KR20010007842A - The system and method of a dialogue form voice and multi-sense recognition for a toy - Google Patents

Abstract

PURPOSE: Interactive type audio and multi-sense cognition system in toy satisfies curiosity and interest of children by various vocal expression for identical question of user, improves culture of sentiments and speech study of children by various vocal and motion expression of children sense. CONSTITUTION: Interactive type audio and multi-sense cognition system in toy comprises first step to extract selected vocal sample to study data on audio cognition training process using IDMLP(Input Driven Multi Layer Perceptron) nerve circuit net algorithm through computer; second step to store study data and vocal scenario corresponding to study data in memory part(20) through PC interface part(12) of toys; third step to extract study data corresponding to vocal signal using IDMLP nerve circuit net algorithm after passing whole treatment process in control part(10); fourth step to select vocal expression and motion expression among scenario for vocal output and motion expression through audio and working composition part(11), memory part(20) in control part(10); fifth step to send control signal to output selected vocal expression and motion expression through output part(40); sixth step to output vocal and motion expression by control signal of control part(10).

Description

The system and method of a dialogue form voice and multi-sense recognition for a toy}

The present invention relates to an interactive voice and multi-sensory recognition system and a method thereof for toys, and more particularly, to a variety of voices, actions, and surroundings of a user by embedding a voice recognition and multi-sense recognition system in toys such as dolls and toys. The present invention relates to a voice and multisensory recognition system and a method thereof for communicating with a user implemented in software and hardware to properly respond to a situation.

In general, toys are closely developed with children, satisfying their curiosity, and at the same time, they are constantly being developed in various forms in consideration of educational aspects such as creativity and sensitivity. It has been developed around the characters, but more and more types of toys are being produced reflecting the demand for educational aspects.

In particular, the appropriate voice expressions and movements according to the various voices and actions of the children satisfy the curiosity of the children and the educational effect is very excellent. Most of the toys are developed in this direction. Until now, the technology level has embedded a predetermined storage medium in the toy, and then stores sound and voice expressions to satisfy children's curiosity, and operates the touch sensor to express the predetermined sound and voice. Due to the burden of cost and the limitation of technology, there is a problem that it is difficult to continue the desire for satisfying curiosity due to the characteristics of the young children who are sensitive to the epidemic and severe.

In recent years, in order to solve the above problems all over the world, children's curiosity and educational production considering the educational aspects have been vigorous. In particular, a voice recognition system using voice recognition technology is embedded in the toy for appropriate situations. It outputs the right voice, but it is still in the elementary stage, so the stored voice is simply output, and the voice is output in a more advanced form with a certain scenario, but the output voice is also uniform. It is still a problem that children lose interest when a certain amount of time becomes a voice expression and the educational effect is weakened.

In addition, there is a case in which a predetermined driving means is embedded in toys such as dolls and robots to express a certain movement in order to induce children's interest. There is a problem that it is insufficient to satisfy the desire for.

Therefore, there is a need for a toy that can implement educational functions while satisfying interests by enabling communication with children by expressing voices and actions corresponding to various voices, actions, and surroundings of children.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-mentioned conventional problem, and recognizes a user's voice input through a predetermined input device using an IDMLP neural network algorithm, and among various voice expressions of a voice scenario corresponding to the recognized voice. By randomly selecting one voice expression and outputting it through the corresponding output device, even if the user asks the same question, the toy can give a variety of new answers depending on the situation. The purpose is to add additional functionality.

In addition, the built-in tactile sensor, ultrasonic sensor, infrared sensor in the toy implemented by the present invention detects the tactile touch by the user, detects a certain distance with the user, and detects light and darkness to a predetermined voice By expressing and expressing movements, it aims to help children's emotional development by making them feel the same effect as talking with living creatures.

In order to achieve the above object, the present invention uses the IDMLP neural network algorithm to recognize a variety of voice signals without limitation to the user and the training data extracted through the voice recognition training process and the voice output and operation output corresponding to the learning data The control unit detects a voice signal input by a user and a signal supplied by a predetermined sensor by storing the scenario in a predetermined storage medium embedded in the toy, and the control signal of the controller controls the scenario of the voice scenario stored in the predetermined storage medium. The present invention is to implement an interactive voice and a multi-sensory toy that enables a variety of conversations with a user by selecting one of various voice expressions and outputting the selected voice expression and a predetermined motion expression through a corresponding output device.

1 is a block diagram showing the configuration of an embodiment of the present invention

2 is a flowchart illustrating a voice recognition training process according to the present invention.

3 is a flowchart illustrating a voice recognition process according to the present invention.

