A method for intelligently pushing Chinese pronunciation training programs

Abstract

The invention discloses a method for intelligently pushing a Chinese articulation training program, which comprises the following steps: step A, establishing an articulation disorder feature code according to speech evaluation data of an evaluation object; step B, querying corresponding database data according to the articulation disorder feature code to obtain a Chinese articulation training program; and step C, pushing the Chinese articulation training program.

Description

Technical Field

[0001] The invention belongs to the technical field of auxiliary training for autism assessment subjects and relates to an intelligent push method for a Chinese pronunciation training program. Background Art

[0002] For various speech and articulation disorders in autistic children, corresponding correction plans are pushed out to help grassroots rehabilitation institutions improve the efficiency of rehabilitation training. It is like developing a set of retrieval tools in the vast articulation speech database, establishing a convenient channel between screening and evaluation and training plans.

[0003] The articulation and speech ability assessment examines children's ability to master the speech articulation of each phoneme. Based on previous research, a group of monosyllabic words for the Chinese articulation ability test were developed for each initial consonant. Then, through several pairs of initial consonant phonetic confusion analysis, the articulation ability of Mandarin initial consonants and the type of disorders were evaluated. This can achieve high validity in measuring the speech errors of patients with speech disorders, and provides a scientific basis for diagnosing the causes of articulation disorders and formulating correction plans. Summary of the invention

[0004] In order to solve the deficiencies in the prior art, the present invention aims to provide a method for intelligently pushing a Chinese pronunciation training program.

[0005] The present invention provides a method for intelligently pushing a Chinese pronunciation training program, the method comprising the following steps:

[0006] Step A, establishing a dysarthria feature code based on the speech assessment data of the assessment subject;

[0007] Step B, querying corresponding database data according to the articulation disorder characteristic code to obtain a Chinese articulation training program;

[0008] Step C: pushing the Chinese pronunciation training program.

[0009] The acquisition of the evaluation data comprises the following steps:

[0010] Step A1, collecting the voice of the evaluation object;

[0011] Step A2: Processing and evaluating the collected speech to obtain a speech evaluation result of the evaluation object;

[0012] Step A3: Based on the speech evaluation result, speech evaluation data is formed in combination with the age of the evaluation object and the target sound.

[0013] In step A1, the speech is tested and collected according to a test word list; the test word list is obtained by summarizing and arranging according to the Chinese phonetic combination rules in the "Basic Rules of Chinese Phonetic Orthography".

[0014] In step A1, the voice collection of the evaluation object is carried out through the following steps:

[0015] 1) Use a microphone and ADC to connect to a computer, and the microphone collects the voice of the person being evaluated;

[0016] 2) The ADC device samples at a rate of 48,000 times per second. Each sampling is cascaded through a multi-stage amplifier and comparator, and the analog signal of the voice is sampled into level signals of different voltages through the analog-to-digital conversion circuit. It is connected and transmitted to the computer via a 3.5mm interface or USB interface, and stored as a sound file for analysis.

[0017] In step A2, the processing of the collected speech includes:

[0018] First, the audio in the sound file is filtered using the following algorithm:

[0019] The state equation of the system: k =φ k，k1 *x k-1 +Γ k-1 w k-1 ,

[0020] The measurement equation of the system: Z k =H k *x k +v k ;

[0021] Where: w k-1 is the process white noise;

[0022] x k is the system state at time k;

[0023] H k are the parameters of the measurement system;

[0024] φ k，k-1 is the state transition moment;

[0025] ν k To measure noise;

[0026] Γ is the noise driving matrix;

[0027] Z k Measure the output of the equation for the system;

[0028] Take the optimal estimate x at time k-1 k-1 As a benchmark, predict the state change at time k At the same time, observe the state and obtain the observed variable Z k, and then analyze between prediction and observation, or correct the prediction with the observation, so as to obtain the optimal state estimate x at time k k ;

[0029] The filtered audio is the human voice audio with environmental noise removed, and the audio is further analyzed;

[0030] The filtered human voice audio is divided into frames, 12,000 samples of 0.25 seconds are taken as one frame, and the entire audio is sliced; and the corresponding combination is found in the Chinese speech database. The algorithm involved is as follows:

[0031] Assume that Q is the set of all possible hidden states and V is the set of all possible observed states, that is:

[0032] Q={q1,q2,…,q N}, V = {v1, v2, ... v M};

[0033] Where N is the number of possible hidden states and M is the number of all possible observed states;

[0034] For a sequence of length T, I is the state sequence corresponding to sequence T, and O is the observation sequence corresponding to sequence T, that is:

[0035] I={i1,i2,...,i T}, O = {o1, o2, ... o T};

[0036] Among them, any hidden state i t ∈Q, any observation state o t ∈V;

[0037] The pronunciation of each Mandarin Chinese character is decomposed into a combination of multiple frames. In the Chinese speech database, there is a sequence set of Q and V containing each character; the input audio slice O is compared with V to find the corresponding character.

