KR102364032B1 - The Articulatory Physical Features and Sound-Text Synchronization for the Speech Production and its Expression Based on Speech Intention and its Recognition Using Derencephalus Action - Google Patents

Abstract

The present invention provides a sensor unit adjacent to one surface of a speaker's head and neck for measuring physical characteristics of a sound articulation organ; a data analysis unit configured to identify the speaker's utterance characteristics based on the position of the sensor unit and the physical characteristics of the articulation organ; a data conversion unit converting the position of the sensor unit and the speech characteristics into language data; a text information unit 600 for generating the language data of the data conversion unit 300 into voice information; a data expression unit interworking with the text information unit to externally express the voice information based on language data; The sensor unit provides a speech realization system including an oral tongue sensor corresponding to the oral tongue.

Description

{The Articulatory Physical Features and Sound-Text Synchronization for the Speech Production and its Expression Based on Speech Intention and its Recognition Using Derencephalus Action}

The present invention recognizes the physical characteristics of the articulation organs of the head and neck, including the oral tongue, through an articulation sensor and an imaging sensor, measures changes in the overall head and neck area of speech, and grasps the intention of speech through this, through visual, auditory, and tactile speech. The present invention relates to a system and method for providing an intention of a utterance to a speaker himself or herself or to the outside.

When the characters generated by the articulatory organ are for communication, which is linguistic information transfer, they are called speech or verbal sounds, and in non-linguistic cases, they are called vocalizations. The main organs of the human body involved in the creation of letters are the nervous system and the respiratory system.

The central nervous system and the peripheral nervous system are involved in the nervous system. Among the central nervous system, the brainstem contains the skull or cranial nerve cell nuclei necessary for the production of language, the cerebellum has the function of fine-tuning the muscle control for movement, and the hemispheres of the cerebrum are language play a dominant role in function. The cranial nerves involved in speech sound generation include the 5th cranial nerve involved in jaw movement, the 7th cranial nerve involved in lip movement, the 10th cranial nerve involved in the movement of the pharynx and larynx, and the 11th cranial nerve involved in the movement of the pharynx. , and the 12th nerve involved in tongue movement. Among peripheral nerves, the superior laryngeal nerve and semicircular laryngeal nerve branching from the vagus nerve are directly involved in laryngeal movement.

In addition, speech sounds are generated by the close interaction between the lower respiratory system, the larynx and the vocal tract. The vocal cords are the source of letters, and the flow of exhaled air from the lungs vibrates the vocal cords, and exhalation control during vocalization provides adequate and efficient sound energy. When the vocal cords are closed with moderate tension, the vocal cords vibrate by exhalation and open and close the glottis at regular intervals to control the exhalation flow that passes through the glottis. This intermittent flow is the sound source of characters.

In order for a person to use speech for the purpose of communication, various physiological processes must be passed. The articulation process refers to the process of forming a phoneme, a unit of speech sound, after the uttered sound is amplified and supplemented through a resonance process.

Although the tongue is considered the most important organ of articulation, not only the tongue but also various structures of the oral cavity and face are involved in making phonemes. These articulatory organs include movable structures such as the tongue, lips, soft palate, and jaw, and immovable structures such as teeth and hard palate. These articulation organs block or restrict the flow of air to form consonants and vowels.

It is not easy to distinguish the tongue as the first articulatory organ because its parts do not show clear boundaries, but distinguishing the external structure of the tongue from a functional point of view helps to understand pathological articulation as well as normal articulation. The tongue can be divided into the apex, tip, blade, dorsum, body, and root of the tongue. The tip of the tongue is the part used when we stick out our tongue or articulate /r/ (e.g. "lalala"), which comes as the first sound of a syllable, and the blade of the tongue is made in the front of the mouth like alveolar sounds. It is mainly used to articulate phonemes, and the tongue is a part of the tongue mainly used to articulate back-sounding phonemes such as velar sounds.

As the second articulatory organ, the lips form the entrance to the mouth and play an important role in facial expression and articulation of the head and neck. In particular, various vowels are differentiated by the shape of the lips as well as the movement of the tongue, and bilabial consonants (bilabial sounds) can be pronounced only when the lips are closed. The shape of the lips is modified by the surrounding muscles. For example, the orbicularis oris muscle surrounding the lips plays an important role in pronouncing double lip consonants and round vowels such as /u/ by closing or retracting the lips, and the levator lip muscle ( The quadratus labii superior muscle and the quadrates labii inferior muscle cause the lips to open. In addition, the risorius muscle plays an important role when smiling by pulling the corners of the lips or making sounds such as /i/ that should be pronounced by contracting the lips.

Among the jaws and teeth as the third articulatory organ, the jaw is divided into an immovable upper jaw (maxilla maxilla) and a lower jaw (mandible mandible) that moves up and down and left and right. These jaws are the strongest and largest of the facial bones, and are moved by four pairs of muscles. Because the movement of the lower jaw changes the size of the mouth, it is important not only for chewing but also for vowel production.

Among the gums and hard roof of the mouth as the fourth articulating organ, the gum is the area where speech sounds of the /c/ and /s/ series are articulated, and the hard palate is the hard and rather flat part behind the gums where the /j/ series sounds are articulated. is the part

The soft palate as the last articulatory organ is classified as a moving articulatory organ, because the muscles of the soft palate contract to form a capillary closure and articulate oral sounds accordingly.

<Articulation process>

Some of the sounds are produced by the oral cavity, or more precisely, in the middle of the oral passages, while the airflow through the vocal cords passes through the vocal cords with some obstruction (obstruction), while others are produced without any obstruction. Usually, the former is called a consonant and the latter is called a vowel.

1) articulation of consonants

Consonants should be looked at according to the method and position in which they are uttered.

First of all, if classified according to the articulation method, it can be broadly divided into da-mak-eum and dul-maeum sounds according to what kind of interference the airflow receives from the central part and is articulated. Daumakeum sound is a sound made by completely blocking the airflow in the oral cavity and then popping it. Deulcaeum sound is a sound made by narrowing a part of the vocal tract and allowing airflow to pass through the narrowed passageway.

Dajakeum sounds can be divided into nasal resonance and non-accompanied sounds. The former is the sound of nasal obstruction (nasal stop), which is made while completely blocking a part of the vocal tract while lowering the soft palate to open the nasal passage and accompanying the resonance of the nasal cavity. The latter belongs to the closed-mouth sounds (oral stops) produced by blocking the passage of airflow into the nasal cavity. Depending on the length and method of closing, the closing sound can be thought of as a closing sound (stop sound) or plosive sound (popping sound, plosive), electric sound (trembling sound, trill), or snoring sound (or flap/tap sound) depending on the length and method of closing. can see.

In addition, the deulmakeum sound is divided into a fricative sound and an approximate sound. When the airflow path is made on the side of the tongue, it is collectively called a lateral sound.

Also, there is an affricate, which uses the articulation method of Damakgal and Deulmaekeum in combination. Lastly, in the alphabet, it is expressed as "r" or "l", but in the case of Korean, it is expressed as /r/. (liquid) and Korean, but there are electric sounds that make sounds by vibrating the articulatory organ.

Classified according to the articulation position, bilabial refers to the sound related to the articulation of the two lips. All of the positive labial consonants in modern Korean (standard language) are sounds made by closing the two lips, but by narrowing the gap between the two lips, the airflow can be rubbed between them (bilabial fricatives) and can also be made by separating the two lips (bilateral electric consonants). .

Labiodentals refer to sounds related to articulation between the lower lip and upper teeth, and do not exist in Korean. There is no pure apical consonant in Korean, but [f, v] in English belongs to this pure apical consonant.

The term “dental” refers to the sound of airflow constriction or occlusion at the back of the upper teeth.

Alveolar is a sound produced when airflow is narrowed or obstructed in the vicinity of the upper gums. In Korean, /ㅅ and ㅆ/ are sounds produced by constriction of airflow at the alveolar part, and in English, /s and z/ and the place where airflow is constricted are almost the same.

Palatoalveolar, also called postalveolar, is a sound produced when the tip of the tongue or tongue touches the postalveolar region. It does not exist in Korean, but does exist in English and French.

Alveolopalatal is also called prepalatal because it is articulated in front of the hard palate, close to the alveolar canal. The three fascia consonants of the Korean language /j, ch, and tk/ belong to this category.

Unlike other tongue consonants in which the tip of the tongue or the upper surface of the tongue touches or approaches the roof of the mouth to articulate, retroflex has a distinct difference in that the lower surface of the tongue is articulated by touching or approaching the roof of the mouth.

A palatal is a sound articulated by the tongue touching or approaching the palatal area.

A velar is a sound articulated by the tongue touching or approaching the soft palate. Korean closing consonants /a, ㅋ, ㄲ/ and nasal /ㅇ/ belong to this category.

The uvular is the sound articulated by the tongue touching or approaching the palate, which is the tip of the soft palate.

Pharyngeal refers to the sound whose articulation is made in the pharyngeal cavity.

Lastly, glottal refers to sounds that are articulated by using the vocal cords as articulation organs.

2) articulation of vowels

In the articulation of vowels, the three most important variables are the tongue's height and anteroposterior position, and the shape of the lips.

As the first variable, the opening degree of vowels, that is, the degree of opening the mouth, is determined by the height of the tongue. is called an open vowel, or a low vowel. And the sound made between the high and low vowels is called a mid vewel. It can be subdivided into open-mid vewels or half-open vewels with a greater degree of spread.