4 is an exemplary diagram showing a table of learning data stored on a memory unit according to the present invention.

5 is an exemplary diagram showing a table of voice scenarios stored on a memory unit according to the present invention;

* Description of Signs of Main Parts of Drawings *

PC 10. Control part

11. Voice and operation synthesis unit 12. PC interface unit

13. Power Supply 20. Memory

21. First memory section 22. Second memory section

23. Third memory unit 30. Input unit

31.Microphone 32.Tactile Sensor

33. Infrared sensor 34. Ultrasonic sensor

35. Filter section 36. First amplifier section

40. Output section 41. Second amplifier section

42. Drive means 43. Speaker

44.LED

Hereinafter, the configuration of the present invention will be described in detail with reference to the accompanying drawings, the following terms are defined in consideration of the functions in the present invention, which may vary according to the intention or custom of those skilled in the art, The definition should be made based on the contents throughout the specification.

The present invention enables speech-independent speech recognition without user restriction by voice recognition technique using IDMLP (Input Driven Multi Layer Perceptron) neural network to enable proper voice and motion expression according to the scenario suitable for the situation, and tactile It is equipped with a sensor, a distance sensor and a light sensor to detect various actions of the user and respond appropriately.

After recognizing various voice signals input through the microphone 31 using the microprocessor 10 by IDMLP algorithm-based voice recognition technology, the corresponding learning data stored in the first memory unit 21 through a learning process. Recognizes and transmits a control signal to the voice and operation synthesis unit 11, selects the scenario from the voice output scenarios stored in the second memory unit 22 in the voice and operation synthesis unit 11 and then various voices After randomly selecting one of the expression data, one voice expression data is extracted and implemented through software and hardware to output through the speaker 43 so that the user can have a conversation, and the user repeats the above process. Even if the toys such as dolls and toys implemented by the present invention always outputs a different voice to interest children To, and configured to simultaneously add the educational function.

In addition, by connecting the tactile sensor 32, the light sensor 33, and the distance sensor 34 to the microprocessor 10, the ADC detects a certain action such as a user touching or approaching light and darkness. The microprocessor 10 recognizes and transmits a control signal to the voice and operation synthesis unit 11 through an analog to digital converter, and transmits the control signal to the second memory unit 22 from the voice and operation synthesis unit 11. Induces children's interest by extracting the corresponding motion expression data and implementing them in software and hardware to express a certain motion through the driving means 42. And educational functions.

In order for toys such as dolls and toys according to the present invention to respond appropriately to the situation according to various voices and actions of the user, the toys may remember and recognize the characteristics in advance about the predictable voices that can be heard in the scenario situation. The above training process requires the normal PC (1) to record about 30 children's voices for the same questions and conditions, and then normalizes them through signal difference accumulation method and interval detection. After the preprocessing process, IDMLP neural network algorithm classifies each voice by phoneme, assigns a certain weight to the phoneme, repeats the training process of extracting data, and extracts a predetermined result value. Voice of various children by storing in the memory unit 20 through the interface unit 12 The so when input from the microphone 31, the microprocessor 10 can recognize to retrieve the data.

Looking at the recognition process of the toy is completed through the above training process, when the user's voice is transmitted to the microprocessor 10 through the microphone 31, the microprocessor 10 receives a voice signal and accumulates the signal processing After the preprocessing process, which is a normalization process through law and section detection, the input voice is extracted through the memory unit 20 by the IDMLP neural network algorithm, and the control unit 10 is applied to the recognized learning data. In order to select a corresponding voice and operation scenario, a control signal is transmitted to the voice and operation synthesis unit 11 to select a corresponding scenario, and then an arbitrary voice expression is selected from various voice expressions and operation expressions and output through the speaker 43. And output through the drive means 42 to perform a certain operation expression.

In addition, the user's tactile sense, the user's distance, and the amount of light are sensed by various sensors, and the corresponding signal is transmitted to the microprocessor 10, and the microprocessor 10 recognizes whether it is ON / OFF and the voice and operation synthesis unit ( 11) through the speaker 43 and the drive means 42 outputs the corresponding voice expression and operation expression.

As such, toys and software can be implemented in software and hardware so that users can express various voices and behaviors appropriately according to the situation. It is designed to add educational elements that can be done.

Hereinafter, with reference to the accompanying drawings in the toy constituted by an embodiment of the present invention will be described in detail for the interactive speech and multi-sensory recognition system and method thereof.

