 9 is a side cross-sectional view of the front leg corresponding to a standing posture of the voice recognition robot toy according to the present invention, Figure 10 is a present invention Side cross-sectional view showing a state in which the front leg of the voice recognition robot toy according to the bending.

That is, the robot mechanism portion 5 is divided into the head, the robot body 40 and the leg portion, as shown in FIG. The head is made by the pastoral moving part 43 to be rotated up and down, the leg has two forelimbs and two hind limbs, and the robot body 40 is installed in a dog shape. The forefoot installed in the first half of the robot main body 40 is provided with a walking driving part 45 for generating a walking driving force to walk based on the robot control signal from the system control part 1, and the working driving part 45 is walking. It is associated with the hind limb to operate in conjunction with the hind limb in the city. The rear leg installed in the rear half of the robot main body 40 is provided with a rear joint driving portion 47 which bends so that its rear knee is bent almost simultaneously based on the robot control signal from the system controller 1. Here, the rear joint driving unit 47, the walking driving unit 45 and the pastoral driving unit 43 are accommodated in the robot body 40 is installed. And, on the upper side of the rear half of the robot body 40, that is, the upper side of the rear joint driving unit 47, a tail turning unit 49 for rotating the tail from side to side based on the robot control signal from the system control unit 1 is installed. It is. Each of the forelimbs has a front joint driving unit 51 for driving the knee joint and the ankle joint of the forelimb.

The pastoral drive unit 43 guides the rotational motion of the first motor driving unit M1 and the first motor driving unit M1 to a linear reciprocating motion to generate a rotational motion having a predetermined ratio by the robot control signal. The rotating cam 53, the support bar 55 to perform a linear motion by the reciprocating motion guided by the rotating cam 53, and the support bar 55 is connected to the robot body according to the linear motion ( 40 is provided with a rotating plate 59 that rotates about a rotation pin 57 provided to be fixed to one side. It can be shown in Table 2 as follows.

Composition water Contents Characteristic Reference structure Head, Mouth-turning Plate (59) Lifelike plastic structure Support Bar (55) Areas that support the movement of the head and mouth Rotary Cam (53) Role of converting rotational motion into two-dimensional shaking motion (rotation of eccentric shaft cam) Motor and Gear Box First motor drive unit (M1) Nodding and opening and closing the mouth at the same time.

The walking driving unit 45 is to allow the operation of the forelegs and the hind limbs to be approached at the same time as the movement of the forearm coupled to both shoulders 60 for the walking of the dog, having a predetermined ratio by the robot control signal A third motor driving unit (M3) for generating a rotational movement, a pair of lifting cams (63) for moving up and down the shoulders of the front leg based on the rotational movement of the third motor driving unit (M3), and installed on the front leg And a lifting guide part 65 for guiding the lifting movement in the height direction, and connected between one side of the lifting guide part 65 and the shaft of the lifting cam 63, and the front leg forward by a predetermined circumferential angle. Rotation induction pin 67 to induce rotation, and the power transmission bar 69 is installed on one side of the forelimb and one side of the hind limb to transmit the hind limb to move backward as the forelimb moves forward. Have.

The front joint driving unit 51 is connected to the walking driving unit 45, as shown in Figures 8 to 10, is formed integrally with the lifting guide portion 65 of the walking driving unit 45 to accommodate a predetermined A rectangular shoulder connection portion 70 having a space, a front leg portion 73 provided at the end of the shoulder portion 70, and a front surface contact portion 75 provided at the end of the front leg portion 73 to be in contact with the ground. And, the front limb joint (77) for the rotation of the front limbs (73) by coupling between the forelimbs (73) and the shoulder connecting portion (70), the forelimbs (73) and the front contact portion (75) It has a front ankle joint (79) for the rotation of the front contact portion 75 by coupling between, and in the shoulder connecting portion 70 in accordance with the robot control signal in front of the limbs (73) and the front At least one of the ground contact portions 75 In order to rotate at a circumferential angle, at least one of the forefoot joint 79 and the front elbow joint 77 is provided with a fourth motor driving unit M4 for supplying a driving force for the rotation. The fourth motor driving unit M4 is a front of the multi-stage that simultaneously rotates the motor 80 operating at a predetermined rotation speed and the front support leg 73 by the rotational force of the motor, based on the robot control signal. It has a joint gear portion 81 and a rope 83 connected to the front joint gear portion 81 to rotate the front contact portion 75. The front ankle joint 79 is provided with a resilient elastic member 85 for restoring the front contact portion 75 that is rotated by the rope (83). The front joint driving unit 51 and the walking driving unit 45 may be shown in Table 3 as follows.