[0038] In step A2, the speech evaluation results include: correct interpretation, misinterpretation, loss, confusion; wherein,

[0039] The correct answer refers to the result of speech evaluation: the Chinese pinyin of the single word recognized by the speech of the evaluation object is consistent with the initial consonant and final vowel of the target Chinese pinyin;

[0040] The missing refers to the result of speech evaluation that the Chinese pinyin of the single word obtained by the speech of the evaluation object is consistent with the final vowel of the target sound, and the initial consonant is not pronounced. express;

[0041] The confusion refers to the speech evaluation result that the Chinese Pinyin initial consonant of the single word obtained by the speech of the evaluation object is different from the Chinese Pinyin initial consonant of the target sound, but is related to the Chinese Pinyin initial consonant of the target sound in pronunciation mode or pronunciation position; the speech evaluation result is recorded in the actual speech of the evaluation object;

[0042] The misinterpretation refers to a speech evaluation result in which the speech evaluation result of the evaluation object is not a correct answer, missing, or confused, and the speech evaluation result is represented by an "×".

[0043] If the assessment result is misinterpretation or loss, the subject has deficiencies in both the pronunciation method and the place of articulation of the initial consonant;

[0044] If the evaluation result is confusion, the confusion sound table is used to determine the deficiencies in the pronunciation method and articulation position of the initial consonant;

[0045] The subject pronounced the word three times during the test. In each of the three pronunciations,

[0046] If all three answers are correct, relevant exercises F should be pushed in the push plan to consolidate them;

[0047] If there is a deficiency in pronunciation method or pronunciation position in one of the three tests, it means that the object of evaluation has learned the sound, but it is not stable enough and needs targeted practice and consolidation. The push practice method is: D→E→F;

[0048] If there are deficiencies in pronunciation or articulation position in two of the three tests, it means that the object of evaluation has not yet stably acquired this sound and needs to be strengthened with the following practice: C→D→E;

[0049] If the pronunciation method or pronunciation position is insufficient in all three tests, it means that the subject has problems with the pronunciation position and / or pronunciation method. In this case, basic training will be conducted on the pronunciation position and / or pronunciation method. The training method is as follows: if only the pronunciation position is insufficient, B→C→D will be pushed; if only the pronunciation method is insufficient, A→C→D will be pushed; if both the pronunciation position and pronunciation method are insufficient, A→B→C→D will be pushed;

[0050] The push schemes for different reasons of pronunciation errors and the number of pronunciation errors are shown in Table 1 below.

[0051] Among them, A is the pronunciation part exercise, B is the pronunciation method exercise, C is the pronunciation induction exercise, D is the phoneme training exercise, E is the phoneme comparison exercise, and F is the phonetic reinforcement exercise.

[0052] Table 1 Push schemes for different pronunciation error reasons and pronunciation error times

[0053] Wrong 0 times Wrong 1 time Wrong 2 times Wrong 3 times Place of pronunciation F D→E→F C→D→E B→C→D Pronunciation F D→E→F C→D→E A→C→D Both F D→E→F C→D→E A→B→C→D

[0054] Among them, the pronunciation position, pronunciation method, and both are the reasons for pronunciation errors, and 0 times wrong, 1 time wrong, 2 times wrong, and 3 times wrong are the number of pronunciation errors.

[0055] In step A, the dysarthria feature encoding is Indicates; among them,

[0056] The α represents the age of the evaluation subject in months, α=0-999;

[0057] For different initial consonants, there are corresponding preferred age ranges. For example, for the sound b, the preferred age range for the sound b is 30 months and above. If the evaluation object is less than 30 months old, no relevant solutions will be pushed to it. The corresponding relationship between the specific initial consonants and the preferred age range is shown in Table 2.