1 is a block diagram showing the configuration of an embodiment of the present invention, Figure 2 is a flow chart showing a voice recognition training process according to the present invention, Figure 3 is a flow chart showing a voice recognition process according to the present invention, 4 is an exemplary diagram showing a table of learning data stored on a memory unit according to the present invention, and FIG. 5 is an exemplary diagram showing a table of a voice scenario stored on the memory unit according to the present invention.

1 is a block diagram illustrating a configuration according to an embodiment of the present invention.

As shown, the present invention includes an input unit 30, a control unit 10, a memory unit 20, a voice and operation synthesis unit 11, an output unit 40, a power supply unit 13, and a PC interface unit 12. The input unit 30 includes a microphone 31, a tactile sensor 32, an ultrasonic sensor 34, an infrared sensor 33, a filter unit 35, and a first amplifier unit 36. The memory unit 20 includes a first memory unit 21, a second memory unit 22, and a third memory unit 23, and the output unit 40 includes a second amplifier 41 and a drive unit. It consists of a means part 42, a speaker 43, and a light emitting diode (LED) 44.

The controller 10 controls the system using the one-chip microprocessor Intel 80C196KC, which converts an analog signal received through the input unit 30 into a digital signal and compares the extracted voice data through a learning process. To recognize the user's voice, and to decode the voice data stored in the predetermined compression form in the first memory unit 21 to compare the input voice data in the third memory unit 23. Search for the corresponding voice data, receive signals input through various sensors, recognize ON / OFF status, and also perform memory unit 20, voice and operation synthesis unit 11, input unit 30, and output unit ( The control unit 10 collectively controls the control signal by sending control signals to the surrounding components such as 40 and the power supply unit 13.

The memory unit 20 is composed of a first memory unit 21, a second memory unit 22, and a third memory unit 23. The first memory unit 21 uses the 32K byte EPROM. Stores the operating software to operate the control unit 10, and stores the learning data obtained through the learning process using a IDMLP neural network algorithm in a predetermined compressed form to receive a random voice signal to the control unit 10 If so, the voice data may be extracted in association with the learning data of the first memory unit 21.

The second memory unit 22 uses a microphone 31 to store voice scenario data suitable for each situation in a predetermined compression form so that a toy can express a predetermined voice and an operation using a 256K byte EPROM. When the ON / OFF signal is received through the voice signal and the sensor of the control unit 10, in the case of the voice signal in conjunction with the first memory unit 21 to extract the corresponding voice data to the second memory unit 22 The controller 10 randomly selects one of the stored voice scenario data and outputs it through the voice and operation synthesizing unit 11, and when the ON / OFF signal is provided through various sensors, the controller 10 A predetermined voice expression and an operation expression stored in the second memory unit 22 are selected to be output through the voice and operation synthesizing unit 11.

The third memory unit 23 is an element in charge of internal data signal processing using a 32K byte RAM, and serves as a use space for recognizing a voice signal input through the microphone 31.

The voice and operation synthesis unit 11 receives a control signal from the control unit 10 and selects a voice expression and an operation expression scenario stored in the second memory unit 22 and then selects the control unit 10 from the corresponding scenarios. Any voice expression and motion expression are extracted and transmitted to the speaker 43 and the driving means 42 to output a predetermined voice and operation.

The input unit 30 includes a microphone 31, a tactile sensor 32, an infrared sensor 33, an ultrasonic sensor 34, a filter unit 35, and a first amplifier unit 36. 31) converts a sound pressure, which is acoustic energy, into electrical energy after receiving the user's voice, and supplies an analog voice signal to the controller 10, and the tactile sensor 32 detects the touch when the user touches the toy. By supplying an ON / OFF signal to the control unit 10, the ultrasonic sensor 34 detects a predetermined distance of the user and supplies an ON / OFF signal to the control unit 10, the infrared sensor 33 is When the surroundings are dark or bright, light and dark are recognized and the ON / OFF signal is supplied to the controller 10, and the filter unit 35 uses a low pass filter (LPF) to use a microphone ( 31) and the control unit 10 by passing only the low-pass portion of the signal input through the various sensors To filter the noise signal mixed with the voice signal so as to be easily recognized, and to remove the constant noise due to the user's contact or ambient noise, the first amplifier 36 using an operational amplifier (OP AMP) The control unit 10 amplifies the signal to a predetermined size so that the control unit 10 recognizes and operates the weak analog signal supplied by the filter unit 35.