Composition water Contents Characteristic Reference structure Front Contact (75) Lifelike plastic structure Forelegs Branch (73) Lifelike plastic structure Shoulder connection (70) Lifelike plastic structure Power Transmission Bars (69) Transmission of shoulder drive motor to hind legs joint Forefoot Joint (79) Interlocking movement is made by rope () and restoring spring connected to knee joint Anterior Knee Joint (77) 180 degree rotation (any angle function) Lifting Cam (63) 1) 360-degree rotational movement (rotation of eccentric shaft cam) 2) Movement up and down 5Cm (guided by guider) 3) Front leg is rotating and vertical movement 4) Power transmission bar: Power transmission. Motor and Gear Box Third motor drive unit (M3) Bidirectional Rotation Movement (Walking Forward and Backward) 4th motor drive (M4) Anti-rotation movement (bidirectional) (Knee straight and bent) connector Rotary guide pin (67) Prevents separation from the forelimbs and body sensor Contact sensor Sitting posture judgment

The tail pivot 49 has a second motor driver M2 for generating a rotational motion having a predetermined ratio by the robot control signal, and a rotational motion of the second motor driver M2 with a predetermined circumferential angle. It has a rotation cam 87 to be converted to the movement, and the tail member 89 to rotate by the circumferential angle by the rotational movement of the rotation cam 87, the tail member 89 is actually a dog's tail shape It has a predetermined length to be formed, and is composed of a shape deformation material so that the user can bend in a desired shape. Between the tail member 89 and the rotation cam 87 is provided with a rotation support portion 91 for transmitting the rotation movement to the tail member (89). It can be shown in Table 4 as follows.

Composition water Contents Characteristic Reference structure Tail member Mock-up plastic structure Pivotal support Area supporting the movement of the tail Homing Cam Role of converting rotational motion into two-dimensional shaking motion Motor and Gear Box 2nd motor drive part (M2) Rock the tail from side to side or back and forth

The rear joint driving unit 47 is to bend the joint of the rear leg installed in the rear half of the robot body 40 to take a posture, and at least one of the forward rotation and the reverse rotation based on the robot control signal A fifth motor driving unit M5 driven by a predetermined rotation angle, a power transmission gear unit 93 for transmitting rotational force from the fifth motor driving unit M5 to the pair of hind legs, and the power transmission gear A seating cam 95 and a seating cam 95 provided at an end portion of a shaft provided on a rotation shaft of the part 93 to guide the rotational movement in a direction in which the rear half of the robot body 40 sits downward. The hip support part 97 and the lower support part of the hip support part 97 are coupled to each other so as to be relatively rotatable with each other and the upper end of the rear leg at a predetermined distance from the rotational axis of the rear joint 99. On of Has a rear leg supporting part 100, a rear surface contact portion 101 in contact with the ground is connected to the lower end of said rear leg support portion 100 which rotates by a predetermined wonjugak. A rear ankle joint 103 is installed between the rear leg support part 100 and the back ground contact part 101 to assist the rotation of the back ground contact part 101. Between the one side of the rear joint 99 and the robot body 40 located above the power transmission gear 93, the rear joint 99 and the rear leg 100 and the back contact ( 101 is prevented from separation from the robot body 40 and the separation prevention portion 105 for assisting the bending operation of the rear joint 99 is coupled to each other. The power transmission bar 69 of the front joint driving unit 51 is connected to one side of the hip support 97 of the separation space between the seating cam 95 and the detachment preventing unit 105 to perform a walking operation. This can be shown in Table 5.

Composition water Contents Characteristic Reference structure Back ground contact part (101) Lifelike plastic structure Rear Leg Branch (100) Lifelike plastic structure Hip Supports (97) Lifelike plastic structure Power Transmission Gear Parts (93) Delivering motor rotation to both rear legs Departure prevention part (105) The goal is to drive the entire hind limb with only one motor with a stick to induce ankle-knee movement. joint Back Ankle Joint (103) The linkage is performed by the string connected to the knee joint and the restoring spring. Posterior knee joint (99) 90 degree rotation Sitting Cam (95) 1) 180 degree rotational movement (rotation of eccentric shaft cam) 2) Power transmission bar (): Receive power of front leg. Motowa Gearbox 5th motor drive (M5) In both directions. Half-turn rotation (sitting and standing) sensor Contact sensor Judging state by top and bottom judgment Contact sensor Determine sitting position (installed on one leg only)

Definitions of the first to fifth motor driving units M1, M2, M3, M4, and M5 may be shown in Table 6 below.