[0058] Table 2 Correspondence between different initial consonants and their optimal learning age range

[0059]

[0060] Said It indicates that the target sounds of the test belong to the initial consonants of Mandarin Chinese;

[0061] The Mandarin Chinese initial consonants include: b, m, p, f, d, t, n, l, g, k, h, j, q, x, z, c, s, zh, ch, sh, r;

[0062] The β represents the target sound, that is, the target initial consonant selected for testing among the Mandarin Chinese initial consonants;

[0063] The γ represents the evaluation result of the speech of the evaluation object, which is represented by ε, η, λ, and specific confused sounds; the ε represents that the evaluation result of the speech of the evaluation object is correct; the η represents that the evaluation result of the speech of the evaluation object is distorted; the λ represents that the evaluation result of the speech of the evaluation object is lost; the specific confused sound represents that the evaluation result of the speech of the evaluation object is a specific confused sound related to the pronunciation method or pronunciation position of the target sound;

[0064] The Δ represents the number of occurrences of the speech evaluation result corresponding to the code γ, and the numerical range of the number is 1-3.

[0065] The method to construct the coding of dysarthria features from speech assessment results is:

[0066] Let α represent the age in months of the subject of this test, and fill in 0 on the left to make it a 3-digit number;

[0067] use Indicates that the target sounds of this test are Mandarin Chinese initial consonants;

[0068] Use β to represent the specific initial consonant of the target sound in this test;

[0069] Let γ represent the evaluation result of the speech of the evaluation object in this test;

[0070] Δ represents the number of occurrences of the speech evaluation results of the evaluation object in this test;

[0071] The characters α, β, γ, and Δ are concatenated to obtain the final dysarthria feature code:

[0072] In the data in the database,

[0073] Database initial consonant table:

[0074] Each initial consonant has unique attributes, and the initial consonants and their corresponding attributes are summarized as a database initial consonant table in the database; the unique attributes of each initial consonant are: target sound, target sound acquisition age, target sound detailed information, target sound type, and single-, double-, and three-syllable test words corresponding to the target sound; each attribute of the initial consonant corresponds to a row of entries in the database;

[0075] Database γ-Δ encoding-strategy chain correspondence table:

[0076] For each initial consonant, because of the different evaluation results, the γ-Δ composition code is also different, and each evaluation result code of each initial consonant is summarized as a database γ-Δ code-strategy chain correspondence table in the database; the database γ-Δ code-strategy chain correspondence table has the following attributes: γ-Δ code, target sound, evaluation result category, confused sound, number of errors, strategy chain; each attribute in the database γ-Δ code-strategy chain correspondence table corresponds to a row of entries in the database;

[0077] Database training suggestion chain corresponding table:

[0078] For each initial consonant, the specific code in the strategy chain has a corresponding training suggestion chain, and a database training suggestion chain correspondence table is established; the database training suggestion chain correspondence table includes the following attributes: specific code in the strategy chain, target sound, and training suggestion chain; each attribute in the database training suggestion chain correspondence table corresponds to a row of entries in the database.

[0079] Database training method table:

[0080] For each training suggestion chain, the training method id in the training suggestion chain has corresponding specific content, and a database training method table is established, which includes the following attributes: target sound, method id in the training suggestion chain, and specific content corresponding to the method id; each attribute in the database training method table corresponds to a column of entries in the database. When querying to obtain a specific training strategy, first locate the database initial consonant table through the target sound, and read the age in months that the initial consonant should be learned; if the age of the evaluation object is younger than the age in months that the initial consonant should be learned, the training plan will not be pushed;

[0081] Secondly, according to the γ-Δ coding of the test, according to the target sound and the γ-Δ coding, the strategy chain corresponding to the target sound of the test is obtained in the database γ-Δ coding-strategy chain correspondence table;

[0082] According to the specific code in the strategy chain corresponding to the target sound and the target sound, the database training suggestion chain corresponding table is queried to obtain the training suggestion chain of the specific code of the strategy chain.

[0083] Finally, according to the target sound and the method ID in the training suggestion chain, query the database training method table, obtain the specific content corresponding to each method ID and combine them to obtain the complete training plan for this test.

[0084] The method of the present invention adopts a Chinese pronunciation training program intelligent push system, which includes a collection module, an evaluation module, an input module, an analysis module, a database module, and a push module;

[0085] The acquisition module is used to call the target sound test words in the database module and collect the speech of the evaluation object once or multiple times;

[0086] The evaluation module is used to process and evaluate the collected speech to obtain a speech evaluation result; the evaluation result is combined with the patient's age and target sound to form speech evaluation data;

[0087] The input module is used to input the speech evaluation data of the evaluation object into the analysis module;

[0088] The analysis module is used to establish an articulation disorder feature code according to the speech evaluation data, query the corresponding database data in the database module according to the articulation disorder feature code, and obtain a training program corresponding to the articulation disorder feature code;

[0089] The database module is used to store data tables; the data tables include: database initial consonant information table, database γ-Δ coding-strategy chain correspondence table, database training suggestion chain correspondence table, and database training method table;

[0090] The push module is used to push the training plan obtained by the analysis module.

