System and method for diagnosing articulation disorders

Abstract

The present invention relates to an automatic obstruent sound generation system for diagnosing articulation disorders, comprising: a word input unit in which a reference word and a language code of said reference word are input; a phoneme sequence conversion unit linked to said word input unit and decomposing said reference word into phoneme units; and a feature annotation unit linked to said phoneme sequence conversion unit and, by referring to a language-specific articulatory feature mapping table selected in correspondence with said language code, assigning consonant articulatory feature information including the place and method of articulation in the case of a consonant, and assigning vowel articulatory feature information including the height of the tongue according to the vowel angle, the anterior-posterior position of the tongue, and roundness in the case of a vowel, for each phoneme decomposed from said reference word, thereby generating phoneme-specific feature annotation information. A phonological change rule matching unit linked to the above feature annotation unit and, according to language-specific phonological change rules, detects phonemes among the decomposed phonemes to which the language-specific phonological change rules are applicable by utilizing consonant articulation feature information or vowel articulation feature information included in the phoneme-specific feature annotation information, and detects candidate phonemes that can be substituted, omitted, distorted, or added to the applicable phonemes; a candidate combination unit linked to the above phonological change rule matching unit and generates obstruent candidate words by combining the detected candidate phonemes; a verification unit linked to the above candidate combination unit and determines the final obstruent candidate words based on syllable structure rules and phoneme combination rules for the generated obstruent candidate words; an alignment unit linked to the above verification unit and analyzes the correspondence relationship by aligning the reference word and the final obstruent candidate words; a labeling unit linked to the above alignment unit and automatically labels detailed phonological change codes for each of the final obstruent candidate words; and an output unit linked to the above labeling unit and provides a list of the final obstruent candidate words.According to the present invention, by automatically generating and verifying candidate obstruent words of normal words that can be converted according to phonological change rules for each language, the process of diagnosing articulation disorders can be automated, and the deviation between examiners is significantly reduced compared to cases relying on subjective listening or manual transcription, thereby having the effect of significantly improving the objectivity and consistency of the judgment of articulation disorders.

Description

Technology Field

[0001] The present invention relates to a system and method for automatically generating obstruent sounds for diagnosing articulation disorders, and more specifically, to a system and method for automatically generating obstruent sounds based on language-specific phonological rules to assist in determining whether a speaker has an articulation disorder. Background Technology

[0002] Recently, there has been a growing need for automation technology to efficiently diagnose articulation disorders in the fields of speech therapy and speech processing.

[0003] Articulation disorder refers to a condition in which specific phonemes cannot be produced normally due to abnormalities in articulatory organs such as the tongue, lips, and soft palate, or inaccuracies in articulatory movements, which frequently results in pronunciation errors such as substitution, omission, distortion, and addition.

[0004] Traditional diagnosis of articulation disorders has primarily involved speech therapists in clinical settings directly listening to and transcribing the subject's speech and determining the type of error by comparing it to normal target words. This manual-based diagnosis has limitations, such as being time-consuming and lacking consistency and objectivity in diagnostic results due to varying judgment criteria among examiners.

[0005] Meanwhile, some automated systems receive speech data from test subjects and perform acoustic analysis and statistical error detection; however, since they primarily rely on acoustic characteristics to perform only simple similarity judgments or fail to sufficiently reflect phonological rules or articulatory features, there is a problem in that it is difficult to provide a clinically interpretable classification of error types.

[0006] Furthermore, since most existing systems rely on the actual speech data of test subjects, there is a problem in that it is difficult to accumulate large-scale articulation disorder patterns by language or to construct obstruent datasets that can be scalable in multilingual environments.

[0007] In other words, conventional technologies for diagnosing articulation disorders suffer from a low level of automation in the analysis process, the inability to systematically diagnose errors reflecting language-specific phonological rules, and a lack of a technical foundation to automatically generate and label large-scale obstruent data.

[0008] Accordingly, there is a need for technology that can improve the objectivity and efficiency of the articulation disorder diagnosis process by receiving a normal word as input, automatically generating possible obstruent candidates using language-specific phonological change rules and articulation feature information, and aligning and analyzing them with a reference word to automatically calculate error types and detailed phonological change codes. Prior art literature

[0009] Republic of Korea Registered Patent No. 10-2860659 The problem to be solved

[0010] The objective of the present invention is to solve the aforementioned conventional problems and to provide an automatic obstruent sound generation system and method for diagnosing articulation disorders, which can automatically generate obstruent sounds based on language-specific phonological rules to assist in determining whether a speaker has an articulation disorder. means of solving the problem

[0011] The above objective is achieved, according to the present invention, by a word input unit in which a reference word and a language code of the reference word are input; a phoneme sequence conversion unit linked to the word input unit and decomposing the reference word into phoneme units; a feature annotation unit linked to the phoneme sequence conversion unit and, by referring to a language-specific articulatory feature mapping table selected in correspondence with the language code, assigns consonant articulatory feature information including the place of articulation and method of articulation in the case of a consonant, and assigns vowel articulatory feature information including the height of the tongue according to the vowel angle, the fore-aft position of the tongue, and roundness in the case of a vowel, to each phoneme decomposed from the reference word, thereby generating phoneme-specific feature annotation information; and a phonological change rule matching unit linked to the feature annotation unit and, according to language-specific phonological change rules, detects phonemes among the decomposed phonemes to which the language-specific phonological change rules are applicable by using the consonant articulatory feature information or the vowel articulatory feature information included in the phoneme-specific feature annotation information, and detects candidate phonemes that can be substituted, omitted, distorted, and added for the applicable phonemes. This is achieved by an automatic obstruent generation system for diagnosing articulation disorders, comprising: a candidate combination unit linked to the above-mentioned phonological change rule matching unit and generating obstruent candidate words by combining detected candidate phonemes; a verification unit linked to the above-mentioned candidate combination unit and determining final obstruent candidate words based on syllable structure rules and phoneme combination rules for the generated obstruent candidate words; an alignment unit linked to the above-mentioned verification unit and analyzing correspondence relationships by aligning the reference word and the final obstruent candidate words; a labeling unit linked to the above-mentioned alignment unit and automatically labeling detailed phonological change codes to each of the final obstruent candidate words; and an output unit linked to the above-mentioned labeling unit and providing a list of the final obstruent candidate words.