Motor Drive Sign Mounting position Motor type Motor driver Action M4 Upper left leg DC motor + A type gear TA 7279P Standing, sitting, stopping M4 Upper right leg DC motor + A type gear TA 7279P Standing, sitting, stopping M5 Hip left side DC motor + A type gear TA 7279P Standing, sitting, stopping M3 Right chest side DC motor + B type gear TA 7279P Walk forward, backward, stop M2 Tail DC motor + A type gear TA 7279P Tail shaking, stop M1 Neck area DC motor + A type gear TA 7279P Open mouth.

Operation for the voice recognition robot system as described above is as follows. 15 is a schematic system diagram showing an operation flow of the voice recognition robot toy according to the present invention, Figure 16 is a system flow diagram showing the operation flow from the sensing of the robot mechanism unit to the motor drive unit, Figure 17 Operation flow chart showing the calibration operation of the voice recognition robot toy according to the present invention.

First, as shown in FIG. 15, when power is supplied (S1), the system controller 1 detects the initial mode from the sensor of the robot mechanism unit 5 (S2), and determines the posture (S3). If the standing posture is not a standing posture, a sitting posture, or a prone posture, which is a basic posture, the calibration mode is performed to become the basic posture (S4). At this time, the posture other than the basic posture is a posture stopped during the walking operation, a forefoot posture, and the like and a posture that is formed by other users. If the operation by the calibration mode is not performed, the posture is determined to be an undefined posture (S5), and the display returns to the initial mode (S2), and the posture attitude is determined again (S3). When the operation in the calibration mode is performed, the system control unit 1 waits for the user's voice command to be input (S6), and when the command is input (S8), operates the execution mode so that the user desires the operation. A robot control signal is generated to the robot mechanism unit 5 so as to take different motions and postures that can be derived from the basic posture (S9).

In more detail, as shown in FIG. 16, when the power is first supplied or the name of the toy robot is called, the system controller 1 checks the sensor of the robot mechanism unit 5 to determine the initial mode. It is detected (S10). The initial state decision program synthesizes the information input through the 12 sensors to determine the posture of the initial state and transfers the result to the correction mode (S11). If the posture is in an undefined state after calibration, it immediately induces the operation of the motor according to the specified algorithm and returns to the initial mode, and the operation is repeated until proper calibration is made. In this case, the calibration means a method of inducing an operation in one of the basic states, the standing posture, the sit posture, and the lie flat posture so that it is easy to move to the requested state when a voice command is issued. . When one of the basic states is reached, a standby state is entered and a voice command is waited (S13). The voice recognition unit 19 performs the operation according to the algorithm of the play mode and the instruction of the system controller 1 according to the voice command recognized. If an undefined state occurs after the command is executed, it returns to the initial mode again and repeats the above operation. When the power supply voltage is lower than 2.8V during the operation, the operation is forcibly stopped and the low voltage state is warned to facilitate the flow of the status mode. When the system is started for the first time, it is divided into a start by a power supply and a start by a user's voice command.

Then, after the initial state is detected by the sensor and the method is not in the basic state, the calibration mode method predicts the initial situation that can occur to make various initial states defined. Algorithms should be prepared to correct this by the motor operation to the basic state. In particular, an undefined state is defined to drive all the motors in the worst case to reach the basic state. The input signal for the basic posture is shown in Table 7 below.

division Normal input signal Other Sitting position S2, S6, S10, S4, S8, S11 S2, S6 and S10 are basic signals for stop operation Standing posture S2, S6, S10, S4, S8, S12 S4, S8, S11 are normal input signals of sitting motion Prone S2, S6, S10, S3, S7, S11 S3, S8 and S12 are normal input signals

As a result, the system controller generates a robot control signal to the robot mechanism unit 5 so that a normal basic posture as shown in Table 6 is achieved before the voice command of the user is performed.

In addition to the basic posture and other postures that can be operated, the abnormal input signal is shown in Table 8 below, and it can be determined from the abnormal signal as the posture of the initial state. The system control unit 1, which determines the posture of the initial state, operates the first to fifth motor driving units M1, M2, M3, M4, and M5 of the robot mechanism unit 5 to correct the basic position.