[0091] The method of the present invention is not intended for the treatment or diagnosis of a disease.

[0092] The beneficial effects of the present invention include: the present invention provides a method for intelligently pushing a Chinese articulation training program, which can be used to implement a method for planning a Chinese articulation disorder correction program for an autism assessment subject, comprising the following steps: step A, establishing an articulation disorder feature code according to the speech assessment data of the assessment subject; step B, querying the corresponding database data according to the articulation disorder feature code to obtain a Chinese articulation disorder correction training program; step C, outputting the articulation disorder correction training program. The present invention can obtain higher validity in measuring speech errors of patients with speech disorders, and provides a scientific basis for diagnosing the cause of articulation disorders and formulating correction programs. BRIEF DESCRIPTION OF THE DRAWINGS

[0093] Figure 1 This is a structural diagram of the articulation disorder planning system of the present invention.

[0094] Figure 2 It is a basic rule diagram of the Chinese Pinyin orthography in the present invention.

[0095] Figure 3 The flowchart of the dysarthria planning method of the present invention is shown in FIG.

[0096] Figure 4 It is a flow chart of the single word test of "initial consonant b" in the present invention.

[0097] Figure 5 This is a confusing sound diagram that may be produced by the "initial consonant b" of the present invention.

[0098] Figure 6 This is a diagram of the pronunciation training strategy for the "initial consonant b" of the present invention.

[0099] Figure 7 This is a diagram of the pronunciation rule library of "initial consonant b" of the present invention.

[0100] Figure 8 It is a flow chart of the single word test of "initial consonant m" in the present invention.

[0101] Fig. 9 This is a confusing sound diagram that may be produced by the "initial consonant m" of the present invention.

[0102] Fig.10 This is a strategy diagram for the pronunciation training of the "initial consonant m" of the present invention.

[0103] Fig.11 This is the pronunciation rule library diagram of the "initial consonant m" of the present invention. DETAILED DESCRIPTION

[0104] The present invention is further described in detail with reference to the following specific examples and drawings. The process, conditions, experimental methods, etc. for implementing the present invention, except for the contents specifically mentioned below, are all common knowledge and common common sense in the art and are not particularly limited by the present invention.

[0105] The initial consonant b is a bilabial unaspirated voiceless stop, which is caused by the upper lip and lower lip touching or approaching, thus blocking the airflow.

[0106] The initial consonant m is a bilabial voiced nasal sound. When pronouncing it, the lips are closed, the soft palate is lowered, the nasal passage is opened, the vocal cords vibrate, and the airflow is sent out from the nasal cavity.

[0107] The present invention provides a Chinese pronunciation training program intelligent push method, using a Chinese pronunciation training program intelligent push system, such as Figure 1 As shown, the system includes a collection module, an evaluation module, an input module, an analysis module, a database module, and a push module;

[0108] The acquisition module is used to call the target sound test words in the database module and collect the speech of the evaluation object once or multiple times;

[0109] The evaluation module is used to process and evaluate the collected speech to obtain a speech evaluation result; the evaluation result is combined with the patient's age and target sound to form speech evaluation data;

[0110] The input module is used to input the speech evaluation data of the evaluation object into the analysis module;

[0111] The analysis module is used to establish an articulation disorder feature code according to the speech evaluation data, query the corresponding database data in the database module according to the articulation disorder feature code, and obtain a training program corresponding to the articulation disorder feature code;

[0112] The database module is used to store data tables; the data tables include: database initial consonant information table, database γ-Δ coding-strategy chain correspondence table, database training suggestion chain correspondence table, and database training method table;

[0113] The four data tables stored in the database module are shown in Table 3-6:

[0114] Table 3 Database initial consonant information table

[0115]

[0116]

[0117] Table 4 Database γ-Δ encoding-strategy chain correspondence table

[0118] Data serial number Data name Data Types Data Description Data Example 1 name VARCHAR γ-Δ encoding ε-3 2 Phonetic VARCHAR Target sound b 3 genre VARCHAR Evaluation results (correct / misinterpreted) correct answer 4 replacement VARCHAR This method is suitable for confusing what sound b 5 occurrence INTEGER The number of target tone errors 0 6 chain VARCHAR Policy Chain F