[0012] In addition, the above-mentioned phonological change rules by language include syllable omission, syllable addition, initial consonant omission, final consonant omission, stop omission, fricative omission, affricate omission, nasal omission, liquid omission, bilabial omission, alveolar omission, velar omission, glottal omission, alveolar forwarding, palatal forwarding, velar forwarding, glottal forwarding, bilabial backwarding, alveolar backwarding, palatal backwarding, velar backwarding, labialization, alveolarization, palatalization, velarization, glottalization, labiodentalization, interdentalization, stop assimilation, fricativeization, affricateization, liquidization, nasalization, bilabial assimilation, alveolar assimilation, palatal assimilation, velar assimilation, glottal assimilation, stop assimilation, fricative assimilation, affricate assimilation, aspirated assimilation, tense assimilation, tense, laxation, It may include aspiration and deaspiration.

[0013] In addition, the above detailed phonological change code may include a language type code, an error type code, a phonological change rule code, a rule application location code, and a surrounding phonological condition code.

[0014] In addition, the above error type codes can be classified according to error types including substitution, omission, distortion, and addition.

[0015] The above other objectives are, according to the present invention, a word input step of inputting a reference word and a language code of the reference word by a word input unit; a phoneme sequence conversion step of decomposing the reference word into phoneme units by a phoneme sequence conversion unit; a feature annotation step of assigning articulation place and articulation method features to each decomposed phoneme by a feature annotation unit; a phonological change rule matching step of generating candidate phonemes that can be substituted, omitted, distorted, or added by applying language-specific phonological change rules based on the features by a phonological change rule matching unit; a candidate combination step of generating obstruent candidate words by combining the phonemes by a candidate combination unit; a verification step of determining final obstruent candidate words by filtering data among the obstruent candidate words that conform to syllable structure rules and phoneme combination rules by a verification unit; an alignment step of aligning the reference word and the final obstruent word and analyzing the correspondence relationship by an alignment unit; and a labeling step of automatically labeling each of the final obstruent candidate words with a detailed phonological change code by a labeling unit. This is achieved by an automatic generation method for articulation disorder diagnosis that includes an output step providing a list of the final obstruent word candidates by means of an output unit. Effects of the invention

[0016] According to the present invention, by automatically generating and verifying candidate obstruent words of normal words that can be converted according to phonological change rules for each language, the process of diagnosing articulation disorders can be automated, and the deviation between examiners is significantly reduced compared to cases relying on subjective listening or manual transcription, thereby having the effect of significantly improving the objectivity and consistency of the judgment of articulation disorders.

[0017] In addition, according to the present invention, types of articulation errors can be classified based on linguistic grounds, and detailed phonological change codes are automatically assigned for each error type, thereby having the effect of being usable for clinical analysis and the generation of learning data using artificial intelligence.

[0018] In addition, according to the present invention, by adding a language-specific phonological rule library, the diagnosis of articulation disorders can be expanded to multiple languages, such as English and Japanese, in addition to Korean.

[0019] In addition, according to the present invention, when the generated obstruent candidate words and labeling results are stored in a database format, it is possible to secure standardized diagnostic data required in the field of speech therapy, such as clinical evaluation, speech therapy education, and learning of artificial intelligence models for articulation disorders.

[0020] Meanwhile, the effects of the present invention are not limited to those mentioned above, and various effects may be included within the scope obvious to a person skilled in the art from the contents described below. Brief explanation of the drawing

[0021] FIG. 1 briefly illustrates the process of an automatic sound generation system for diagnosing articulation disorders according to an embodiment of the present invention performing a speech error analysis procedure, and FIG. 2 illustrates the overall configuration of an automatic sound generation system for diagnosing articulation disorders according to an embodiment of the present invention, and FIG. 3 illustrates the electrical connections between the components of an automatic sound generation system for diagnosing articulation disorders according to an embodiment of the present invention, and FIG. 4 is a table showing the articulation method and place of articulation for each phoneme decomposed by the phoneme sequence conversion unit of an automatic sound generation system for diagnosing articulation disorders according to an embodiment of the present invention, and FIG. 5 illustrates Korean and English vowel quadrilaterals to which features are assigned by the feature annotation section of an automatic obstruent sound generation system for diagnosing articulation disorders according to an embodiment of the present invention, and FIG. 6 is a flowchart of a method for automatically generating obstruent sounds for diagnosing articulation disorders according to an embodiment of the present invention. Specific details for implementing the invention

[0022] Hereinafter, some embodiments of the present invention will be described in detail with reference to exemplary drawings. It should be noted that in assigning reference numerals to the components of each drawing, the same components are given the same reference numeral whenever possible, even if they are shown in different drawings.

[0023] In addition, when describing embodiments of the present invention, if it is determined that a detailed description of related known configurations or functions would hinder understanding of the embodiments of the present invention, such detailed description is omitted.

[0024] In addition, terms such as first, second, A, B, (a), (b), etc., may be used when describing the components of the embodiments of the present invention. These terms are used merely to distinguish the components from other components, and the essence, order, or sequence of the components is not limited by such terms.

[0025] Additionally, in describing embodiments of the present invention, each "part," "module," or "step" may be implemented through a processor and memory. The term "processor" should be broadly interpreted to include general-purpose processors, central processing units (CPUs), microprocessors, digital signal processors (DSPs), controllers, microcontrollers, and state machines, etc. In some environments, the term "processor" may refer to an application-specific integrated circuit (ASIC), a programmable logic device (PLD), a field programmable gate array (FPGA), etc. The term "processor" may also refer to a combination of processing devices, such as, for example, a combination of a DSP and a microprocessor, a combination of multiple microprocessors, a combination of one or more microprocessors combined with a DSP core, or any other combination of such configurations.

[0026] Additionally, memory should be broadly interpreted to include any electronic component capable of storing electronic information. Memory may also refer to various types of processor-readable media, such as random access memory (RAM), read-only memory (ROM), non-volatile random access memory (NVRAM), programmable read-only memory (PROM), eraseable-programmable read-only memory (EPROM), electrically eraseable PROM (EEPROM), flash memory, magnetic or optical data storage devices, registers, etc. If a processor can read information from memory or write information to memory, memory is said to be in an electronic communication state with the processor, and each "part," "module," or "stage" may be implemented through a program or application based on the processor and memory in an electronic communication state.