Initial posture Calibration method Error input signal Default posture (final state) No. Motor signal Sit right foot up M4 + S7 Sitting One Sit left foot M4 + S3 Sitting 2 Stand up right foot M4 + S7 clerk 3 Stand up and raise left foot M4 + S3 clerk 4 Not straight M4 + S1 and S5, and S9 clerk 5 Posture in an undefined state M4 + S3 M4 (stop), clerk 6 M4 + S7 M4 (stop), clerk 7 M2 + S9 M5 (Pause), Clerk 8 M1 + S1, S5, S9 M3 (Pause), Clerk 9

If the third motor driving unit (M3) is made during walking, the information of one leg can be known if the other three legs are also in a situation. The sitting or standing motion should be made, but in reality this is impossible, so every time a sitting or standing motion is induced, a slight deviation occurs in the original walking direction. The correction method is to find the initial state by the sensor and place the two legs completely in place, then stand up. If this is not corrected, the direction of travel will change when you stand up. It must be calibrated by a combination with the sensor.

After the calibration operation as described above is performed, the system controller 1 prepares to receive a user voice command, as shown in FIG. When the user inputs a desired command, that is, a command (walking, standing, sitting, front hoof, or lying down) that causes a dog's motion or sound, such as a robot toy, to the first microphone 6 and the second microphone 7, the system control unit (1) recognizes a command by extracting only the voice of the user, and controls the robot mechanism unit 5 according to the recognition.

For example, when the user's voice command comes in, the robot mechanism part 5 taking the basic position is driven by a transition operation corresponding to the voice command to take a posture or action desired by the user.

That is, when the dog, which is a toy robot, is sitting, when the user commands the "right foot", the system control unit 1 recognizes the voice command and judges it as "right foot" in the current state of the dog sitting. The driving command is input to the fourth motor driving unit M4 of the robot mechanism unit 5 so as to lift the right foot. The puppy's motion mechanism is the same when inducing other actions to be performed, as shown in Table 9 below.

Initial state Command How To Final sensor input signal Motor signal Sitting position Sit right foot up M4 - S2, S6, S10, S4, S7, S11 Sit left foot M4 - S2, S6, S10, S3, S4, S11 Standing up Walking forward M3 + S1 (S2), S5 (S6), S9 (S10) Walking back M3 - S1 (S2), S5 (S6), S9 (S10) Stand up right foot M4 - S2, S6, S10, S4, S7, S12 Stand up and raise left foot M4 - S2, S6, S10, S3, S4, S12 Sit right foot sit M4 + S2, S6, S10, S4, S8, S11 Sit in one's left foot M4 + S2, S6, S10, S4, S8, S11 Walking forward stand up M3 s S2, S6, S10, S4, S8, S12 Walking posture M3 s S2, S6, S10, S4, S8, S11 Stand up M4 + S2, S6, S10, S4, S8, S11 Stand up M4 + S2, S6, S10, S4, S8, S11

Eventually, before entering the operation according to the voice command, the system first finds out the initial state and enters the corrective operation in the basic state. Here, the corrective operation means that the basic position is placed in an ideal state for the operation before performing the operation by the voice signal. It depends on the initial state and depends on a predefined algorithm to move to which basic posture. One of the basic postures, once corrected, performs a voice command. The order of operations performed by the robot mechanism unit 5 and the system control unit 1 is an initial state ⇒ a calibration operation ⇒ a basic state ⇒ a voice command () ⇒ an initial state.

If an unrecognized condition occurs, the motor is operated to forcibly move to the initial mode. The method is based on a separate algorithm. Although various other operations can be implemented, the algorithm is basically configured to enable a total of six operations.

In order to produce a desired action, it must go through a series of flows from the initial state. Some actions can be performed as desired after passing up to three or more steps depending on the initial state. For example, considering the initial state of prone state, this is a requirement that the most complicated steps can be configured up to three operations. In other words, it refers to the execution of lying posture ⇒ sitting motion ⇒ standing motion ⇒ walking motion. If the flow of the step by step is grasped, the flow can be naturally grasped since the whole flow is also part of the element. The flow of such an operation and its reverse operation can be shown in Tables 10 and 11.