The second variable, the anteroposterior position of the tongue, is actually a front-to-back basis based on which part of the tongue is most narrowed, that is, which part of the tongue is closest to the roof of the mouth. The vowel at the front of the tongue with the narrowed part is called the front vowel, the vowel at the back is called the back vowel, and the vowel in the middle is called the central vowel.

Finally, an important variable in the articulation of vowels is the shape of the lips. A vowel in which the lips are rounded and protruding forward during articulation is called a rounded vowel, and a vowel that does not protrude is called an unrounded vowel.

Speech disorder refers to a condition in which pitch, intensity, sound quality, and fluidity are not appropriate for gender, age, body size, social environment, and geographic location. It can be made congenitally or acquired, and it is possible to treat it to some extent by lengthening or reducing the vocal cords, a part of the larynx, through surgery. However, it is not a perfect treatment, and the effect cannot be said to be accurate either.

The functions of the larynx include swallowing, coughing, obstruction, respiration, and vocalization, and there are various evaluation methods for this (ex. utterance history examination, speech pattern, acoustic examination, aerodynamic examination...). . Through this evaluation, it is possible to determine to some extent whether there is a speech disorder.

*44 There are various types of speech disorders, and they are largely divided into functional speech disorders and organic speech disorders. In most of these types, abnormalities occur in the vocal cords, which are part of the larynx, and these vocal cords are often impaired due to swelling, tearing, or abnormal substances due to external environmental factors.

In order to replace the function of the vocal cords, a vibration generator capable of artificially generating vibration may be used. The method of the vibration generator can use the principle of the speaker. If you look at the structure of the speaker, it consists of a magnet and a coil, and if the direction of the current is reversed while the current flows through these coils, the pole of the magnet is reversed. Therefore, attractive and repulsive forces act depending on the direction of the current between the magnet and the coil, which causes the coil to reciprocate. The reciprocating motion of the coil vibrates the air to generate vibration.

As another method, there is a method using the piezoelectric phenomenon, and the piezoelectric crystal unit receives a low-frequency signal voltage to generate distortion, whereby the diaphragm vibrates to generate sound. Therefore, it is possible to perform the function of the vocal cords using a vibration generator using these principles.

However, in the case of this method, because it is located outside and merely functions to vibrate the vocal cords, the sound produced is very inaccurate and it is difficult to understand the speaker's intention to speak. In addition, it has a vibration generator and is located on the vocal cords, so it must be carried at all times, and when speaking, one hand is used, which makes daily life difficult. For the above-mentioned speech disorders and these speech abnormalities, therapeutic methods such as surgery for part of the larynx or vocal cords can be sought, but there are cases in which these surgical methods or treatments are not possible, so it is not a complete solution.

In particular, in the related industry, the Rion EPG, which was widely commercialized under the name of Rion, developed in 1973 by the University of Reading, Fujimura, and Tatsumi of Japan, which are being used by companies such as WinEPG and Articulate Instruments Ltd. Kay Palatometer developed and used by UCLA Phonetics Lab for research purposes based on this application, and Complete Speech (Logomertix) developed by Schmidt.

However, conventional techniques have limitations in uttering based on a passive articulatory organ, and it is difficult to implement utterance according to the actual articulation method by using the oral tongue, which is an active articulation organ itself, or by connecting the oral tongue with other articulation organs. There were clear limitations.

Various sensors have been developed to detect changes in state or movement, and it is possible to detect changes in pressure, temperature, distance, friction, etc. based on the sensors.

In addition, speech and expression synchronization (Lip Sync) reproduces speech and expression including spoken voice and articulation, which are the most important factors determining the identity of an object or object, to characters, robots, various electronic products, autonomous vehicles, etc. It is a key means of determining and expanding an individual's identity by applying it. In particular, since a professional animation team works to create a high-quality Lip Sync animation, it requires high cost and a lot of time, and there are difficulties in mass production. Conventional general techniques have been limited to creating low-level animations using a simple lip shape library. Overseas animation content producers such as Pixar and Disney spend a lot of money and time to create realistic character animations through Lip Sync.

The present invention has been proposed to solve the above problems, and an object of the present invention is to determine the user's articulation method according to the user's utterance intention through the sensor of the head and neck including the oral tongue, An object of the present invention is to provide an apparatus for supplementing speech, and a method therefor, in which speech formation of good quality, that is, speech can be expressed in the form of speech.

Another object of the present invention is to implement high-quality and appropriate utterance in a case in which a normal function cannot be performed in utterance and correction or treatment is impossible.

Another object of the present invention is to provide a device for supplementing speech located inside/outside, and a method for controlling the same, in which an accurate speech desired by a user can be expressed to the outside according to an articulation intention for speech.

In order to achieve the above object, the present invention, a sensor unit for measuring the physical characteristics of the articulation organ adjacent to one surface of the speaker's head and neck;

a data analysis unit configured to identify the speaker's utterance characteristics based on the position of the sensor unit and the physical characteristics of the articulation organ;

2 a data conversion unit for converting the position of the sensor unit and the speech characteristics into language data;

a text information unit for generating the language data of the data conversion unit 300 into voice information;

a data expression unit interworking with the text information unit to externally express the voice information based on language data;

The sensor unit provides a speech implementation system including an oral tongue sensor corresponding to the oral tongue.

And, the oral tongue sensor is fixed to one side of the oral tongue, wraps around the surface of the oral tongue, or is inserted into the oral tongue, and the x-axis, y-axis, z-axis of the oral tongue according to the utterance By grasping the amount of change of the vector quantity over time based on the direction, the physical characteristics of at least one of the low elevation, anteroposterior, flexion, extension, rotation, tension, contraction, relaxation, and vibration of the oral tongue can be grasped. there is.

In addition, the oral tongue sensor is fixed on one side of the oral tongue, wraps around the surface of the oral tongue, or is inserted into the oral tongue, and the x-axis, y-axis, z-axis of the oral tongue according to the utterance By determining the amount of change in the rotation angle per unit time based on the direction, the physical characteristics of the articulatory organ including the oral tongue may be grasped.

In addition, the oral tongue sensor is fixed to one side of the oral tongue or surrounds the surface of the oral tongue, and polarization due to a change in crystal structure according to physical force generated by contraction and relaxation of the oral tongue according to ignition By grasping the degree of bending of the oral tongue through a piezoelectric element that generates an electrical signal corresponding to At least one of the physical properties may be identified.

In addition, the sensor unit, the degree of rupture, friction, resonance, and approach according to the tribo electric generator corresponding to the approach and contact caused by the interaction of the oral tongue with other articulatory organs inside and outside the head and neck It may include a triboelectric charging element for detecting at least one physical characteristic.

And, the data analysis unit, the consonants uttered by the speaker through the physical characteristics of the oral tongue and other articulation organs measured by the sensor unit, lexical stress (Lexical Stress), sentence unit stress (Tonic stress) At least one speech characteristic may be identified.

In addition, the data analysis unit, in recognizing the speech characteristics according to the physical characteristics of the articulatory organ measured by the sensor unit, the speaker's pronunciation based on a standard speech characteristic matrix composed of a number including a binary number or a real number At least one speech characteristic of hyperstress, similarity, and intention to speak may be measured.

And, when the data analysis unit recognizes the speech characteristics of the physical characteristics of the articulation organ measured by the sensor unit, the step of recognizing the physical characteristics of the articulation organ as a pattern of each consonant unit, the consonant unit extracting the characteristics of the pattern of , classifying the extracted characteristics of the pattern of the consonant-vowel unit according to the degree of similarity, recombining the classified characteristics of the pattern of the consonant-vowel unit, and uttering the physical characteristics of the articulation organ According to the step of interpreting the feature as the feature, the utterance feature may be identified.

In addition, the data analysis unit, according to the physical characteristics of the articulation organ measured by the sensor unit, Assimilation, Dissimilation, Elision, Attachment, Stress of consonants Asperation, Syllabic cosonant, Flapping, Tensification, Labilalization, Velarization, and Gingivization caused by reduction Dentalizatiom), palatalization, nasalization, stress shift, and lengthening, which are secondary articulation phenomena of at least one, can be measured.

In addition, the oral tongue sensor may include a circuit part for sensor operation, a capsule part surrounding the circuit part, and an adhesive part attached to one surface of the oral tongue.

In addition, the oral tongue sensor, in the form of a film with a thin film circuit, may operate adjacent to the oral tongue.

And, the sensor unit, at least one reference sensor for generating a reference potential for measuring the nerve signal of the head and neck muscles, and at least one positive sensor and at least one negative sensor for measuring the nerve signal of the head and neck muscles. may include

In addition, the data analysis unit, in obtaining the position of the sensor unit based on the face sensor, grasps the potential difference between the at least one positive sensor and the at least one negative sensor with respect to the reference sensor to determine the facial sensor location can be found.

In addition, the data analysis unit, when acquiring the speech characteristics of the speaker based on the face sensor, grasps the potential difference between the at least one positive sensor and the at least one negative sensor with respect to the reference sensor to determine the speaker It is possible to grasp the speech characteristics by the physical characteristics of the articulation organ generated in the head and neck of the.

In addition, the sensor unit may include a vocal cord sensor that detects electromyography or tremor of the vocal cords adjacent to the vocal cords of the speaker's head and neck, and grasps information on the speech history of at least one of the speaker's speech start, speech stop, and speech end. can

In addition, the sensor unit may include a tooth sensor that detects a signal generation position according to a change in electrical capacity generated due to the contact between the oral tongue and the lower lip adjacent to one surface of the tooth.