The output unit 40 includes a second amplifier 41, a driving means 42, a speaker 43, and a light emitting diode (LED) 44. The second amplifier 41 is an input unit. The driving means unit drives the weak analog audio signal and the operation signal supplied through the voice and operation synthesizing unit 11 using an operational amplifier (OP AMP) similarly to the first amplifier unit 36 of 30. (42), the speaker 43 and the LED 44 to amplify the signal to a predetermined size to operate, the speaker 43 has a voice expression data supplied through the speech recognition process from the control unit 10 Converts an electrical signal into a sound vibration and outputs it as a voice that a user can recognize, and the LED 44 emits light at a predetermined interval by a control signal of the controller 10 using a light emitting diode, The driving means unit 42 is composed of various driving devices to receive the control signal of the control unit 10 And the devices so as to output a predetermined operation according to the control signal.

The PC interface unit 12 easily accesses a general PC which derives predetermined learning data by performing a learning process and prevents a bus collision with signals transmitted from the control unit 10 in receiving the corresponding data. Accessible device that has input and output ports.

The power supply unit 13 maintains an arbitrary voltage with a voltage change at a predetermined voltage at which the toy implemented by the present invention can operate, so that a certain amount of power can be supplied to each circuit and the device. In view of the characteristics, it is safe to use a predetermined battery for easy movement.

2 is a flowchart illustrating a voice recognition training process according to the present invention.

In order to recognize the voices of various users, the toys implemented by the present invention should generate predetermined learning data by assigning each weight to each phoneme extracted through a predetermined training process. Use) to work.

In order to enable speaker-independent speech recognition, voices of about 30 local children are sampled (100), recorded through a predetermined input device, and normalized by signal difference accumulation method and interval detection for each voice signal. After the preprocessing process (110) using the IDMLP neural network algorithm to classify each phoneme of the voice, extract the feature points of each phoneme (120) weighted (130) and store in a predetermined storage medium 140, a series of voice recognition training processes are repeatedly performed (150) to extract common result values to recognize voices of various users.

When the result value for the predetermined voice recognition is extracted through the general PC 1 through the above process, the corresponding data is stored in the first memory unit 21 through the PC interface unit 12 of the system according to the present invention. Make it recognizable.

3 is a flowchart illustrating a voice recognition process according to the present invention.

After the training data is stored in the first memory unit 21 through a training process for speech recognition as shown in FIG. 2, the user uses a microphone 31 which is an input unit 30 of a toy implemented by the present invention. When a certain voice is input (200), the filter unit 35 filters a predetermined noise signal (210), and the first amplifier 36 controls the voice signal from which the noise signal is removed from the filter unit 35. After the amplification to a predetermined size (220) to facilitate processing in (10) and transmits the voice signal to the control unit (10).

The control unit 10 converts the analog voice signal received through the first amplifier 36 into a digital voice signal, and then normalizes the signal difference accumulation method and interval detection, which is a preprocessing process, to recognize the voice signal. In step 230, the speech signal, which has undergone the pre-processing process, is compared with the learning data stored in the first memory unit 21 by the IDMLP neural network algorithm, and then the corresponding data is extracted (240). 250), the voice and operation scenarios corresponding to the learning data are loaded into the second memory unit 22 through the voice and operation synthesizing unit 11 in a predetermined compressed form, and the scenario data are loaded. The controller 10 arbitrarily selects one voice expression data (260), selects the corresponding operation expression data, and outputs a control signal to the corresponding device of the output unit 40. Each W devices to a phonetic representation and operation expression according to the control signal 270.

For example, when a user makes a voice expression of 'what is your name' on a toy implemented by the present invention, the controller 10 recognizes the voice expression and recognizes the voice through a process as shown in FIG. 3. As shown in FIG. 4, the control unit 10 compares and searches the learning data stored in the first memory unit 21, and extracts the learning data of 'what is your name', and performs voice and operation. After extracting the voice scenario from the voice scenario table shown in FIG. 5 stored in the second memory unit 22 through the synthesis unit 11, 'My name is Saojung', 'Tell me first'. ‥ The controller 10 randomly selects one voice expression from among various voice expressions of 'what do you know' and then assigns an operation expression to it and supplies a control signal to the corresponding device of the output unit 40 to provide a predetermined voice. Expression and Action Table Do the strings.

As described above, the toy implemented by the present invention, which has undergone a voice recognition training process using the IDMLP neural network algorithm, recognizes various voice expressions of the user and implements them in software and hardware to enable different voice expressions for the same question. It is a useful toy with educational function as it can learn language by satisfying the interest and various voice expressions.

The present invention described above is capable of various substitutions, modifications, and changes without departing from the technical spirit of the present invention for those skilled in the art to which the present invention pertains. It is not limited to the drawings.