Action sensor Drive motor Motor drive signal Other kneel down S2, S6, S10, S3, S7, S11 No activity No activity - kneel down Sit S2, S6, S10, S3, S7, S11 M4 + Forelegs M4 + Forelegs Sit get up S2, S6, S10, S4, S8, S11 M5 + Hind legs get up walk S2, S6, S10, S4, S8, S12 M3 + or- Up and down movement

Action sensor Drive motor Motor drive signal Other walk S1 (S2), S5 (S6), S9 (S10) M3 + or- Up and down movement walk get up S1 (S2), S5 (S6), S9 (S10) M3 S Fixed to the bottom get up walk S2, S6, S10, S4, S8, S12 M5 - Hind Leg Bending Sit kneel down S2, S6, S10, S3, S7, S11 M4 - Foreleg bending M4 - Abdominal bending

Meanwhile, the process of recognizing the voice command of the user in the system controller 1 will be described.

That is, the two first and second microphones 6 and 7 receive the user's voice signal and convert it into an electric signal, thereby converting the analog voice signal of the A / DD / A converter (Codec) 11 described in the above configuration. Send to. At this time, the two input voice signals are transmitted to the system controller 1 in a form independent of each other to remove noise to perform a correlation operation. The system controller 1 sends a control signal (data input ready signal) to the A / DD / A converter (Codec) 11 to indicate that it is ready to accept it, unless there is a special situation, and the A / DD / A converter ( Codec (11) uses 2.048 MHz, which is a value of x256FS for interpolation, and its synchronization frequency (SYNC Frequecy) is 8 kHz, and a sampling rate for improving speech recognition in the speech recognition unit 19 is provided. rate). In particular, the 8 kHz sampling rate has become an important processing criterion for the recognition algorithm in the speech recognition unit 19 of the system controller 1. On the other hand, the input voice signal is sent to the system control unit 1 by A / DD / A converter (Codec) 11 to perform A / D conversion, and the independent data is transmitted through the first and second microphones 6 and 7. The noise inputted by the correlation operation is filtered.

The digital voice sampling signal from which the noise is removed is temporarily stored in the circular buffer 17 in units of frames, and then the eigenvalues of the user voices of the pre-emphasis 39 and the speech recognition unit 29 are determined. Calculated as feature vector data, the crossword passes through the zero crossing plate 31, the energy calculation unit 33, and the unit speech detection unit 35 at the same time to detect each word end point. Words are divided into words for speech recognition. Then, when the list controller 25 extracts the compressed data of the first memory unit 13 corresponding to the voice recognition word from the preprocessor 37, the data and beater are transferred to the second memory unit 15. Analyze non-algorithms and perform associative actions for recognition.

In other words, the operation actually consists of a speech signal sampled at 8 kbps-> preprocessing (speech feature detection)-> Speech Detection-> Speech Recogniton. Pre-processing is performed through the computational steps of Power, Hamming Window, Preemphasis, etc. Calculate Separately, zero crossing rate 31 and power energy 33 are calculated from the voice to detect the start and end points of the voice. Based on the results of these two voice detections, it is decided whether to start or stop the voice recognition or reset, and finally recognize the voice by using Viterbi algorithm for Melscale's Cepstrum coefficient sequence and HMM. Of course, the constants necessary to perform such a large number of calculations are stored in the first memory unit 13 and used whenever necessary. In addition, the second memory unit 15 should be used for the task of calculating and recording the necessary value, and the List Controller 25 is used due to the vastness of the data calculation. Here, the end point detection of the voice for speech recognition and compression is performed in the unit speech detection unit 35 used to increase the recognition rate and the compression rate.

However, the zerocrossing rate 31 and the power energy 33 show high efficiency and hit rate in a laboratory or relatively quiet room, but are fundamental problems in speech endpoint detection that responds to small noises. It is true that it has to be operated with Mel Cepstrum.

In other words, a sampling signal mixed with voice, noise, and silence is input to the unit voice detection unit by obtaining a power energy unit, a zerocrossing rate, and a mel scale cepstrum. If you put it in, the voice (mixed noise) part comes out. The results from the two modules are sent to the preprocessor 37 to recognize the voice signal.

As such, when the voice recognition unit 19 recognizes the user's voice, the system controller recognizes the sensing signals output from at least eight sensors of the robot mechanism unit 5 to determine the current posture of the robot, and After determining whether the current posture and the voice command of the user are the same, for example, the robot control signal is generated so that the current posture is performed by the walking operation.