Also, the data analysis unit may acquire the speaker's voice according to the utterance through a voice acquisition sensor adjacent to one surface of the speaker's head and neck.

In addition, the sensor unit detects at least one of non-verbal expressions of the speaker's head and neck articulatory organs, location change information, the speaker's head and neck facial expression change information, and the head and neck, rib cage, upper extremities, and lower extremities that move according to the speaker's intention to speak. In order to do this, it may include an imaging sensor that images the speaker's head and neck.

In addition, the speech implementation system may further include a power supply for supplying power to at least one of the oral tongue sensor, the facial sensor, the voice acquisition sensor, the vocal cord sensor, the tooth sensor, and the image sensor of the sensor unit.

In addition, the speech implementation system may further include a wired or wireless communication unit capable of interworking and communicating when the data analysis unit and the database unit are located outside and operate.

In addition, the data analysis unit may be linked with a database unit including the location of the sensor unit, the speaker's utterance characteristics, and at least one language data index corresponding to the speaker's voice.

In addition, the database unit includes at least one of a duration of utterance, frequency according to utterance, amplitude of utterance, electromyography of head and neck muscles according to utterance, position change of head and neck muscles according to utterance, and position change according to bending and rotation of the oral tongue. Based on the information of .

In addition, the data expression unit is linked with the language data index of the database unit to display the speaker's utterance characteristics in phoneme units, at least one word unit, at least one phrase unit (Citation Forms), at least one At least one speech expression among sentence units and continuous speech units of

In addition, the speech expression displayed by the data expression unit may be visualized as at least one of a character symbol, a picture, a special symbol, and a number, or may be audibly provided to the speaker and the listener.

In addition, the speech expression displayed by the data expression unit may be provided to the speaker and the listener through at least one tactile method of vibration, snoozing, tapping, pressing, and relaxation.

The speech realization system based on the physical characteristics of the head and neck articulatory organ of the present invention identifies the intention of speech in the use of the head and neck articulatory organ centering on the oral language of the speaker and expresses it in the form of hearing, sight, and tactile sense to form good quality speech, i.e., vocalization. This has an effect that can be expressed. In addition, it is possible for the speaker to improve his/her own utterance through feedback through comparison between the content of the first utterance and at least the content of the second utterance.

In the present invention, articulation organs inside and outside the head and neck, including the oral tongue, are used to determine the intention to speak. Degree of occlusion, degree of rupture, degree of friction, and resonance caused by interaction with one or more of the active articulators composed of one or more of the following. At least one characteristic of the approach should be grasped, and in order to grasp these characteristics, various sensors that can grasp azimuth, elevation, rotation, pressure, friction, distance, temperature, sound, etc. are used.

In the case of the previously proposed artificial vocal cords, there were disadvantages in that the movement of one hand was unnatural and the quality of speech was very low, to the extent of making sound through vibration from the outside. there was.

In addition, articulatory phonetics, which attempts to measure the speaker's utterance using the artificial palate, has been accepted as the mainstream phonetics, but only the presence or absence of discrete utterance according to the articulation of specific consonant vowels in utterance measurement has been accepted. could figure it out However, this argument of articulatory phonetics is not that human speech has discrete characteristics. For each phoneme, especially vowel, each vowel cannot exist as a segment and cannot be articulated as a continuous system, which is being called into question by Acoustic Phonetics. More specifically, human speech is articulated to "speech." Or, it is not that it can be divided discretely, such as "Failed to ignite", but has proportional, proportionate, and step-by-step characteristics according to the degree of similarity.

Therefore, acoustophonology quantifies the physical properties of speech sounds according to the speaker's utterance, and grasps the similarity or proximity of the utterances. The possibility of fire measurement according to severity is left open.

When referring to these prior art trends and related academic backgrounds, the present invention is a very innovative method that can identify and implement a more accurate utterance intention according to the scaling of articulation that acoustic phonetics seeks based on articulation phonetics. It can be said that it has advantages.

Specifically, in the present invention, the quality of communication and living convenience are very excellent because the intention of speech is intuitively presented in the form of auditory, visual and tactile senses by scaling the articulation degree generated by the speaker's articulation organ action. that is expected to happen.

In addition, when expressing the intention of the speaker according to the speaker's utterance as text, it is applied as Speech to Text, enabling silent speech. Through this, when communicating with the hearing impaired, the speaker makes a utterance and the hearing impaired, the listener, perceives it as visual data, so there is no communication difficulty. In addition, it can be used in public transportation, public facilities, military facilities and operations, and underwater activities that are affected by noise in communication.

In addition, by capturing the trauma of the speaker's head and neck articulatory organs that change according to speech, the relationship between speech and external changes in the articulatory organ according to speech is identified, and used in the area of linguistics, complementary and alternative communication, and humanoid facial realization. can be

In particular, in the animation and film production industry, it is difficult to achieve the matching of speech and facial expressions of video objects including animated characters to date. The most problematic part is the operation of the articulation organ and the area of speech. It does not properly reflect the physical characteristics of the complex human articulatory organ, so even giant companies such as Walt Disney and Pixar are only at the level of development where the characters are just bubbly, showing a low degree of correspondence between dialogue, utterance, and facial expression. In order to solve this problem, a high cost is paid to the video production team and feature points are attached to the voice actor or the copying actor on the whole body. However, this method does not solve the fundamental part of the speech or expression of an image object, and there is a limitation in that the main purpose is to express macroscopic movements throughout the body. However, the present invention measures the physical characteristics of the articulation organ of an actual human speaker and maps them to the head and neck of the image object, so that the speech or expression of the image object can be implemented similar to that of the actual human speaker.

In particular, the present invention reproduces head and neck movements including articulation, utterance, and expression of a robot similar to a human speaker by transmitting and matching speaker articulation information to an actuator that implements the movement of the head and neck of a robot object, which is a Japanese Mori It has the effect of overcoming the "Uncanny Valley", a chronic cognitive dissonance that humanoid robots claim by Mori Masahiro in humans. In addition, as humanoids and other general robots can implement human-friendly articulation, human roles of robots and androids can be replaced, and further, human-robot dialogue is achieved, which is emerging as the elderly population increases due to aging. It has the effect of preventing the isolation of the elderly and mental/psychological diseases such as depression.

1 is a view showing a sensor unit of an utterance realization system according to a first embodiment of the present invention.
2 is a view showing the position of the sensor unit of the utterance realization system according to the first embodiment of the present invention.
3 is a diagram illustrating an utterance realization system according to a first embodiment of the present invention.
4 is a view showing the positional name of the oral tongue used in the speech realization system according to the first embodiment of the present invention.
5 is a diagram illustrating the operation of oral language for vowel utterance utilized in the utterance realization system according to the first embodiment of the present invention.
6 to 10 are views showing various oral tongue sensors of the speech realization system according to the first embodiment of the present invention, respectively.
11 and 12 are a cross-sectional view and a perspective view showing the attachment state of the oral tongue sensor of the speech realization system according to the first embodiment of the present invention, respectively.
13 is a view showing a circuit part of the oral tongue sensor of the speech realization system according to the first embodiment of the present invention.
14 is a view showing various utilization states of the oral tongue sensor of the speech realization system according to the first embodiment of the present invention.
15 is a diagram illustrating an utterance realization system according to a second embodiment of the present invention.
16 is a diagram illustrating a principle of a data analysis unit identifying an utterance characteristic of the utterance realization system according to the second embodiment of the present invention.
17 is a diagram illustrating a principle of identifying the physical characteristics of the articulation organ measured by the data analysis unit of the speech realization system according to the second embodiment of the present invention as speech characteristics.
18 is a diagram illustrating a standard utterance feature matrix for vowels utilized by the data analysis unit of the utterance realization system according to the second embodiment of the present invention.
19 is a diagram illustrating a standard speech feature matrix for consonants utilized by the data analysis unit of the speech realization system according to the second embodiment of the present invention.
20 is a diagram illustrating an algorithm process used by the data analysis unit of the speech realization system according to the second embodiment of the present invention to determine the physical characteristics of the articulation organ as speech characteristics.
21 is a diagram illustrating in detail an algorithm process used by the data analysis unit of the speech realization system according to the second embodiment of the present invention to determine the physical characteristics of the articulatory organ as speech characteristics.
22 is a diagram illustrating in detail the principle of an algorithm process used by the data analysis unit of the speech realization system according to the second embodiment of the present invention to identify the physical characteristics of the articulatory organ as speech characteristics.
FIG. 23 is a diagram illustrating an algorithm process in which an oral tongue sensor of a speech realization system according to a second embodiment of the present invention recognizes a specific vowel uttered as a speech feature.
24 is a diagram illustrating a case in which the data analysis unit of the speech realization system according to the second embodiment of the present invention utilizes Alveolar Stop.
25 is a diagram illustrating a case in which the data analysis unit of the utterance realization system uses Bilabial Stop according to the second embodiment of the present invention.
26 is a view showing an experimental result using Voiced Bilabial Stop by the data analysis unit of the speech realization system according to the second embodiment of the present invention.
27 and 28 are diagrams each illustrating a case in which the data analysis unit of the speech realization system according to the second embodiment of the present invention utilizes Voiced Labiodental Fricative.
29 is a diagram illustrating interworking of a data analysis unit and a database of the speech realization system according to the second embodiment of the present invention.
30 is a diagram illustrating a case in which the data analysis unit of the speech realization system according to the second embodiment of the present invention recognizes a specific word.
31 is a diagram illustrating a database unit of a speech realization system according to a second embodiment of the present invention;
32 is a diagram illustrating an utterance realization system according to a third embodiment of the present invention.
33 and 34 are diagrams each showing an actual form of a database unit of an utterance realization system according to a third embodiment of the present invention;
35 is a diagram illustrating an utterance realization system according to a fourth embodiment of the present invention;
36 is a diagram showing the interworking of a sensor unit, a data analysis unit, a data expression unit, and a database unit of the speech realization system according to the fourth embodiment of the present invention.
37 to 41 are views each showing means for expressing language data by the data expression unit of the speech realization system according to the fourth embodiment of the present invention;
42 is a diagram illustrating a case in which the data expression unit of the speech realization system according to the fourth embodiment of the present invention expresses language data visually and audibly.
43 is a diagram illustrating a case in which the data expression unit of the speech realization system according to the fourth embodiment of the present invention visually expresses language data.
44 is a diagram illustrating a case in which the data expression unit of the speech realization system visually expresses language data according to the fourth embodiment of the present invention.
45 is a diagram illustrating a case in which the data expression unit of the speech implementation system according to the fourth embodiment of the present invention expresses language data in units of continuous speech.
46 is a diagram illustrating a confusion matrix utilized by the utterance realization system according to the fourth embodiment of the present invention.
47 is a diagram illustrating a Confusion Matrix utilized by the utterance realization system according to the fourth embodiment of the present invention as a percentage;
48 is a diagram illustrating a case in which the speech realization system according to the fourth embodiment of the present invention assists the speaker with language correction and guidance through the screen.
49 is a diagram illustrating a case in which the speech realization system according to the fourth embodiment of the present invention captures and detects trauma to the head and neck articulatory organ.
50 is a diagram illustrating a case in which the speech implementation system according to the fourth embodiment of the present invention combines mutual information through a standard speech feature matrix.
Fig. 51 is a diagram showing a non-verbal expression of a speaker captured by the imaging sensor of the present invention;
52 is a diagram illustrating a case in which the text information unit of the present invention generates voice information.
53 is a view showing a partial example of the frequency table of the phoneme unit index utilized by the text information unit of the present invention;
54 is another view showing a partial example of the frequency table of the phoneme unit index utilized by the text information unit of the present invention.