[0027] In addition, the communication network may be a wired or wireless communication network and may utilize at least one of LAN (local area network), WLAN (wireless LAN), Wibro (wireless broadband), WiMAX (world interoperability for microwave access), HSDPA (high speed downlink packet access), and various other communication standards. Furthermore, the communication network may be a mobile communication network, and it may be a synchronous mobile communication network or an asynchronous mobile communication network. As an example of the asynchronous mobile communication network, a WCDMA (Wideband Code Division Multiple Access) network may be cited. In this case, although not shown in the drawings, the mobile communication network may include, for example, an RNC (Radio Network Controller). Meanwhile, although the WCDMA network was cited as an example, it may be a next-generation communication network such as a 3GLTE network, 4G network, or 5G network, as well as other low-power communication networks such as WI-FI, Bluetooth, and Zigbee, and wired / wireless IP networks.

[0029] Now, with reference to the attached drawings, an automatic sound generation system (100) for diagnosing articulation disorders according to an embodiment of the present invention will be described in detail.

[0030] FIG. 1 briefly illustrates the process of an automatic sound generation system for diagnosing articulation disorders according to an embodiment of the present invention performing a speech error analysis procedure; FIG. 2 illustrates the overall configuration of an automatic sound generation system for diagnosing articulation disorders according to an embodiment of the present invention; FIG. 3 illustrates the electrical connections between the components of an automatic sound generation system for diagnosing articulation disorders according to an embodiment of the present invention; FIG. 4 is a table showing the articulation method and articulation location for each phoneme decomposed by the phoneme sequence conversion unit of an automatic sound generation system for diagnosing articulation disorders according to an embodiment of the present invention; FIG. 5 illustrates the vowel quadrilaterals of Korean and English to which features are assigned by the feature annotation unit of an automatic sound generation system for diagnosing articulation disorders according to an embodiment of the present invention.

[0031] As illustrated in FIGS. 2 and 3, an automatic sound generation system (100) for diagnosing articulation disorders according to an embodiment of the present invention may include a word input unit (110), a phoneme sequence conversion unit (120), a feature annotation unit (130), a phonological change rule matching unit (140), a candidate combination unit (150), a verification unit (160), an alignment unit (170), a labeling unit (180), and / or an output unit (190).

[0032] The word input unit (110) receives a reference word and a language code of the reference word, and is linked with the user terminal unit and the phoneme sequence conversion unit (120) described later.

[0033] The reference word refers to a word that serves as a criterion for determining whether there is an articulation disorder, and includes, for example, "cat," "rice soup," "processing," etc., but is not limited thereto.

[0034] The user terminal is a device in which a user inputs words into the word input unit (110), and includes, but is not limited to, a desktop PC, laptop PC, smartphone, tablet, UMPC, etc.

[0035] The phoneme sequence conversion unit (120) decomposes a reference word into phoneme units and is linked with the word input unit (110) and the feature annotation unit (130) described later.

[0036] More specifically, the phoneme sequence conversion unit (120) can normalize a reference word input into the word input unit (110) through the user terminal into a format suitable for standard phonetic symbol processing using a G2P (Grapheme-to-Phoneme) conversion algorithm, and then convert the normalized string into a phoneme sequence according to the orthography-to-phoneme rule of the corresponding language. For example, in the case of Korean, the phoneme sequence conversion unit (120) can convert the combination of initial consonant, medial vowel, and final consonant into phoneme units (e.g., ㄱ( / k / ), ㄴ( / n / ), ㅁ( / m / ), ㅏ( / a / ), ㅗ( / o / )), and in the case of English, can convert the spelling sequence into phonetic symbols (e.g., / æ / , / ð / ).

[0037] The feature annotation section (130) is linked with the phoneme sequence conversion section (120) and assigns articulation position, articulation method, and vowel quadrant features to each decomposed phoneme, and is linked with the phoneme sequence conversion section (120) and the phonological change rule matching section (140) described later.

[0038] The feature annotation unit (130) can, more specifically, receive a sequence of phonemes from the phoneme sequence conversion unit (120), and by referring to a feature database for each phoneme or a mapping table for articulatory features for each phoneme, analyze articulatory features such as place of articulation, manner of articulation, and vowel quadrilateral, and assign the corresponding feature information to each phoneme. For example, the feature annotation section (130) can assign features to the phoneme "ㄱ( / k / )" by analyzing that in the case of Korean, the place of articulation of "ㄱ( / k / )" is a velar, the manner of articulation is a stop, and the voiced / unvoiced type is a voiceless sound, and can assign features to the phoneme "ㅁ( / m / )" by analyzing that the place of articulation of "ㅁ( / m / )" is a bilabial sound and the manner of articulation is a nasal sound, and in the case of English, can assign features to the phoneme " / ð / " by analyzing that the place of articulation of " / ð / " is an interdental sound, the manner of articulation is a fricative, and the voiced / unvoiced type is a voiced sound. In addition, in the case of vowels, " / æ / " can be analyzed that the height of the tongue is mid-low, the frontness of the tongue is front, and the roundness is unrounded.

[0039] The phonological change rule matching unit (140) detects candidate phonemes that can be replaced, omitted, distorted, or added for each decomposed phoneme according to language-specific phonological change rules, and is linked with the feature annotation unit (130) and the candidate combination unit (150) described later.

[0040] Here, substitution refers to the conversion of at least one of the initial, medial, or final consonants into a different phoneme; omission refers to the replacement of an initial consonant with "ㅇ" or the omission of a final consonant; distortion refers to pronunciation that is similar to the target sound but different from the standard sound; and addition refers to the insertion of a new phoneme or syllable.

[0041] The above language-specific phonological change rules include syllable omission, syllable addition, initial consonant omission, final consonant omission, stop omission, fricative omission, affricate omission, nasal omission, liquid omission, bilabial omission, alveolar omission, velar omission, glottal omission, alveolar forwarding, palatal forwarding, velar forwarding, glottal forwarding, bilabial backwarding, alveolar backwarding, palatal backwarding, velar backwarding, labialization, alveolarization, palatalization, velarization, glottalization, labiodentalization, interdentalization, stop assimilation, fricativeization, affricateization, liquidization, nasalization, bilabial assimilation, alveolar assimilation, palatal assimilation, velar assimilation, glottal assimilation, stop assimilation, fricative assimilation, affricate assimilation, aspirated assimilation, tense assimilation, tense assimilation, laxation, aspirated and / or deaeration may be included.