In addition, in response to the recognized voice command, that is, singing (generating a melody selected as a dog sound) and barking are input, the system controller 1 fetches the compressed melody sound from the first memory unit 13. Expand and restore the signal, and amplify the restored signal by a predetermined magnitude gain in the power amplifier 10 to output to the speaker 8.

Then, such a corrective operation is terminated and the user's voice command for executing the operation, that is, the walking operation, the sitting operation, the standing operation, the tail shaking operation, the head shaking operation, and the forefoot extruding operations are performed by the first and second microphones 6, 7) is entered as follows. 5 is an operation cross-sectional view showing only the front and rear legs corresponding to the standing posture of the voice recognition robot toy according to the present invention, Figure 6 shows only the front and rear legs corresponding to the walking posture of the voice recognition robot toy according to the present invention. 7 is an operation cross-sectional view showing only the front and rear legs corresponding to the sitting position of the voice recognition robot toy according to the present invention, Figures 11 to 12 is a neck operation of the voice recognition robot toy according to the present invention 13 is a cross-sectional view showing a tail operation of the robot toy for speech recognition according to the present invention.

As shown in FIGS. 5 and 6, the walking operation moves four legs at once by the third motor driving unit M3 and the power transmission bar 69 attached to the right side of the chest. When the robot control signal is applied to the third motor driving unit M3 from the system control unit 1, the shoulder connecting portion of the front joint driving unit 51 by the elevating cam 63 subjected to the rotational force of the third motor driving unit M3. 70 is moved up and down along the lifting guide portion 65, and at the same time the shoulder connecting portion 70 is rotated toward the front and rear by the pivot guide pin 67 to perform the walking operation as a whole. At this time, the walking is performed as the right forefoot and the left forefoot alternately step forward. In fact, all joints except for the knee joint of the forelimbs should be involved in this movement, so the hind legs are also involved in the walking movement. Here, when rotated by the distance of the front and rear movement along the shoulder connecting portion 69 rotated by the elevating cam (63), the power transmission bar 69 is the rear joint driving portion 47 by the front and rear movement distance. Hip support 97 of the) to move backward and forward respectively to perform the overall walking. That is, the power transmission bar 69 performs the front and rear movement of the right front leg and the left rear leg together, and performs the front and rear movement of the left front leg and the right rear leg together.

As shown in FIG. 10, the standing operation is performed by two fourth motor driving units M4 of the front legs. In order to achieve the standing position, the ankle joint must be properly positioned, and the driving method is semi-automatically determined by the situation of the knee joint. The ankle joint is made of a restoring elastic member 85 and a rope 83, the restoring elastic member 85 provides a restoring force such that the front contact portion 75 and the front leg support portion 73 are in a straight line, The rope 83 folds the ankle joint such that the front contact portion 75 and the front limb portion 73 are vertically formed by the front joint gear portion 81.

As shown in FIG. 7, the seating motion is in a standing posture with the front legs standing and the back legs bent in an overall sitting position. This is the normal behavior of the dog, which is distinguished from its lying on its forelimbs. The sitting position is induced in a smooth motion by the fifth motor driving unit M5 mounted on the left side of the hip and the separation prevention unit 105. That is, the sitting motion is different from the front knee joint 77 of the front joint driving unit 51, the seating cam 95 is rotated by one fifth motor driving unit M5 and the power transmission gear unit 93, power transmission A seating cam 95 having another shaft spaced apart from the shaft of the gear portion 93 on one side of the hip support portion 97 rotates the release preventing portion 105 to move the hip support portion 97 back to the knee joint 99. When folded in), the sitting motion is naturally made. At this time, the power transmission bar 69 does not transmit power to the front joint driving unit (51).

As shown in FIGS. 11 and 12, the head shaking operation includes a rotary cam 53 for converting the rotational movement of the first motor driving unit M1 into a linear movement, a support bar 55 for linear movement, and the support. It is made by the rotating plate 59 which forms the head of the dog by the bar 55. Then, the pivoting plate 59 nods back and forth around the pivoting pin 57 because the length variation of the support bar 55 changes in the height direction by the eccentric separation distance of the rotary cam 53.

The tail swinging operation is performed as shown in FIGS. 13 and 14, which is substantially the same as the head shaking operation, and simply rotates the rotational motion of the second motor driver M2 to the left and right rotational motion 87. And, by the tail member 89 and the rotation support portion 91 for transmitting the power of the rotation cam 87 to the tail member (87).