Hereinafter, a speech implementation system using the physical characteristics of the head and neck articulatory organ for speech improvement guide and feedback according to an embodiment of the present invention will be described in detail with reference to the accompanying drawings.

It goes without saying that the following examples of the present invention are not intended to limit the scope of the present invention or to limit the scope of the present invention only to embody the present invention. What can be easily inferred by an expert in the technical field to which the present invention belongs from the detailed description and examples of the present invention is construed as belonging to the scope of the present invention.

The contents for carrying out the present invention will be described in detail with reference to FIGS. 1 to 54 .

1 is a view showing the sensor unit of the speech realization system according to the first embodiment of the present invention, Figure 2 is a view showing the position of the sensor unit of the ignition improvement guiding and feedback system according to the first embodiment of the present invention and FIG. 3 is a diagram illustrating an utterance realization system according to a first embodiment of the present invention.

1, 2 and 3, in the speech implementation system according to the first embodiment of the present invention, the sensor unit 100 includes an oral tongue sensor 110 and a facial sensor 120 located in the head and neck region. , is composed of a voice acquisition sensor 130 , a vocal cord sensor 140 , and a tooth sensor 150 .

In more detail, the oral tongue sensor 110, the facial sensor 120, the voice acquisition sensor 130, the vocal cord sensor 140, and the dental sensor 150 located in the head and neck region are the position 210 of the sensor unit where each sensor is located. ), utterance characteristics 220 according to the utterance of the speaker 10 , the speaker's voice 230 , utterance history information 240 , and utterance variation 250 are provided.

The data analysis unit 200 acquires these data, and the data conversion unit 300 processes these data as language data 310 .

4 is a view showing the positional names of oral tongues used in the speech realization system according to the first embodiment of the present invention, and FIG. 5 is a vowel utterance used in the speech realization system according to the first embodiment of the present invention. It is a diagram showing the action of oral seol for

4 and 5, in the case of the oral tongue sensor 110, it is fixed to one side of the oral tongue 12, wraps around the surface, or is inserted therein, and is low-altitude, anterior-posterior, and flexion. Determine the independent physical properties of the oral tongue itself at least one of degree, extension degree, rotation degree, tension degree, contraction degree, relaxation degree, and vibration degree.

6 to 10 are views showing various oral tongue sensors of the speech realization system according to the first embodiment of the present invention, respectively.

6 and 7, in determining the independent physical properties of the oral tongue 12 itself, the oral tongue sensor 110 rotates per unit time and acceleration in the x-axis, y-axis, and z-axis directions. By grasping at least one of the amount of change of the angle (angular velocity), the speech characteristics 220 based on the physical characteristics of other articulatory organs including the oral tongue 12 are grasped.

As shown in FIG. 8 , the oral tongue sensor 110 is polarized by a change in the crystal structure 111 according to the physical force generated by the contraction or relaxation of the oral tongue 12 according to the ignition, thereby generating an electrical signal. By grasping the degree of bending of the oral tongue 12 through the piezoelectric element 112 to, it is possible to grasp the firing characteristics 220 according to the physical characteristics of the articulatory organ including the oral tongue 12.

As shown in Fig. 9, the oral tongue sensor 110 uses a tribo electric generator generated by the approach and contact caused by the interaction of the oral tongue 12 with other articulatory organs inside and outside the head and neck. In order to grasp the associated physical characteristics, the speaker's utterance characteristics 220 are identified using the triboelectric charging element 113 .

As shown in FIG. 10, the integrated oral tongue sensor 110 uses acceleration and angular velocity in the x-axis, y-axis, and z-axis directions, an electrical signal by piezoelectricity, and friction electricity by contact. The speech characteristics 220 by the physical characteristics of the articulation organ including

11 and 12 are sectional views and perspective views, respectively, showing the attachment state of the oral tongue sensor of the speech realization system according to the first embodiment of the present invention.

11 and 12, the oral tongue sensor 110 may be implemented in the form of a single film composed of a complex thin film circuit. At this time, the oral tongue sensor 110 includes the circuit part 114 for operating the sensor part 100, the capsule part 115 surrounding the circuit part 114, and the oral tongue sensor 110 on one surface of the oral tongue 12. It is composed of an adhesive portion 116 for fixing.

As shown in Figures 6 to 9, the oral tongue sensor 110 is a degree of rupture, friction, resonance, and approach caused by proximity or contact with other articulatory organs inside and outside the head and neck according to the characteristics of each sensor. One or more physical properties may be identified.

13 is a diagram illustrating a circuit part of the oral tongue sensor of the utterance realization system according to the first embodiment of the present invention.

13, the circuit unit 114 of the oral tongue sensor 110 is composed of a communication chip, a sensing circuit, and an MCU.

14 is a view showing various usage states of the oral tongue sensor of the speech realization system according to the first embodiment of the present invention.

As shown in FIG. 14 , the oral tongue sensor 110 may determine the state of the oral tongue 12 according to the speaker's utterance of various consonant consonants, and thereby determine the speech characteristics 220 according to the consonant utterance.

For example, the oral tongue sensor 110 may recognize the speech characteristics 220 according to a bilabial sound (labial sound), an alveolar sound (alveolar sound), and a Palatal sound (palatal sound).

15 is a diagram illustrating an utterance realization system according to a second embodiment of the present invention.

15, in the speech implementation system according to the second embodiment of the present invention, the oral tongue sensor 110, the facial sensor 120, the voice acquisition sensor 130, the vocal cord sensor 140, and the tooth sensor The sensor unit 100 in the vicinity of the head and neck articulatory organ consisting of 150 includes a position 210 of the sensor in which the sensor is located in the head and neck articulation organ, a speech characteristic according to the speech 220, the speaker's voice according to the speech 230, The utterance history information 240 including utterance start, utterance stop, and utterance end is identified.

At this time, the speech characteristics 220 include one or more basic physical speech characteristics of closed rupture, fricative, fricative, nasal, consonant, active, sibilization, unvoiced, and glottal consonants that occur when a human utters. it means. Also, the speaker's voice 230 is an auditory utterance characteristic that accompanies the utterance characteristic. And, the utterance history information 240 is transmitted through the vocal cord sensor 140, and the information is grasped by electromyography or tremor of the vocal cords.

The data analysis unit 200 is a sensor unit 100 near the head and neck articulatory organ consisting of an oral tongue sensor 110 , a facial sensor 120 , a voice acquisition sensor 130 , a vocal cord sensor 140 , and a tooth sensor 150 . ), the speech variation 250 that occurs according to the speaker's gender, race, age, and native language is identified from the speaker's articulation organ physical characteristics measured by .

At this time, the speech variation 250 is assimilation, dissimilation, elimination, attachment, stress, and reduction of consonant vowels. Syllabic cosonant, Flapping, Tensification, Labilalization, Velarization, Dentalizatiom, Palatalization, Nasalization, Stress change ( Stress Shift) and lengthening include one or more secondary articulation phenomena.

The data conversion unit 300 includes a position 210 of the sensor unit measured by the head and neck articulatory organ sensors 110 , 120 , 130 , 140 , 150 , an utterance characteristic 220 according to the utterance, and the speaker's voice according to the utterance ( 230 ), the utterance history information 240 , and the utterance variation 250 are recognized and processed as the language data 310 .