[0042] Meanwhile, in the case of Korean, the initial consonants are "ㄱ", "ㄲ", "ㄴ", "ㄷ", "ㄸ", "ㄹ", "ㅁ", "ㅂ", "ㅃ", "ㅅ", "ㅆ", "ㅇ", "ㅈ", "ㅉ", "ㅊ", "ㅋ", "ㅌ", "ㅍ", "ㅎ", and the final consonants are "ㄱ", "ㄴ", "ㄷ", "ㄹ", "ㅁ", "ㅂ", "ㅇ", and

[0043] Plosives are the initial consonants "ㄱ", "ㄲ", "ㄷ", "ㄸ", "ㅂ", "ㅃ", "ㅋ", "ㅌ", "ㅍ" and the final consonants "ㄱ", "ㄷ", "ㅂ", and

[0044] The affricates are the initial consonants "ㅈ", "ㅉ", and "ㅊ", and

[0045] The fricatives are the initial consonants "ㅅ", "ㅆ", and "ㅎ", and

[0046] Nasal sounds are the initial consonants "n" and "m" and the final consonants "n", "m", and "ng", and

[0047] Liquids are the initial consonant "ㄹ" and the final consonant "ㄹ", and

[0048] The bilabial consonants are the initial consonants "ㅂ", "ㅃ", "ㅁ", "ㅍ" and the final consonants "ㅁ", "ㅂ", and

[0049] The alveolar sounds are the initial consonants "n", "d", "tt", "l", "s", "ss", "tt", and the final consonants "n", "d", and "l".

[0050] Palatal consonants are the initial consonants "ㅈ", "ㅉ", and "ㅊ", and

[0051] The velar consonants are the initial consonants "ㄱ", "ㄲ", "ㅋ" and the final consonant "ㅇ", and

[0052] The glottal sound is the initial consonant "ㅎ", and

[0053] The tense consonants are the initial consonants "ㄲ", "ㄸ", "ㅃ", "ㅆ", and "ㅉ", and

[0054] The aspirated sounds are the initial consonants "ㅋ", "ㅌ", "ㅍ", "ㅊ", and "ㅎ", and

[0055] The loose consonants are the initial consonants "ㄱ", "ㄷ", "ㅂ", "ㅅ", and "ㅈ".

[0056] Syllable omission means that the number of syllables in a base word is omitted. For example, when "cat" is changed to "cat," this constitutes syllable omission.

[0057] Syllable addition means that at least one syllable is added to a syllable of a base word. For example, when "cat" is changed to "goyaangi," this constitutes syllable addition.

[0058] Initial consonant omission refers to the replacement of the initial or medial consonant of a base word with "ㅇ". For example, when "고구마" (sweet potato) is transformed into "고우마" (sweet potato), "오구마" (sweet potato), "고구아" (sweet potato), "오우마" (sweet potato), "고우아" (sweet potato), or "오구아" (sweet potato), it constitutes initial consonant omission.

[0059] Final consonant omission means that a final consonant is omitted from at least one of the first, middle, or last syllables of a reference word. For example, when "spring vacation" is transformed into "bobanghak," "bobahak," "bobaha," "bombaha," "bobangha," and "bombangha," it is considered final consonant omission.

[0060] Plosive omission refers to the replacement of a plosive initial consonant included in the initial or medial consonant position of a reference word with "ㅇ," or the omission of a plosive final consonant in the same position. For example, when "바람" changes to "아람" or "김밥" changes to "김바," this constitutes plosive omission.

[0061] Fricative omission refers to the replacement of a fricative consonant contained in the initial or medial consonant of a reference word with "ㅇ". For example, when "사슴" (deer) is transformed into "아슴", "사음", or "아음", it constitutes fricative omission.

[0062] Affricate omission refers to the replacement of an affricate consonant found in the initial or medial consonant position of a reference word with "ㅇ". For example, when "기준" changes to "기운", it constitutes affricate omission.

[0063] Nasal omission refers to the replacement of a nasal initial consonant contained in the initial or medial consonant position of a reference word with "ㅇ," or the omission of a nasal final consonant in the same position. For example, when "manna" is transformed into "anna," "mana," or "maa," it constitutes nasal omission.

[0064] Liquid omission refers to the replacement of a liquid initial consonant contained in the initial or medial consonant position of a reference word with "ㅇ," or the omission of a liquid final consonant in the same position. For example, when "ramyeon" changes to "amyeon" or "jorong" changes to "joong," this constitutes liquid omission.

[0065] Labial omission refers to the replacement of a labial initial consonant in the initial or medial position of a reference word with "ㅇ," or the omission of a labial final consonant in the same position. For example, cases where "빵꾸" changes to "앙꾸," or "엄마" changes to "어마" or "어아," constitute labial omission.

[0066] Alveolar omission refers to the replacement of an alveolar initial consonant in the initial or medial position of a reference word with "ㅇ," or the omission of an alveolar final consonant in the same position. For example, when "다람쥐" (squirrel) changes to "아암쥐" (aam-jwi) or "다암쥐" (daam-jwi), or when "안아" (an-a) changes to "아아" (a-a), this constitutes alveolar omission.

[0067] Velar omission refers to the replacement of a velar initial consonant in the initial or medial position of a reference word with "ㅇ," or the omission of a velar final consonant in the same position. For example, cases where "강가" changes to "앙가" or "강아," or where "악기" changes to "아이" or "아기," constitute velar omission.

[0068] Glottal omission refers to the replacement of a glottal consonant included in the initial or medial consonant of a reference word with "ㅇ". For example, when "하루" changes to "아루", it constitutes glottal omission.

[0069] Alveolar forwarding refers to the conversion of an alveolar initial consonant in the initial or medial position of a reference word into a bilabial initial consonant, or an alveolar final consonant in the final position into a bilabial final consonant. For example, cases where "나라" changes to "바라" or "나마," or "다림질" changes to "다림짐" or "다림질," constitute alveolar forwarding.