As shown in FIG. 10, the front foot pushing operation may be performed by lifting the front foot by rotating the front leg support 73 by the fourth motor driving unit M4 and the front joint gear unit 81. This can be done with the front foot in the standing state or the front foot in the sitting state.

The operation method of the voice recognition robot toy according to the present invention as described above is as shown in FIG.

First of all, there are two types of dog's action and security features. The security function is put in consideration of the characteristics of the dog to protect the house, and the security mode functions only when the security switch is in operation. Recognition of any external noise through the first US second microphones 6 and 7 responds with an already programmed algorithm.

Its main functions include many functions such as sitting motion, standing motion, down motion, walking motion (forward and backward motion), forefoot motion (right and left foot motion), barking motion and dog ring melody sound. The function is a sensor of the robot mechanism part (5) and the working driving unit 45, the front joint driving unit 51, the rear joint driving unit 47, the pastoral driving unit 43 and the tail rotating unit (49) ), And achieves the user's desired action (predetermined motion) in a continuous motion in a basic posture. In this case, the algorithm is also configured to follow the given voice command while following the change using a sensor.

As described above, all movements start with standing, sitting, and lying down, which is the basic posture, which is always detected by the initial mode and leads to this mode. Of course, this mode is executed even when the switch is operated, but if it is not defined, the correct operation is immediately performed and the command is executed.

As an example, consider a stand up command / standing command / weather command. Recognition of the user's command in these three initial situations will result in an immediate correction in case of undefined situations and take one of the three basic postures. Stand up and perform commands / standing commands / weather commands. In other words, if the basic posture is standing, it stops, if it is sitting, it stands up, and if it is lying down, it stands up and then stands up. Here, after such a stand-up operation is performed, the tail is shaken for a predetermined time and the standby state is entered for the next operation.

Sit command, stop command, bark command, forefoot command, prone command, walk command and sing command are performed through the same process. At this time, if there is no special command for a long time, when standing, shaking the tail and entering the standby state after walking backwards, when sitting, shaking the tail and entering the waiting state after lying down, immediately lying down, After shaking its tail, it enters the standby state.

In the case of the song command, the user selects one of a plurality of melodies which are stored randomly in the basic posture and tells the user to enter the waiting state for the next command.

The security function barks fiercely for a predetermined time after taking a basic posture when listening to a random sound, and then waits after making a rough breath and then barks fiercely for a predetermined time when a popular chuck occurs again.

According to the present invention, the user's voice command is speech-independent and the user performs the operation to generate the appropriate dog action and dog sound according to the corresponding robot mechanism (doggy mechanism), and the user has a price competitiveness and operation reliability. Maximizes the interest and interest of, and does not get tired of using it has the effect of improving the life as a toy.

Claims (14)