At this time, when the data conversion unit 300 recognizes and processes the language data 310 , the data analysis unit 200 is interlocked with the database unit 350 .

The database unit 350 includes a phoneme unit 361 of consonants and vowels, an index syllable unit index 362 , a word unit index 363 , a phrase unit index 364 , a sentence unit index 365 , and a continuous utterance index 366 . ), has a linguistic data index 360 including a high and low index 367 of pronunciation. Through the language data index 360 , the data analysis unit 200 can process various utterance-related information acquired from the sensor unit 100 as language data.

16 is a diagram illustrating the principle of the data analysis unit grasping the speech characteristics of the speech realization system according to the second embodiment of the present invention, and FIG. 17 is a data analysis unit of the speech realization system according to the second embodiment of the present invention. It is a diagram showing the principle of recognizing the measured physical characteristics of the articulatory organ as speech characteristics, and FIG. 18 shows a standard speech feature matrix regarding vowels used by the data analysis unit of the speech realization system according to the second embodiment of the present invention. 19 is a diagram illustrating a standard speech feature matrix for consonants utilized by the data analysis unit of the speech realization system according to the second embodiment of the present invention.

16 , 17 , 18 and 19 , the data analysis unit 200 first acquires the physical characteristics of the articulation organ measured from the sensor unit 100 including the oral tongue sensor 110 . . When physical characteristics of the articulation organ are acquired by the oral tongue sensor 110 , the oral tongue sensor 110 creates a matrix value of the sensed physical characteristics while sensing the physical characteristics of the articulation organ.

Thereafter, the data analysis unit 200 detects the speech feature 220 of the consonant vowel corresponding to the matrix value of the physical property from the standard speech feature matrix 205 of the consonant. In this case, in the standard utterance feature matrix 205 of consonants, values therein may exist as one or more of consonant utterances, binary numbers, or real numbers.

20 is a diagram illustrating an algorithm process used by the data analysis unit of the speech realization system according to the second embodiment of the present invention to determine the physical characteristics of the articulatory organ as speech characteristics.

As shown in FIG. 20, the algorithm process utilized by the data analysis unit 200 includes the steps of acquiring the physical characteristics of the articulation organ in recognizing the physical characteristics of the articulation organ measured by the sensor unit 100; Identifying the pattern of each consonant unit of the acquired physical characteristics of the articulatory organ, extracting unique features from each consonant pattern, classifying the extracted features, recombining the features of the classified pattern It consists of steps, and through this, it is finally identified as a specific utterance characteristic.

21 is a diagram illustrating in detail an algorithm process used by the data analysis unit of the speech realization system according to the second embodiment of the present invention to identify the physical characteristics of the articulatory organ as the speech characteristics, and FIG. 22 is the second embodiment of the present invention. It is a diagram showing in detail the principle of an algorithm process used by the data analysis unit of the speech realization system according to the embodiment to identify the physical characteristics of the articulatory organ as the speech characteristics, and FIG. 23 is the speech implementation according to the second embodiment of the present invention. It is a diagram showing an algorithmic process in which the oral tongue sensor of the system identifies a specific vowel uttered as a utterance feature.

As shown in FIGS. 21, 22 and 23 , in the speech characteristic determination algorithm performed by the data analysis unit 200, the step of recognizing the unit pattern of each consonant is a sensor that grasps the physical characteristics of the articulation organ When the part 100 is the oral tongue 12, the pattern of the consonant-vowel unit is determined based on the x, y, and z axes.

At this time, the algorithm is one or more of K-nearset Neihbor (KNN), Artificial Neural Network (ANN), Convolutional Neural Network (CNN), Recurrent Neural Network (RNN), Restricted Boltzmann Machine (RBM), and Hidden Markov Model (HMM). can be based on

For example, in FIGS. 22 and 23 , when the oral tongue sensor 110 is driven as a sensor that detects the amount of change in the vector amount or the amount of change in the angle, the amount of change in the vector amount and the amount of change in the angle are determined by measuring the speaker's utterance, and this It is recognized as a vowel [i] with Tongue Height and Tongue Frontness.

In addition, when the oral tongue sensor 110 is a sensor driven by the principle of a piezoelectric signal or a triboelectric signal, an electrical signal change according to the piezoelectric force and the oral tongue sensor 110 and the oral cavity sensor 110 and the inside and outside of the oral cavity are in proximity or friction. It recognizes the triboelectric signal and recognizes it as a vowel [i] with old-fashioned and legendary properties.

In the case of vowel [u], based on the same principles, Tongue Height: High and Backness are measured and identified as the corresponding vowel. In the case of [ ] , based on the same principles, Tongue Height: Low)r and Tongue Frontness are measured and identified as the corresponding vowel.

In FIG. 23 , the oral tongue sensor 110 measures vowels such as [i], [u], and [ ] generated according to the speaker's utterance as the speech feature 220 . The speech feature 220 of such a vowel corresponds to the phoneme unit index 361 of the consonant vowel of the database unit 350 .

24 is a diagram illustrating a case in which the data analysis unit of the utterance realization system uses Alveolar Stop according to the second embodiment of the present invention.

As shown in FIG. 24 , the oral tongue sensor 110 measures a specific consonant uttered by the speaker as a speech feature 220 . The speech characteristics 220 of these consonants correspond to the phoneme unit index 361 of the consonants and vowels of the database unit 350 , and the data analysis unit 200 recognizes them as Alveolar Stop, which is the language data 310 .

25 is a diagram illustrating a case in which the data analysis unit of the utterance realization system uses Bilabial Stop according to the second embodiment of the present invention.

25 , the oral tongue sensor 110 and the facial sensor 120 measure a specific consonant uttered by a speaker as a speech feature 220 . The utterance characteristics 220 of these consonants correspond to the phoneme unit index 361 of the consonants of the database unit 350 , and the data analysis unit 200 recognizes this as a bilabial stop that is the language data 310 .

26 is a diagram illustrating an experimental result using Voiced Bilabial Stop by the data analysis unit of the speech realization system according to the second embodiment of the present invention.

As shown in FIG. 26 , the oral tongue sensor 110 and the facial sensor 120 measure a specific consonant uttered by a speaker as a speech feature 220 . The speech characteristics 220 of these consonants correspond to the phoneme unit index 361 of the consonant vowels of the database unit 350, and the data analysis unit 200 uses the language data 310, Voiced Bilabial Stop, /bur/ and It was identified as /Pur/, which is a Voiceless Bilabial Stop.

27 and 28 are diagrams illustrating a case in which the data analysis unit of the speech realization system according to the second embodiment of the present invention utilizes Labiodental Fricatives, respectively.

27 and 28 , the oral tongue sensor 110 , the facial sensor 120 , and the voice acquisition sensor 130 . The vocal cord sensor 140 and the tooth sensor 150 measure a specific consonant uttered by the speaker as the speech feature 220 . The speech characteristics 220 of these consonants correspond to the phoneme unit index 361 of the consonants of the database unit 350, and the data analysis unit 200, in the case of FIG. , in the case of Figure 28 is identified as Voiced Labiodental Fricative.

29 is a diagram illustrating the interworking of the data analysis unit and the database of the speech realization system according to the second embodiment of the present invention.

As shown in FIG. 29 , the imaging sensor 160 includes the speaker's oral tongue sensor 110 , a facial sensor 120 , and a voice acquisition sensor 130 . Information on the position of the articulatory organ of the head and neck that occurs when one or more of the vocal cord sensor 140 and the tooth sensor 150 is used in a situation where one or more of them is used (161), information on the change in facial expression of the head and neck (162), non-verbal expression information (163) is recognized as voice data 310 and processed.

In particular, the facial sensor located on one surface of the head and neck region has a potential difference between the positive sensor 122 and the negative sensor 123 with respect to the reference sensor 121 and provides its own position, which is captured by the imaging sensor 160 . The physical head and neck articulatory organ position change information 161 , head and neck expression change information 162 , and non-verbal expression information 163 are transmitted to the data conversion unit 300 as verbal data 310 .

30 is a diagram illustrating a case in which the data analysis unit of the speech realization system according to the second embodiment of the present invention recognizes a specific word.

30, the oral tongue sensor 110, the facial sensor 120, the voice acquisition sensor 130, the vocal cord sensor 140, and the dental sensor 150 detect specific consonants and vowels uttered by the speaker. Measuring, the data analysis unit 200 identifies consonants and vowels as speech characteristics 220 . [b], [i], and [f], which are the utterance features 220 of each consonant, respectively, correspond to the phoneme unit index 361 of the consonant in the database unit 350, and the data analysis unit converts it to /beef/ or the word [bif].

31 is a diagram illustrating a database unit of an utterance realization system according to a second embodiment of the present invention.

31, the language data index 360 of the database unit 350 includes a phoneme unit index 361 of consonants, a syllable unit index 362, a word unit index 363, and a phrase unit index 364. ), sentence unit index 365, continuous utterance index 366, and pronunciation high and low index 367.

32 is a diagram illustrating an utterance realization system according to a third embodiment of the present invention.

As shown in Fig. 32, when one or more of the data analysis unit 200 and the data expression unit (500 in Fig. 34) is located outside and operates, a communication unit 400 capable of interworking and communication is included. The communication unit 400 is implemented as wired and wireless, and in the case of wireless, various methods such as Bluetooth, Wi-Fi, 3G, 4G, NFC, etc. may be used.