Recognizes user's voice using IDMLP neural network algorithm, randomly selects one voice expression from various voice expressions of voice scenarios corresponding to the recognized voice, and outputs it through a predetermined output device. It detects the tactile sensation of touch, detects a certain distance approach, detects light and darkness, and expresses the corresponding voice and motion, so that the user's curiosity and interest can always be changed due to various new voice expressions for the user's same question. It can satisfy and improve children's emotional development and language learning due to various voice expressions and movement expressions that reflect children's senses.

Claims (6)

In the interactive voice and multi-sensory recognition system for toys and methods thereof, A first step of extracting the selected speech sample into predetermined learning data through a speech recognition training process using an IDMLP neural network algorithm through a predetermined computer; A second step of storing the predetermined learning data extracted in the first step and a voice scenario corresponding to the learning data in a predetermined memory unit through a PC interface unit of a toy implemented according to the present invention; After the second step, when a predetermined voice signal is input through the input unit, a third step of extracting learning data corresponding to the input voice signal using the IDMLP neural network algorithm after performing a preprocessing process in the control unit; A fourth step of optionally selecting, by the controller, a predetermined voice expression and a motion expression from a voice output and motion expression scenario corresponding to the training data through a voice and operation synthesis unit and a predetermined memory unit; A fifth step of sending a control signal to a corresponding device to output the voice expression and the operation expression selected in the fourth step through an output unit; And a sixth step of outputting a predetermined voice expression and an operation expression by the corresponding device of the output unit in response to a control signal of the control unit. The method of claim 1, After the second step, when a predetermined signal is input through a predetermined sensor as an input unit, the controller recognizes the predetermined signal and selects a predetermined voice expression from among voice expression and operation expression scenarios stored in the predetermined memory unit through the voice and operation synthesis unit. A first step of arbitrarily selecting a motion expression; A second step of sending a control signal to a corresponding device to output the voice expression and the operation expression selected in the first step through an output unit; And a third step in which the corresponding device of the output unit outputs a predetermined voice expression and an operation expression according to a control signal of the controller. In the interactive voice and multi-sensory recognition system for toys and methods thereof, A controller for controlling the apparatus to recognize a voice signal supplied through an input unit using an IDMLP neural network algorithm and to output a predetermined voice expression and an operation expression by recognizing a signal supplied through a predetermined sensor; A first memory unit for storing the operating software of the control unit and storing learning data obtained through a learning process using an IDMLP neural network algorithm, a second memory unit for storing voice expression and operation expression scenarios, and internal data signal processing A memory unit comprising a third memory unit in charge of the; A voice and operation synthesizing unit for supplying predetermined voice expression and operation expression data stored in a second memory unit to the output unit in association with the control unit; A microphone for receiving a user's voice and converting it into an electrical analog voice signal, a tactile sensor for sensing a touch to supply a predetermined signal to the controller, an ultrasonic sensor for sensing a predetermined distance and supplying a predetermined signal to the controller; Infrared sensor for detecting light and darkness and supplying a predetermined signal to the control unit, Filter unit for removing noise and ambient noise from the user's voice signal, and a predetermined analog signal so that the controller can recognize and operate An input section comprising a first amplifier section amplifying to a magnitude; A second amplifier for amplifying a weak analog signal supplied through the voice and the operation synthesizer to a predetermined size to operate the corresponding device of the output unit, the user can recognize the electrical signal having a predetermined voice expression data An output unit including a speaker for converting and outputting a voice, an LED for emitting light at a predetermined interval according to a control signal of the controller, and a driving unit for outputting an operation corresponding to predetermined operation expression data; A PC interface unit for easily accessing a computer from which learning data is extracted by performing a predetermined learning process; Interactive voice and multisensory recognition system for toys, comprising a power supply unit for supplying a constant power to each circuit and element while maintaining a predetermined voltage The method according to any one of claims 1, 2 or 3, The sensor installed in a predetermined portion of the toy is configured as a tactile sensor for sensing the user's touch to output a predetermined voice expression and motion expression, interactive voice and multi-sensory recognition system in the toy and Method of objection The method according to any one of claims 1, 2 or 3, The sensor installed at a predetermined portion of the toy is configured with an ultrasonic sensor to detect a user approaching a certain distance and output a predetermined voice expression and motion expression, interactive voice and multi-sensory recognition in the toy System and Method The method according to any one of claims 1, 2 or 3, The sensor installed at a predetermined portion of the toy is configured with an infrared sensor to detect light and darkness to output a predetermined voice expression and motion expression, interactive voice and multi-sensory recognition system and a method thereof in the toy