[0070] Palatal forwarding refers to the transformation of a palatal initial consonant contained in the initial or medial position of a reference word into a bilabial or alveolar initial consonant. For example, when "자정거" (bicycle) transforms into "다던거," "다정거," "마던거," or "마정거," this constitutes palatal forwarding.

[0071] Velar forwarding refers to the transformation of a velar initial consonant in the initial or medial position of a reference word into a labial or alveolar initial consonant, or a velar final consonant in the final position into a labial or alveolar final consonant. For example, when "gabang" changes to "babang" or "gabam," or when "agi" changes to "adi" or "ami," this constitutes velar forwarding.

[0072] Glottal forwarding refers to the transformation of the glottal initial consonant of a reference word into an initial consonant other than "ㅇ". For example, the transformation of "Hama" into "Gama" constitutes glottal forwarding.

[0073] Backwarding of labial consonants refers to the transformation of a labial initial consonant of a reference word into an initial consonant other than "ㅇ," or the transformation of a labial final consonant into a final consonant. For example, when "banana" transforms into "ga-na-na," "ja-na-na," or "dan-na-na," this constitutes backwarding of labial consonants.

[0074] Alveolar backfiring refers to the transformation of an alveolar initial consonant of a reference word into "ㄱ," "ㄲ," "ㅈ," "ㅉ," "ㅊ," "ㅋ," or "ㅎ," or an alveolar final consonant into "ㄱ" or "ㅇ." For example, when "naganda" changes to "ganda," "nagangda," or "nagangga," this constitutes alveolar backfiring.

[0075] Palatal rearrangement refers to the transformation of the palatal initial consonant of a reference word into "ㄱ," "ㄲ," "ㅋ," or "ㅎ." For example, when "자전거" (bicycle) transforms into "가전거," "하전거," or "가건거," this constitutes palatal rearrangement.

[0076] Posterior velarization refers to the change of the initial velar consonant of a reference word to "ㅎ". For example, when "거미" changes to "허미", it corresponds to posterior velarization.

[0077] Labialization refers to the transformation of a non-labial initial consonant into a labial initial consonant or a non-labial final consonant into a labial final consonant in a reference word. For example, when "gaji" changes to "maji" or "mabi," this constitutes labialization.

[0078] Alveolarization refers to the transformation of a non-alveolar initial consonant into an alveolar initial consonant or a non-alveolar final consonant into an alveolar final consonant in a base word. For example, when "강물" (river) changes to "낭물" (nangmul) or "간물" (ganmul), it constitutes alveolarization.

[0079] Palatalization refers to the transformation of a non-palatal initial consonant in a base word into a palatal initial consonant. For example, when "seoltang" changes to "jeoltang" or "seolchang," this constitutes palatalization.

[0080] Velarization refers to the transformation of a non-velar initial consonant into a velar initial consonant or a non-velar final consonant into a velar final consonant in a base word. For example, when "supul" changes to "gupul" or "supung," this constitutes velarization.

[0081] Glottalization refers to the transformation of a non-glottal initial consonant in a reference word into a glottal initial consonant. For example, the transformation of "geomi" into "heomi" is an example of glottalization.

[0082] Labiodentalization means that the bilabial consonant of a reference word is pronounced as the International Phonetic Alphabet (IPA) "f".

[0083] Interdental assimilation means that the alveolar sound of a reference word is pronounced as the International Phonetic Alphabet "θ".

[0084] Plosive assimilation refers to the transformation of an initial consonant in a base word—which is neither a plosive nor "ㅇ"—into a plosive initial consonant, or a final consonant that is not a plosive into a plosive final consonant. For example, the transformation of "gamyeon" into "gakyeon" constitutes plosive assimilation.

[0085] Fricative assimilation refers to the transformation of a non-fricative initial consonant in a base word into a fricative initial consonant. For example, the change from "gabang" to "sabang" is an example of fricative assimilation.

[0086] Affrication refers to the transformation of a non-affricate initial consonant in a base word into an affricate initial consonant. For example, when "gabang" changes to "jjabang," this constitutes affrication.

[0087] Liquidization refers to the transformation of a non-liquid initial consonant into a liquid initial consonant or a non-liquid final consonant into a liquid final consonant in a base word. For example, the transformation of "gamyeon" into "ramyeon" or "jung" into "jul" constitutes liquidization.

[0088] Nasalization refers to the transformation of a non-nasal initial consonant in a base word into a nasal initial consonant, or a non-nasal final consonant into a nasal final consonant. For example, when "gama" changes to "nama" or "hak" changes to "han," this constitutes nasalization.

[0089] Labial assimilation refers to the transformation of a non-labial initial consonant into a labial initial consonant or a non-labial final consonant into a labial final consonant within a reference word containing a labial consonant. For example, the transformation of "mak" into "mam" or "baji" into "babi" constitutes labial assimilation.

[0090] Alveolar assimilation refers to the transformation of a non-alveolar initial consonant into an alveolar initial consonant or a non-alveolar final consonant into an alveolar final consonant within a reference word containing an alveolar sound. For example, the transformation of "ga-na" into "dana" or "seo-raep" into "seo-ran" constitutes alveolar assimilation.

[0091] Palatal assimilation refers to the transformation of a non-palatal initial consonant into a palatal initial consonant in a reference word containing a palatal sound. For example, when "Jajangmyeon" changes to "Jajangjeon," this constitutes palatal assimilation.

[0092] Velar assimilation refers to the transformation of a non-velar initial consonant into a velar initial consonant or a non-velar final consonant into a velar final consonant within a base word containing a velar. For example, the transformation of "gawi" into "gagwi" or "kkamak" into "kkamang" constitutes velar assimilation.

[0093] Glottal assimilation refers to the transformation of a non-glottal initial consonant into a glottal initial consonant within a reference word that contains a glottal sound. For example, the transformation of "Hama" into "Haha" is an example of glottal assimilation.

[0094] Plosive assimilation refers to the transformation of a non-plosive initial consonant into a plosive initial consonant or a non-plosive final consonant into a plosive final consonant within a base word containing a plosive. For example, cases where "gasu" changes to "gadu" or "kkeomjeong" changes to "kkeokjeong" or "kkeokjeong" constitute plosive assimilation.