33 and 34 are diagrams showing the actual form of the database unit of the utterance realization system according to the third embodiment of the present invention, respectively.

As shown in FIGS. 33 and 34 , the database unit 350 interlocked with the data analysis unit 200 has a language data index, and the utterance characteristics 220 according to the actual utterance, the speaker's voice 230, and the utterance history The information 240 and the utterance variation 250 are identified as the language data 310 .

33 shows that the sensor unit 100 measures various consonant speech characteristics including the High Front tense Vowel and the High Back tense Vowel of FIG. 23, the Alveolar Sounds of FIG. 24, and the Voiceless labiodental fricative of FIG. 27, and the data analysis unit ( 200) is the actual data of the database unit 350 reflected.

34 is a database in which the sensor unit 100 measures various consonant speech characteristics, including the High Front lax Vowel of FIG. 23, the Alveolar Sounds of FIG. 24, and the Bilabial Stop Sounds of FIG. 25, and the data analysis unit 200 reflects This is the actual data of the part 350 .

35 is a diagram illustrating a speech realization system according to a fourth embodiment of the present invention, and FIG. 36 is a sensor unit, data analysis unit, data expression unit, and text information unit of the speech realization system according to the fourth embodiment of the present invention. and a diagram showing the interworking of the database unit.

35, the speech implementation system according to the fourth embodiment of the present invention is a sensor unit 100, a data analysis unit 200, a data conversion unit 300, a database unit ( 350), a data expression unit 500 and a text information unit 600 are included. A detailed description thereof will be provided later with reference to FIGS. 52, 53 and 54 .

As shown in FIG. 36 , the sensor unit 100 is located in the actual articulation organ, measures the physical characteristics of the articulation organ according to the speaker's utterance, and transmits it to the data analysis unit 200, and the data analysis unit 200 interpreted as language data. The interpreted language data is transmitted to the data expression unit 500, and according to the processing step S700 from the speaker data acquisition (S710) to the data generation step (S760) of the text information unit 600, the voice information 640 is the data It is transmitted to the expression unit 500 . It can be seen that the database unit 350 is interlocked in the process of interpretation and expression of the language data 310 and the voice information 640 .

37 to 41 are diagrams showing means for expressing language data by the data expression unit of the speech realization system according to the fourth embodiment of the present invention, respectively.

37 to 41, the physical characteristics of the speaker's head and neck articulatory organ acquired by the sensor unit 100 are analyzed through the data analysis unit 200 to determine the position 210 of the sensor unit, the utterance characteristics 220, It is identified as the speaker's voice 230 , utterance history information 240 , and utterance variation 250 .

The imaging sensor 160 captures a change in the appearance of the speaker's head and neck articulation organs, and the data analysis unit 200 grasps the position change information 161 of the speaker's head and neck articulation organs and information on the head and neck facial expression change 162 through this. .

Thereafter, such information is converted into language data 310 in the data analysis unit 200 , and is externally expressed in the data expression unit 500 .

At this time, FIG. 37 shows that the data expression unit 500 audibly expresses the language data 310 in the form of voice information 640 according to the linkage with the text information unit, and FIG. 38 shows the data expression unit ( When 500) visually expresses the language data 310, the physical characteristics of the speaker's articulation organ measured by the data analysis unit 200 are compared with the language data index 360 of the database unit 350, and the actual standard It represents providing a numerical value that measures one or more of stress, similarity, and intention to speak together with a broad description of pronunciation.

39 shows the language data of the database unit 350 showing the physical characteristics of the speaker's articulation organ measured by the data analysis unit 200 when the data expression unit 500 visually and aurally expresses the language data 310 . Compared with the index 360, it is shown that a numerical value measuring one or more of stress, similarity, and intention to speak is provided together with a narrow description of the actual standard pronunciation.

40 is a language data index 360 of the database unit 350 showing the physical characteristics of the speaker's articulation organ measured by the data analysis unit 200 when the data expression unit 500 visually expresses the language data 310. ), a numerical value measuring one or more of stress, similarity, and utterance intent along with a narrow description of the actual standard pronunciation, and the corresponding language data 310 as a word, a word unit index 363 If it corresponds to , it indicates that the corresponding image is provided together.

FIG. 41 shows the physical characteristics of the speaker's articulation organ measured by the data analysis unit 200 when the data expression unit 500 visually and aurally expresses the language data 310 as the language data of the database unit 350 Compared with the index 360, it provides a numerical value that measures one or more of stress, similarity, and utterance intent along with a narrow description of the actual standard pronunciation, and corrects the language data 310 by the speaker, This indicates that a speech correction image capable of uttering the corresponding pronunciation is provided together so that the corresponding pronunciation can be reinforced.

42 is a diagram illustrating a case in which the data expression unit of the speech realization system according to the fourth embodiment of the present invention expresses language data visually and audibly.

As shown in FIG. 42 , in the data expression unit 500 visualizing the language data 310 as text and providing it as audio information 640 according to the text information unit 600 , the data analysis unit ( 200) measured the physical characteristics of the speaker's articulatory organs in the database unit 350, consonant-vowel phoneme unit index 361, syllable unit index 362, word unit index 363, phrase unit index 364, sentence It compares with one or more language data indexes 360 among the unit indexes 365 .

In this language data 310, the data expression unit 500 measures one or more of stress, similarity, and utterance intent together with a narrow description of the actual standard pronunciation related to the speaker's language data 310. By providing text and sound, it helps the speaker to correct and reinforce the language data 310 .

43 is a diagram illustrating a case in which the data expression unit of the speech realization system visually expresses language data according to the fourth embodiment of the present invention.

As shown in FIG. 43, the data expression unit 500 visualizes and provides the language data 310 as one or more of text, pictures, photos, and images.

At this time, the data analysis unit 200 calculates the measured physical characteristics of the speaker's articulation organ in the consonant and vowel phoneme unit index 361, syllable unit index 362, word unit index 363, phrase unit of the database unit 350. It is compared with the language data index 360 of one or more of the index 364 and the sentence unit index 365.

In addition, when visualized as text, both a narrow description and a broad description of the actual standard pronunciation are provided. Through this, the language data 310 is expressed by the data expression unit 500 along with a narrow description and a broad description of the actual standard pronunciation related to the speaker's language data 310. Stress, similar proximity, It helps the speaker to correct and reinforce the language data 310 by providing one or more of the intentions of speech as measured characters and sounds.

44 is a diagram illustrating a case in which the data expression unit of the speech realization system visually expresses language data according to the fourth embodiment of the present invention.

As shown in FIG. 44 , when the data expression unit 500 visualizes and provides the language data 310 as text, the database unit 350 records the physical characteristics of the speaker's articulation organ measured by the data analysis unit 200 . ) of consonant-vowel phoneme unit index (361), syllable unit index (362), word unit index (363), phrase unit index (364), sentence unit index (365), and continuous speech index (366). Compare with index 360. This language data 310 is expressed by the data expression unit 500 along with a narrow description and a broad description of the actual standard pronunciation related to the speaker's language data 310 of stress, similarity, and utterance. It helps the speaker to correct and reinforce the language data 310 by providing characters and sounds of continuous speech units in which one or more of the intentions are measured.

45 is a diagram illustrating a case in which the data expression unit of the speech implementation system according to the fourth embodiment of the present invention expresses language data in units of continuous speech.

As shown in FIG. 45 , when the data expression unit 500 visualizes the language data 310 as text and provides it as auralized sound, the data analysis unit 200 measures the physical characteristics of the speaker's articulation organ. Consonant vowel phoneme unit index 361, syllable unit index 362, word unit index 363, phrase unit index 364, sentence unit index 365, continuous speech index 366 of the database unit 350 , and compares it with one or more language data indexes (360) among the high and low indexes (367) of pronunciation. This language data 310 is expressed by the data expression unit 500 along with a narrow description and a broad description of the actual standard pronunciation related to the speaker's language data 310 of stress, similarity, and utterance. It helps the speaker to correct and reinforce the language data 310 by providing one or more of the intentions as measured characters and sounds.

46 is a diagram illustrating the confusion matrix utilized by the utterance realization system according to the fourth embodiment of the present invention, and FIG. it is one drawing

As shown in FIGS. 46 and 47 , the data analysis unit 200 extracts one or more features using the Time Domain Variance, the Frequency Domain Cepstral Coefficient, and the Linear Predict Coding Coefficient in recognizing the language data 310 . It is used to represent the algorithm.

는 수집된 조음기관 물리 특성 인 데이터의 모집단의 평균, xi는 수집된 조음기관 물리 특성인 데이 터들을 나타낸다.Variance representing the degree of dispersion of data is calculated according to the following [Equation 1]. where n is the network of the population,

is the average of the population of the collected articulation organ physical characteristics data, and xi is the collected articulatory organ physical characteristics data.

[Equation 1]

The Cepstral Coefficient is calculated by the following [Equation 2] to standardize the intensity of the frequency. Here, F-1 denotes an Inverse Fourier Transform, which is an inverse Fourier series transformation, and X(f) denotes a spectrum of a frequency with respect to a signal. In this example, Cepstral's Cofficent was used when n=0.

[Equation 2]

The Linear Predict Coding Coefficient indicates the linear characteristic of the frequency and is calculated according to the following [Equation 3]. Here, n represents the number of samples, and ai is the Linear Predict Coding Coefficient coefficient. Cepstral's coefficient was used when n=1.