In the toy robot for voice recognition following the user's voice; Memory means for compressing and storing a plurality of voice patterns for recognizing a voice signal of the user, and at least one melody sound having a predetermined pitch and time signature; A system control unit for recognizing the voice of the user from the voice pattern of the memory means and outputting a robot control signal having a predetermined operation pattern according to the voice command; And a robot mechanism part operating in at least one pattern based on the robot control signal of the system control part. The method of claim 1; And the system controller extends the compressed melody sound of the memory means based on the recognized voice of the user and outputs the melody sound signal. The method of claim 2; A first microphone for converting the user's voice and outside noise into an electrical signal and outputting the electrical signal to the system controller; A second microphone for converting the outside noise into an electrical signal and outputting the electrical signal to the system controller; The voice recognition robot toy, characterized in that it further comprises a sound generator for hearing the user based on the melody sound signal from the system control unit. The method of claim 3; An A / D D / A converter is provided between the system control unit and the microphone to convert one side of an analog audio signal and a digital audio signal to the other side. The method of claim 1; The memory means, A first memory unit for storing melody sound compression data in which a plurality of melody sounds of digital sound signals are compressed at a predetermined compression rate and a plurality of speech patterns for speech recognition; A speech recognition robot toy, comprising: a second memory unit configured to store a robot algorithm for performing a doggy operation, wherein a computing space is provided for recognizing the voice signal of the user input from the outside. The method of claim 1; The system control unit, A circular buffer for temporarily storing a digital audio signal of a user in units of frames output from the first microphone; A voice recognition unit for dividing the digital voice signal stored in the circular buffer into a voice recognition word according to the voice recognition constant of the compressed data stored in the memory means and recognizing the user's voice with a Viterbi algorithm; A robot control unit controlling a robot mechanism unit according to the robot algorithm for a user's command recognized by the voice recognition unit; Voice recognition robot toy, characterized in that it comprises a sound decoder to extend and restore the melody sound compression data having the dog sound to the sound generator. The method of claim 1; The robot mechanism unit, Working driving means for generating a walking driving force to at least one pair of four legs so that the four legs work based on the robot control signal from the system control unit; Front joint driving means for driving a bent joint of at least one of the two forelimbs based on the robot control signal from the system controller; Rear joint driving means for driving the rear leg joint so that the two rear leg joints are bent at the same time based on the robot control signal from the system controller; And a robot body accommodating the rear joint driving means, the front joint driving means and the working joint driving means and having four legs installed therein. The method of claim 7; Pasting means for rotating the head up and down based on the robot control signal from the system controller; The robot toy for voice recognition further comprising a tail rotating means for rotating the tail to the right and left based on the robot control signal from the system control unit. The method of claim 8; The pastoral movement means, A first motor driver for generating a rotational motion having a predetermined ratio by the robot control signal; A rotary cam for converting the rotational motion of the first motor driver into a linear motion; A support bar linearly moving by the rotary cam; Voice recognition, characterized in that it comprises a rotary plate connected to the support bar to rotate around the rotation pin installed to be fixed to one side of the robot body in accordance with the linear motion The method of claim 8; The tail rotation means, A second motor driver for generating a rotational motion having a predetermined ratio by the robot control signal; A rotation cam means for converting the rotational motion of the second motor driver into a rotational motion having a predetermined circumferential angle; A tail member rotating by the circumferential angle by the rotational movement; A guide bar connecting the rotation cam means and the tail member to guide the rotation force; A voice recognition robot toy, characterized in that it comprises a support member for supporting the rotational force of the guide bar. The method of claim 8; The walking driving means, A third motor driver configured to generate a rotational motion having a predetermined ratio by the robot control signal; A pair of elevating cams for elevating and moving the shoulders of the front legs based on the rotational motion of the third motor driver; An elevating guide means installed on the front leg to guide the elevating movement; A rotation induction pin connected between one side of the elevating guide means and the shaft of the elevating cam and inducing the forward leg forward by a predetermined circumferential angle; A voice recognition robot toy, characterized in that it comprises a power transmission bar installed on one side of the forelimb and one side of the forelimb so that the forelimb and the hind limb are mutually accessible. The method of claim 7; The front joint driving means is connected to the working driving means; A rectangular shoulder portion integrally formed with the lifting guide means of the walking driving means; A foreleg support portion provided at the end of the shoulder portion; A front surface contact part provided at an end of the front leg support part to be in contact with the ground; A front joint gear portion which rotates each of these between the front contact portion and the front leg portion or between the front leg portion and the shoulder portion by a predetermined circumferential angle; A fourth motor driving unit generating a rotational force to the front joint gear unit; A joint part for bending each of the front leg support part and the front surface contact part according to the rotational force generated from the front joint gear part; Voice recognition robot toy, characterized in that it comprises a rope connected to the front joint gear portion to bend the front contact portion. The method of claim 7; The rear joint driving means, A fifth motor driving unit which drives a predetermined rotation angle in one of pre-purification and reverse rotations based on the robot control signal; A power transmission gear unit for transmitting the rotational force from the fifth motor driving unit to the pair of rear legs; A seating cam provided at an end portion of the shaft provided on the rotation shaft of the power transmission gear unit to guide the rotational force in a direction in which the rear half of the robot sits downward; A butt support part coupled to the rotation axis of the seating cam at a predetermined distance from the upper end of the rear leg to be rotatable relative to each other; A rear leg portion connected to the lower end of the hip support portion and rotated by a predetermined circumferential angle by the rear joint; A rear surface contact portion connected to a lower end of the rear leg support portion and in contact with the ground; One side of the rear elbow joint and the robot body located above the power transmission gear portion are mutually coupled to prevent the separation and separation of the rear elbow portion and the rear leg support and the rear surface contact the robot body and the rear elbow joint Voice recognition robot toy, characterized in that it comprises a departure prevention portion to assist the bending operation of the. A method of controlling a voice recognition robot toy having a robot mechanism; Determining an initial posture of the robot from a sensor indicating an initial state of the robot mechanism; Calibrating with at least one basic posture after determining the initial posture; Recognizing a voice command of the user; And controlling the robot mechanism to follow the robot mechanism in accordance with the voice command based on the recognition.