[0095] Fricative assimilation refers to the transformation of a non-fricative initial consonant into a fricative initial consonant in a reference word containing a fricative. For example, the transformation of "seoro" into "seoho" is an example of fricative assimilation.

[0096] Affricate assimilation refers to the transformation of a non-affricate initial consonant into an affricate initial consonant within a reference word containing an affricate. For example, the transformation of "둘째" into "줄째" is an example of affricate assimilation.

[0097] Aspiration assimilation refers to the transformation of a non-aspirated initial consonant into an aspirated initial consonant within a reference word that contains an aspirated sound. For example, the transformation of "기타" into "키타" is an example of aspiration assimilation.

[0098] Tense assimilation refers to the transformation of a non-tense initial consonant into a tense initial consonant within a reference word containing a tense consonant. For example, the change from "gambbang" to "kkambbang" is an example of tense assimilation.

[0099] Tension refers to the transformation of a non-tension initial consonant in a reference word into a tension initial consonant. For example, the transformation of "sori" into "ssori" is an example of tension.

[0100] Relaxation refers to the conversion of a tense or aspirated initial consonant in a reference word into a lax initial consonant. For example, when "kkangchon" changes to "gangjon," this constitutes laxation.

[0101] Aspiration refers to the conversion of a non-aspirated initial consonant in a reference word into an aspirated initial consonant. For example, the conversion of "jeori" to "cheori" is an example of aspiration.

[0102] Deaspiration refers to the conversion of an aspirated initial consonant in a reference word into a non-aspirated initial consonant. For example, the conversion of "tajo" to "dajo" is an example of deaspiration.

[0103] That is, the phonological change rule matching unit (140) can detect phonemes to which the above-described phonological change rule is applicable among phonemes that are decomposed by the phoneme sequence conversion unit (120) and to which features such as articulation position, articulation method, and vowel quadrant are assigned by the feature annotation unit (130).

[0104] The candidate combination unit (150) generates obstruent candidate words by combining detected candidate phonemes and is linked with the phonological change rule matching unit (140) and the verification unit (160) described later.

[0105] More specifically, the candidate combining unit (150) receives phoneme candidates that can be replaced, omitted, distorted, and / or added according to phonological change rules from the phonological change rule matching unit (140), and reconstructs the input phoneme candidates according to syllable structure rules and combines them into actual utterable syllable units, thereby automatically generating multiple obstruent candidate words.

[0106] Here, syllable structure rules refer to constraints on the arrangement of phonemes that constitute a syllable. For example, in the case of Korean, it follows syllable structure rules consisting of an initial consonant, a vowel, and a final consonant.

[0107] That is, for example, the candidate combination unit (150) can generate obstruent candidate words such as "aseum," "saeum," and "aeum" in which the initial consonant "s" or the medial consonant "s" is replaced with "ng" according to the fricative omission rule, when the reference word is "deer" and the candidate phonemes detected as fricative omissionable in the phonological change rule matching unit (140) are the initial consonant "s" and the medial consonant "s".

[0108] The verification unit (160) verifies the generated obstruent candidate words according to language-specific phonological constraints and phonological legality to determine the final obstruent candidate words, and is linked with the candidate combination unit (150) and the alignment unit (170) described later.

[0109] The verification unit (160) more specifically determines whether the obstruent candidate word generated by the candidate combination unit (150) violates the phonological permission rules of the language and whether it is physically possible to utter it in the articulatory organ. For example, in the case of Korean, the verification unit (160) can filter obstruent candidate words that include prohibited consonant clusters such as "ㄹㄹ" and "ㄸㄸ" or non-permitted syllable structures such as initial consonant, initial consonant, medial vowel, and final consonant (CCVC); in the case of English, it can filter obstruent candidate words that include illegal phoneme combinations starting with / ŋ / and obstruent candidate words that include unnatural sequence phonemes such as "ㄹㄹㅃ" and "ㅋㅎㅋ" and are physically impossible to utter.

[0110] That is, the verification unit (160) can determine the final candidate words for obstruents as "Gungbap" and "Gungmam" by determining that "Gungbap" and "Gungmam" are physically utterable without violating Korean phonological constraints, and by filtering out "Gukmam" from the candidate words for obstruents by determining that it has a low probability of utterance due to the unnaturalness of the articulation position transition.

[0111] The alignment unit (170) aligns the reference word and the final obstruent word candidate to analyze the correspondence relationship, and is linked with the verification unit (160) and the labeling unit (180) described later.

[0112] The alignment unit (170) can, more specifically, divide the final obstruent candidate word and the reference word determined by the verification unit (160) into phoneme units, and then compare the phonemes at the same position to determine the phonemes of the reference word that have been substituted, omitted, distorted, or added.

[0113] For example, the alignment unit (170) can determine that the initial consonant "ㄱ" has been labialized into "ㅁ" when the reference word is "가지" and the final obstruent candidate word is "마지".

[0114] The labeling unit (180) automatically labels each final obstruent word with a detailed phonological change code and is linked with the alignment unit (170) and the output unit (190) described later.

[0115] Detailed phonological change codes may include language type codes, error type codes, phonological change rule codes, rule application location codes, and surrounding phonological condition codes.

[0116] The language type code indicates the language type of the reference word and the final obstruent candidate word; for example, Korean can be displayed as [KR], English as [EN], etc.

[0117] The error type code indicates the type of error that occurred in the final obstruent candidate word and can be classified according to error types, including substitution, omission, distortion, and addition. For example, [SUB] may be indicated for substitution, [OMI] for omission, [DIS] for distortion, and [ADD] for addition.

[0118] The phonological change rule code indicates the type of phonological change rule applied to the transformed phoneme, and can be represented, for example, as [R08] for liquid omission, [R24] for palatalization, [R34] for labial assimilation, etc.

[0119] The rule application location code indicates the position of the converted phoneme, and can be displayed as, for example, [ONSET] for the initial consonant, [VOWEL] for the medial vowel, and [CODA] for the final consonant.

[0120] The surrounding phonological condition code indicates the phonological environment surrounding the transformed phoneme and can be displayed as, for example, [NASAL] for a nasal adjacent environment, [LIQ] for a liquid adjacent environment, [VELAR] for a velar adjacent environment, [LABIAL] for a labial adjacent environment, and [NULL] for no influence from surrounding phonemes.