[Equation 3]

In addition, ANN was used to classify each data by grouping the data, which is the physical characteristic of articulatory organs, according to the degree of similarity and generating predictive data. Through this, it is possible for the speaker to grasp the close similarity and utterance intention of the content of his utterance in preparation for the standard utterance with respect to the initial utterance. Based on this, the speaker obtains feedback on the content of the utterance and continuously repeats the utterance to correct the utterance. Through this iterative articulation organ physical property data input method, a lot of articulation organ physical property data is collected and the accuracy of ANN is increased.

Here, the physical characteristics of articulatory organs as input data were selected as 10 consonants, and in the extraction process, they were classified into 5 articulation positions, Bilabial, Alveolar, Palatal, Velar, and Glottal. To this end, 10 consonants corresponding to the five articulation positions were pronounced 100 times in sequence, 100 times in total, and randomly 50 times a total of 500 times.

Accordingly, as shown in FIG. 45, a confusion matrix for classifying consonants was formed. Based on this, considering that the number of consonants uttered at each articulation position is different, as shown in FIG. 46, it is expressed as a percentage.

Through this, it can be seen that the speaker does not properly utter consonants in relation to Palatal, which has low pronunciation and close similarity compared to standard utterances. Also, as shown in FIG. 46 , 17% of the cases in which an attempt was made to utter a consonant related to Palatal but erroneously uttered a consonant related to Alveolar. This means that the speaker does not clearly recognize the difference between the consonants associated with Palatal and those associated with Alveolar.

48 is a diagram illustrating a case in which the speech realization system according to the fourth embodiment of the present invention assists the speaker with language correction and guidance through the screen.

As shown in FIG. 48 , a Korean speaker who does not speak English as a native language intended to utter [kiŋ], and the sensor unit 100 detects the physical characteristics of the articulatory organ according to the utterance.

However, in the case of the speaker, he was inexperienced in articulation and utterance methods for [ŋ], a Velar Nasal Sound that does not exist in Korean.

Accordingly, the data analysis unit 200 identifies [ŋ], which the speaker has not uttered properly, through comparison with the standard speech feature matrix 205 . After that, the data expression unit 300 provided the speaker's utterance and similarity, and the result was only 46%. Thereafter, the data expression unit 300 helps the speaker to correctly pronounce [kiŋ] through a screen or the like.

In this case, the data expression unit 300 provides a Speech Guidance (Image) to intuitively show which articulation organ the speaker should manipulate and how. The Speech Guidance (Image) presented by the data expression unit 300 performs speech correction and guidance based on the sensor unit attached to or adjacent to the articulation organ for uttering the [ŋ]. For example, in the case of [kiŋ], [k] elevates and attaches the back part of the tongue (Tongue Body, Root) in the direction of Velar (soft palate) and makes a plosive sound while pacing, and is a silent sound without vibration of the vocal cords, through the mouth It should be uttered with /k/.

At this time, the oral tongue sensor 110 measures that the rear part of the tongue raises and lowers in the Velar (soft palate) direction and makes a plosive sound. In the case of [i], since it is a Legendary High Front Tense Vowel, the oral tongue sensor 110 also detects Hight and Frontness of the tongue. In addition, when [i] is uttered, both ends of the lips are pulled towards both cheeks, respectively. The facial sensor 120 will grasp this. In the case of [ŋ], the rear part of the tongue (Tongue Body, Tongue Root) must be raised and lowered in the direction of Velar (soft palate) and the nose must be sounded to fire. Therefore, the oral tongue sensor 110 also detects the tongue's high low tongue and anterior and posterior tongue characteristics.

In addition, since the pronunciation is a nasal sound, the nose and surrounding muscles tremble. This phenomenon can be detected by attaching the facial sensor 120 around the nose.

49 is a diagram illustrating a case in which the speech realization system according to the fourth embodiment of the present invention captures and detects trauma to the head and neck articulatory organ.

49, the imaging sensor 160 captures a change in the appearance of the speaker's head and neck articulatory organ according to the utterance, and the data analysis unit 200 uses this information to change the position of the speaker's head and neck articulatory organ 161 , the head and neck expression change information 162 is grasped. At this time, the speech characteristics 210 of the speaker identified through the oral tongue sensor 110 , the facial sensor 120 , the voice acquisition sensor 130 , the vocal cord sensor 140 , and the tooth sensor 150 of the sensor unit 100 are also Together with the data analysis unit 200 is considered.

50 is a diagram illustrating a case in which the speech implementation system according to the fourth embodiment of the present invention combines mutual information through a standard speech feature matrix.

As shown in FIG. 50 , the oral tongue sensor 110 , the facial sensor 120 , the voice acquisition sensor 130 , and the vocal cord sensor 140 of the sensor unit 100 . The dental sensor 150 detects the speaker's speech characteristics 210 , and the imaging sensor 160 detects the speaker's head and neck articulation organ position change information 161 and head and neck facial expression change information 162 . Through this, the data analysis unit 200 combines the speech characteristics corresponding to the position change information 161 of the head and neck articulatory organs and the head and neck expression change information 162 based on the standard speech feature matrix 205 .

As shown in FIG. 51 , the imaging sensor 160 provides not only information on changes in the speaker’s head and neck articulatory organs 161 and information on changes in head and neck facial expressions 162, but also non-verbal expressions 153 that the speaker expresses while speaking. take a picture That is, the non-verbal expression 153 is the inclination of the head and neck that occurs according to the speaker's intention to speak, the shaking of the thoracic region, the tension of the joint and muscle between the head and neck and the thoracic region, the tremor of the upper extremities, the gesture of the upper extremities, and the gesture of the lower extremities. It includes expressions of one or more of tremors and gestures of the lower extremities.

52, the text information unit 600 outputs the language data as voice information, characterized in that it goes through a speaker data processing step (S700). The steps are as follows.

In step (a), the data conversion unit acquires speaker data for acquiring at least one language data. In steps (S710) and (b), based on the word unit index 363 of the language data index of the database unit, at least one function word 611 to blank in the language data is labeled with 0 to NULL and excluded, and at least one A speaker content table 615 is generated by labeling the speaker content word 610 of the data analysis unit 200 with the standard utterance feature matrix 205 . (S720)

The function words 611 are words such as /the/, /to/, and /of/ among the previously acquired speaker language data 310 . The function word 611 does not generate a meaning in a sentence, and means a proposition and ending that cannot be used independently in a sentence. On the other hand, the content word is a minimum unit of a word that can be used independently in a sentence, including a noun, a verb, an adjective, an adverb, and an adjective, among the previously acquired speaker's language data 310 . For example, in "The postman brings the letter." measured in FIG. 34, the function word is /the/, and the content word becomes /postman/, /brings/, and /letter/.

In step (c), onset (initial), Necleus (neutral), Coda ( The phoneme value matrix 620 is generated by separating at least one consonant-vowel phoneme unit of the final consonant). (S730) In step (d), corresponding to the phoneme value matrix 620 generated in step (c), the utterance progress time value of the phoneme unit index 361 and F0, F1, F2 according to the utterance A voice value matrix 630 including at least one formant frequency value of , F3, and F4 and at least one of an utterance amplitude value is generated (S740).

At this time, the phoneme unit index 361 of the language data index 360 is classified into Onset (initial consonant = consonant), Necleus (neutral = vowel), and Coda (final consonant = consonant) for a word having at least one syllable. , and the syllable unit index 362 includes a waveform and a spectrogram for each syllable according to the classification. Here, the Waveform is composed of an x-axis indicating the time of utterance and a y-axis indicating an amplitude indicating the intensity according to the utterance, and is information indicating a change in the utterance amplitude according to the utterance time. The spectrogram is composed of the x-axis representing the time of speech, the y-axis representing the frequency (Hz) according to the speech, and the z-axis (contrast) representing the intensity representing the intensity according to the speech. In the spectrogram, different fundamental frequencies for each consonant, F0 (Hz), and formant frequencies, F1 (Hz), F2 (Hz), and F3 (Hz) can be identified. In the case of F1, it is related to the low height of the vowel, and when it is a high vowel, F1 is formed low, and when it is a low vowel, F1 is formed high. In the case of F2, it is related to the frontness and backness of the vowel, so the higher the front vowel, the lower the back vowel. In the case of a circular hindlingual vowel, F2 tends to be low. For example, in FIG. 52 , among vowels of tr], [ ] , which is the first vowel in bold, corresponds to (4, 2) according to the standard speech feature matrix 205 as a middle front voice.

Some examples of the frequency table of the phoneme unit index 361 shown in FIG. 53 indicate a part of the increase in the average value of F0 to F4 mentioned above. Here, M stands for Man, W stands for Woman, and C stands for Child. For example, among the vowels of FIG. 53 , [ ] , which is the first vowel indicated in bold, is a middle front voice and corresponds to (4, 1) according to the standard speech feature matrix 205 .

Accordingly, in [ ] , in the case of adult male speakers, F0 is 127Hz, F1 is 580Hz, F2 is 1799Hz, and F4 is 3677Hz. As such, each phoneme uttered by the speaker has a different frequency, and through this, each phoneme is distinguished from each other.

Thereafter, at least one of the voice value matrices 630 generated in step (d) corresponding to at least one element of the phoneme value matrix 620 generated in c, and the element of the phoneme value matrix 620 . Allocate and merge the elements of (S750). Based on this, in step (f), at least one element that has passed through step (e) is recombined to generate at least one voice information 640 among phoneme units, word units, phrase units, sentence units, and continuous speech units. . (S760)

In addition to the method described in the drawings, the following may be included in the case of the sensor unit 100 .

1. Pressure sensor: MEMS sensor, piezoelectric (pressure-voltage) method, piezoresistive (pressure-resistance) method, capacitive method, pressure sensitive rubber method, force sensing resistor (FSR) method, inner particle deformation method, buckling measurement method.

2. Friction sensor: micro hair array method, friction temperature measurement method.

3. Static sensor: static electricity consumption method, static electricity generation method.

4. Electrical resistance sensor: DC resistance measurement method, AC resistance measurement method, MEMS, Lateral electrode array method, Layered electrode method, Field Effect Transistor (FET) method (Organic-FET, Metal-oxide-semiconductor-FET, Piezoelectric-oxide -semiconductor -FET, etc.).

5. Tunnel Effect Tactile Sensor: Quantum tunnel composites method, Electron tunneling method, Electroluminescent light method.

6. Thermal resistance sensor: thermal conductivity measurement method, thermoelectric method.

7. Optical sensor: light intensity measurement method, refractive index measurement method.

8. Magnetism based sensor: Hall-effect measurement method, Magnetic flux measurement method.

9. Ultrasonic based sensor: Acoustic resonance frequency method, Surface noise method, Ultrasonic emission measurement method.

10. Soft material sensor: rubber, powder, porous material, sponge, hydrogel, airgel, carbon fiber, nano-carbon material, carbon nanotube, graphene, graphite, composite material, nano-composite, metal-polymer composite, ceramic-polymer composite , a method of measuring pressure, stress, or strain using materials such as conductive polymers, and a Stimuli responsive method.

11. Piezoelectric material sensor: Ceramic materials such as Quartz and PZT (lead zirconate titanate), polymer materials such as PVDF, PVDF copolymers, PVDF-TrFE, and nano materials such as cellulose and ZnO nanowires.

100: sensor unit 110: oral tongue sensor
120: facial sensor 130: voice acquisition sensor
140: vocal cord sensor 150: tooth sensor
200: data analysis unit 205: standard speech feature matrix
210: position of the sensor unit 220: ignition characteristics
230: speaker's voice 240: utterance history information
250: utterance variation 300: data conversion unit
310: language data 350: database unit
360: Language data index
400: communication unit 500: data expression unit
600: text information unit 610: speaker content
615: speaker content table 620: phoneme value matrix
630: voice value matrix 640: voice information

Claims (19)

a sensor unit adjacent to one surface of the speaker's head and neck for measuring physical characteristics of the articulation organ;
a data analysis unit configured to identify the speaker's utterance characteristics based on the position of the sensor unit and the physical characteristics of the articulation organ;
a data conversion unit converting the position of the sensor unit and the speech characteristics into language data;
a text information unit generating the language data of the data conversion unit as voice information;
a data expression unit interworking with the text information unit to externally express the voice information based on language data;
The sensor unit includes an oral tongue sensor corresponding to the oral tongue,

The data analysis unit,
Through the physical characteristics of the oral tongue and other articulation organs measured by the sensor unit, the utterance characteristics of at least one of consonant vowels, lexical stress, and tonic stress, uttered by the speaker, are identified. and,
In recognizing the speech characteristics according to the physical characteristics of the articulatory organ measured by the sensor unit, the similarity of pronunciation and stress of the speaker based on a standard speech characteristic matrix composed of numeric values including binary numbers or real numbers; Measure at least one speech characteristic of speech intent,
The location of the sensor unit, the speaker's utterance characteristics, and at least one linguistic data index corresponding to the speaker's voice is linked with the database unit,

The data expression unit,
By interworking with the language data index of the database unit, the speaker's utterance characteristics are determined by phoneme units of consonants and vowels, at least one word unit, at least one phrase unit (Citation Forms), at least one sentence unit, and continuous utterance unit. A speech implementation system representing at least one speech expression of (Consecutive Speech).
According to claim 1, wherein the oral tongue sensor,
It is fixed to one side of the oral tongue, wraps around the surface of the oral tongue, or is inserted into the oral tongue, and the amount of vector according to time based on the x-axis, y-axis, and z-axis direction of the oral tongue according to the utterance An utterance realization system for grasping at least one physical characteristic of low elevation, anterior-posterior tongue, flexion, extension, rotation, tension, contraction, relaxation, and vibration of the oral tongue by grasping the amount of change.
According to claim 1, wherein the oral tongue sensor,
It is fixed to one side of the oral tongue, wraps around the surface of the oral tongue, or is inserted into the oral tongue, and rotates per unit time based on the x-axis, y-axis, and z-axis direction of the oral tongue according to the utterance. A speech implementation system for grasping the amount of change in the angle, and identifying the physical characteristics of the articulatory organ, including the oral tongue.
According to claim 1, wherein the oral tongue sensor,
An electrical signal corresponding to the polarization caused by a change in the crystal structure according to the physical force that is adhered to one side of the oral tongue or surrounds the surface of the oral tongue, and is generated by the contraction and relaxation of the oral tongue according to ignition is generated. By grasping the degree of bending of the oral tongue through a piezoelectric element, the physical characteristics of at least one of low elevation, anteroposterior, flexion, extension, rotation, tension, contraction, relaxation, and vibration of the oral tongue an utterance implementation system.
According to claim 1, wherein the sensor unit,
At least one of the physical properties of the degree of rupture, friction, resonance, and access according to the tribo electric generator corresponding to the approach and contact caused by the interaction of the oral tongue with other articulatory organs inside and outside the head and neck An ignition realization system including a triboelectric charging element to grasp.
According to claim 1, wherein the data analysis unit,
In recognizing the speech characteristics of the physical characteristics of the articulation organ measured by the sensor unit, the steps of recognizing the physical characteristics of the articulation organ as a pattern of each consonant unit, extracting the characteristics of the pattern of the consonant unit, and , classifying the extracted characteristics of the pattern of the consonant unit according to the degree of similarity, recombining the characteristics of the classified pattern of the consonant unit, and interpreting the physical characteristics of the articulation organ as the speech characteristics. An utterance realization system for recognizing the utterance characteristics.
According to claim 1, wherein the data analysis unit,
According to the physical characteristics of the articulation organ measured by the sensor unit, assimilation, dissimilation, elimination, attachment, stress, and reduction of consonant vowels Caused Asperation, Syllabic cosonant, Flapping, Tensification, Labilalization, Velarization, Dentalizatiom, Palatalization A speech realization system that measures speech variation, which is a secondary articulation phenomenon of at least one of , Nasalization, Stress Shift, and Lengthening. The method of claim 1,
The oral tongue sensor is an utterance implementation system comprising a circuit part for sensor operation, a capsule part surrounding the circuit part, and an adhesive part attached to one surface of the oral tongue.
The method of claim 1,
The oral tongue sensor, in the form of a film having a thin film circuit, a speech implementation system that operates adjacent to the oral tongue.
The method of claim 1,
The sensor unit includes at least one reference sensor for generating a reference potential for measuring a nerve signal of the head and neck muscles, and at least one positive sensor and at least one cathode sensor for measuring the nerve signal of the head and neck muscle. speech implementation system.
11. The method of claim 10,
The data analysis unit, in obtaining the position of the sensor unit based on the face sensor, grasps the potential difference between the at least one positive sensor and the at least one negative sensor based on the reference sensor to determine the position of the facial sensor An utterance implementation system to identify.
11. The method of claim 10,
The data analysis unit, when acquiring the speech characteristics of the speaker based on the face sensor, recognizes a potential difference between the at least one positive sensor and the at least one negative sensor with respect to the reference sensor to determine the head and neck of the speaker A speech implementation system for identifying speech characteristics caused by the physical characteristics of the articulatory organ.
The method of claim 1,
The sensor unit may include a vocal cord sensor that detects electromyography or tremor of the vocal cords adjacent to the vocal cords of the speaker's head and neck, and grasps information on the speech history of at least one of the speaker's speech start, speech stop, and speech end. system.
The method of claim 1,
The sensor unit, adjacent to one surface of the tooth, the speech implementation system comprising a tooth sensor for detecting a signal generation position according to a change in electrical capacity generated due to the contact between the oral tongue and the lower lip.
The method of claim 1,
The data analysis unit may be configured to acquire the speaker's voice according to the utterance through a voice acquisition sensor adjacent to one surface of the speaker's head and neck.
The method of claim 1,
In the sensor unit, the tilting of the head and neck, up, down, left, and right, which occurs according to the speaker's intention to speak, shaking of the thorax, the tension between the joint and muscle between the head and neck and the rib cage, the tremor of the upper extremities, the gestures of the upper extremities, the tremors of the lower extremities, the gestures of the lower extremities The speech realization system of claim 1, further comprising: an image sensor configured to capture an image of the speaker's head and neck in order to detect at least one non-verbal expression.
The method of claim 1,
Speech implementation system further comprising a power supply for supplying power to at least one of the oral tongue sensor, the facial sensor, the voice acquisition sensor, the vocal cord sensor, the tooth sensor, and the imaging sensor of the sensor unit.
The method of claim 1,
The speech expression displayed by the data expression unit is visualized as at least one of a character symbol, a picture, a special symbol, and a number, or is auralized in the form of a sound, and is provided to the speaker and the listener.
The method of claim 1,
The speech expression displayed by the data expression unit is provided to the speaker and the listener in at least one tactile method of vibration, snoozing, tapping, pressing, and relaxation.

Images