[0121] That is, the labeling unit (180) can label the word "Gungbap" with a detailed phonological change code [KR_SUB_R33_CODA_LABIAL], for example, when the final obstruent candidate word of the reference word "Gukbap" is determined to be "Gungbap" by the verification unit (160), since it is Korean and the final consonant "ㄱ" is nasalized in an adjacent environment of the labial consonant "ㅂ", and when the final obstruent candidate word of the reference word "Saseum" is determined to be "Aseum" by the verification unit (160), since it is Korean and the initial consonant "ㅅ" is omitted in an environment without the influence of surrounding phonology, the labeling unit can label the word "Aseum" with a detailed phonological change code [KR_OMI_R06_ONSET_NULL].

[0122] Additionally, the labeling unit (180) can label multiple detailed phonological change codes to the final obstruent candidate word when multiple phonological change rules are applied to the final obstruent candidate word.

[0123] Additionally, the labeling unit (180) may be configured to generate a list of final obstruent candidate words labeled with detailed phonological change codes, and to store the generated list in a database.

[0124] The output unit (190) provides a list of labeled final obstruction sound candidate words and is linked with the user terminal unit.

[0125] That is, the output unit (190) can provide a list of multiple final obstruent candidate words, each labeled with a detailed phonological change code by the labeling unit (180), to the user terminal, and the user can determine whether the speaker has an articulation disorder by comparing the final obstruent candidate words provided by the output unit (190) with the reference word pronounced by the speaker.

[0126] According to an embodiment of the present invention, an automatic obstruent sound generation system (100) for diagnosing articulation disorders, comprising a word input unit (110), a phoneme sequence conversion unit (120), a feature annotation unit (130), a phonological change rule matching unit (140), a candidate combination unit (150), a verification unit (160), an alignment unit (170), a labeling unit (180), and / or an output unit (190), the process of diagnosing articulation disorders can be automated by automatically generating and verifying candidate obstruent words of normal words that can be converted according to phonological change rules for each language. Furthermore, compared to cases relying on subjective listening or manual transcription, the deviation between examiners is significantly reduced, thereby greatly improving the objectivity and consistency of the judgment of articulation disorders. In addition, the types of articulation errors can be classified according to linguistic grounds, and detailed phonological change codes are automatically assigned for each type of error, thus having the effect of being usable for clinical analysis and the generation of learning data using artificial intelligence. Furthermore, by adding language-specific phonological rule libraries, the system can be expanded to diagnose articulation disorders in multiple languages, including not only Korean but also English and Japanese. Additionally, storing the generated candidate obstruent words and labeling results in a database format enables the acquisition of standardized diagnostic data required in the field of speech therapy, such as for clinical evaluation, speech therapy education, and training AI models for articulation disorders.

[0128] Now, with reference to the attached drawings, a method for automatically generating a sound of an articulation disorder (S100) according to an embodiment of the present invention will be described in detail.

[0129] FIG. 6 is a flowchart of a method for automatically generating obstruent sounds for diagnosing articulation disorders according to an embodiment of the present invention.

[0130] As illustrated in FIG. 6, an automatic sound generation system (100) for diagnosing articulation disorders according to an embodiment of the present invention may include a word input step (S110), a phoneme sequence conversion step (S120), a feature annotation step (S130), a phonological change rule matching step (S140), a candidate combination step (S150), a verification step (S160), an alignment step (S170), a labeling step (S180), and / or an output step (S190).

[0131] The word input step (S110) is a step of inputting a reference word and a language code of the reference word, and is performed by the user inputting the reference word and the language code of the reference word into the word input unit (110).

[0132] The phoneme sequence conversion step (S120) is a step of breaking down a reference word into phoneme units, and is carried out by the phoneme sequence conversion unit (120).

[0133] The feature annotation step (S130) is a step of assigning articulation position, articulation method, and vowel quadrilateral features to each decomposed phoneme, and is carried out by the feature annotation unit (130).

[0134] The phonological change rule matching step (S140) is a step of detecting candidate phonemes that can be replaced, omitted, distorted, or added for each decomposed phoneme according to the phonological change rules for each language, and is carried out by the phonological change rule matching unit (140).

[0135] The candidate combination step (S150) is a step of generating obstruent candidate words by combining detected candidate phonemes, and is carried out by the candidate combination unit (150).

[0136] The verification step (S160) is a step of determining the final obstruent candidate word by verifying the generated obstruent candidate word according to language-specific phonological constraints and phonological legality, and is carried out by the verification unit (160).

[0137] The alignment step (S170) is a step of aligning the reference word and the final obscuration candidate word to analyze the correspondence relationship, and is carried out by the alignment unit (170).

[0138] The labeling step (S180) is a step of automatically labeling each final obstruent word with a detailed phonological change code, and is carried out by the labeling unit (180).

[0139] The output step (S190) is a step of providing a list of labeled final obstruent sound candidate words, and is carried out by the output unit (190).

[0140] Meanwhile, the word input unit (110), phoneme sequence conversion unit (120), feature annotation unit (130), phonological change rule matching unit (140), candidate combination unit (150), verification unit (160), alignment unit (170), labeling unit (180), and output unit (190) for implementing the method for automatically generating a sound of an articulation disorder (S100) according to one embodiment of the present invention have been described above, so a redundant description is omitted.

[0141] According to the automatic generation method (S100) for diagnosing articulation disorders according to an embodiment of the present invention, which includes a word input step (S110), a phoneme sequence conversion step (S120), a feature annotation step (S130), a phonological change rule matching step (S140), a candidate combination step (S150), a verification step (S160), an alignment step (S170), a labeling step (S180), and / or an output step (S190), the process of diagnosing articulation disorders can be automated by automatically generating and verifying candidate words for obstruents of normal words that can be converted according to phonological change rules for each language. Furthermore, compared to cases relying on subjective listening or manual transcription, the deviation between examiners is significantly reduced, thereby greatly improving the objectivity and consistency of the judgment of articulation disorders. In addition, the types of articulation errors can be classified according to linguistic grounds, and detailed phonological change codes are automatically assigned for each error type, thus providing an effect that can be utilized for clinical analysis and the generation of learning data using artificial intelligence. Furthermore, by adding language-specific phonological rule libraries, the system can be expanded to diagnose articulation disorders in multiple languages, including not only Korean but also English and Japanese. Additionally, storing the generated candidate obstruent words and labeling results in a database format enables the acquisition of standardized diagnostic data required in the field of speech therapy, such as for clinical evaluation, speech therapy education, and training AI models for articulation disorders.

[0143] Although all components constituting the embodiments of the present invention have been described above as being combined or operating in combination, the present invention is not necessarily limited to such embodiments. That is, within the scope of the purpose of the present invention, all components may be selectively combined in one or more ways to operate.

[0144] Furthermore, terms such as "include," "compose," or "have" described above, unless specifically stated otherwise, mean that the relevant component may be inherent; therefore, they should be interpreted as allowing for the inclusion of additional components rather than excluding them. All terms, including technical or scientific terms, have the same meaning as generally understood by those skilled in the art to which the present invention pertains, unless otherwise defined. Commonly used terms, such as those defined in advance, should be interpreted in accordance with their meaning in the context of the relevant technology and should not be interpreted in an ideal or overly formal sense unless explicitly defined in the present invention.

[0145] Furthermore, the above description is merely an illustrative explanation of the technical concept of the present invention, and those skilled in the art to which the present invention pertains will be able to make various modifications and variations within the scope of the essential characteristics of the present invention.

[0146] Accordingly, the embodiments disclosed in this invention are intended to illustrate, not limit, the technical concept of the invention, and the scope of the technical concept of the invention is not limited by these embodiments. The scope of protection of this invention shall be interpreted by the claims below, and all technical concepts within an equivalent scope shall be interpreted as being included within the scope of rights of this invention. Explanation of the symbols

[0147] 100: Automatic sound generation system for diagnosing articulation disorders according to an embodiment of the present invention 110 : Word input section 120 : Phoneme sequence conversion unit 130 : Quality Commentary 140 : Phonological change rule matching part 150 : Candidate combination 160 : Verification section 170 : Alignment section 180 : Labeling section 190 : Output section S100: Automatic generation method for diagnosing articulation disorders according to an embodiment of the present invention S110: Word input step S120: Phoneme sequence conversion step S130: Feature annotation stage S140: Phonological change rule matching stage S150: Candidate combination step S160: Verification stage S170: Alignment step S180: Labeling step S190: Output stage

Claims

Claim 1 A word input unit receiving a reference word and a language code of the reference word; a phoneme sequence conversion unit linked to the word input unit and decomposing the reference word into phoneme units; a feature annotation unit linked to the phoneme sequence conversion unit and generating phoneme-specific feature annotation information by referring to a language-specific articulatory feature mapping table selected in correspondence with the language code, assigning consonant articulatory feature information including the place and method of articulation in the case of a consonant, and vowel articulatory feature information including the height of the tongue according to the vowel angle, the anterior-posterior position of the tongue, and rounding in the case of a vowel, for each phoneme decomposed from the reference word; a phonological change rule matching unit linked to the feature annotation unit and, according to language-specific phonological change rules, using the consonant articulatory feature information or the vowel articulatory feature information included in the phoneme-specific feature annotation information to detect phonemes among the decomposed phonemes to which the language-specific phonological change rules are applicable, and detecting candidate phonemes that can be substituted, omitted, distorted, or added to the applicable phonemes; the phonological change rules A system for automatically generating obstruents for diagnosing articulation disorders, comprising: a candidate combining unit linked to a matching unit and generating obstruent candidate words by combining detected candidate phonemes; a verification unit linked to the candidate combining unit and determining final obstruent candidate words based on syllable structure rules and phoneme combining rules for the generated obstruent candidate words; an alignment unit linked to the verification unit and analyzing correspondence relationships by aligning the reference word and the final obstruent candidate words; a labeling unit linked to the alignment unit and automatically labeling detailed phonological change codes to each of the final obstruent candidate words; and an output unit linked to the labeling unit and providing a list of the labeled final obstruent candidate words. Claim 2 In claim 1, the language-specific phonological change rules include: syllable omission, syllable addition, initial consonant omission, final consonant omission, plosive omission, fricative omission, affricate omission, nasal omission, liquid omission, labial omission, alveolar omission, velar omission, glottal omission, alveolar forwarding, palatal forwarding, velar forwarding, glottal forwarding, labial backwarding, alveolar backwarding, palatal backwarding, velar backwarding, labialization, alveolarization, palatalization, velarization, glottalization, labiodentalization, interdentalization, plosiveization, fricativeization, affricateization, liquidization, nasalization, labial assimilation, alveolar assimilation, palatal assimilation, velar assimilation, glottal assimilation, plosive assimilation, fricative assimilation, affricate assimilation, aspirated assimilation, tense assimilation, tense, Automatic obstruent generation system for diagnosing articulation disorders, including laxation, aspiration, and deaspiration. Claim 3 An automatic obstruent generation system for diagnosing articulation disorders according to claim 1, wherein the detailed phonological change code comprises a language type code, an error type code, a phonological change rule code, a rule application location code, and a surrounding phonological condition code. Claim 4 An automatic sound generation system for diagnosing articulation disorders according to claim 3, wherein the error type code is classified according to error types including substitution, omission, distortion, and addition. Claim 5 A method for automatically generating obstruents for diagnosing articulation disorders, comprising: a word input step of inputting a reference word and a language code of the reference word by a word input unit; a phoneme sequence conversion step of decomposing the reference word into phoneme units by a phoneme sequence conversion unit; a feature annotation step of assigning articulation location and articulation method features to each decomposed phoneme by a feature annotation unit; a phonological change rule matching step of generating candidate phonemes that can be substituted, omitted, distorted, or added by applying language-specific phonological change rules based on the features by a phonological change rule matching unit; a candidate combination step of generating obstruent candidate words by combining the phonemes by a candidate combination unit; a verification step of determining final obstruent candidate words by filtering data among the obstruent candidate words that conform to syllable structure rules and phoneme combination rules by a verification unit; an alignment step of aligning the reference word and the final obstruent candidate words and analyzing the correspondence relationship by an alignment unit; a labeling step of automatically labeling each of the final obstruent candidate words with a detailed phonological change code by a labeling unit; and an output step of providing a list of the final obstruent candidate words by an output unit.

Images