[0119] Table 5 Database training suggestion chain correspondence table

[0120] Data serial number Data name Data Types Data Description Data Example 1 name INTEGER Specific coding in the strategy chain F 2 Phonetic VARCHAR Target sound b 3 genre VARCHAR Training Advice Chain 11→12→13

[0121] Table 6 Database training method table

[0122]

[0123]

[0124] The push module is used to push the training plan obtained by the analysis module;

[0125] The method for generating and pushing the training plan comprises the following steps:

[0126] Step A, establishing a dysarthria feature code based on the speech assessment data of the assessment subject;

[0127] Step B, querying corresponding database data according to the articulation disorder characteristic code to obtain a Chinese articulation training program;

[0128] Step C: pushing the Chinese pronunciation training program.

[0129] The acquisition of the evaluation data comprises the following steps:

[0130] Step A1, collecting the voice of the evaluation object;

[0131] In step A1, the speech is tested and collected according to a test word list; the test word list is obtained by summarizing and arranging according to the Chinese phonetic combination rules in the "Basic Rules of Chinese Phonetic Orthography".

[0132] Step A2: Processing and evaluating the collected speech to obtain a speech evaluation result of the evaluation object;

[0133] Step A3: Based on the speech evaluation result, speech evaluation data is formed in combination with the age of the evaluation object and the target sound.

[0134] In step A2, the processing of the collected speech includes:

[0135] First, the audio in the sound file is filtered using the following algorithm:

[0136] The state equation of the system: k =φ k，k-1 *x k-1 +Γ k-1 w k-1 ,

[0137] The measurement equation of the system: Z k =Hk *x k +vk;

[0138] Where: w k-1 is the process white noise;

[0139] x k is the system state at time k;

[0140] H k are the parameters of the measurement system;

[0141] φ k，k-1 is the state transition moment;

[0142] ν k To measure noise;

[0143] Γ is the noise driving matrix;

[0144] Z k Measure the output of the equation for the system;

[0145] With the optimal estimate at time kl, x k-1 As a benchmark, predict the state change at time k At the same time, observe the state and obtain the observed variable Z k , and then use the observed value to correct the predicted value, so as to obtain the optimal state estimate x at time k k ;

[0146] The filtered audio is the human voice audio with environmental noise removed, and the audio is further analyzed;

[0147] The filtered human voice audio is divided into frames, 12,000 samples of 0.25 seconds are taken as one frame, and the entire audio is sliced; and the corresponding combination is found in the Chinese speech database. The algorithm involved is as follows:

[0148] Assume that Q is the set of all possible hidden states and V is the set of all possible observed states, that is:

[0149] Q={q1,q2,…,q N}, V = {v1, v2, ... ν M};

[0150] Where N is the number of possible hidden states and M is the number of all possible observed states;

[0151] For a sequence of length T, I is the state sequence corresponding to sequence T, and O is the observation sequence corresponding to sequence T, that is:

[0152] I={i1,i2,...,i T}, O = {o1, o2, ... oT};

[0153] Among them, for any hidden state i t ∈Q, and for any observed state o t ∈V;

[0154] The pronunciation of each Chinese character in Mandarin is decomposed into a combination of multiple frames. In the Chinese speech database, there is a sequence set containing Q and V for each character; the input audio slice O is compared with V to find the corresponding character.

[0155] In step A2, the speech evaluation results include: correct answer, misinterpretation, omission, and confusion; among them,

[0156] The correct answer means that the speech evaluation result is: the initial consonant and final of the Chinese pinyin of the single character recognized from the speech of the evaluation object are both the same as those of the target sound Chinese pinyin;

[0157] The omission means that the speech evaluation result is: the final of the Chinese pinyin of the single character recognized from the speech of the evaluation object is the same as that of the target sound, and the initial consonant is not pronounced. The evaluation result is represented by ;

[0158] The confusion means that the speech evaluation result is: the initial consonant of the Chinese pinyin of the single character recognized from the speech of the evaluation object is different from the initial consonant of the target sound Chinese pinyin, but is related to the initial consonant of the target sound Chinese pinyin in terms of pronunciation method or pronunciation position; the speech evaluation result is recorded based on the actual speech of the evaluation object;

[0159] The misinterpretation means that the speech evaluation result is: the speech evaluation result of the evaluation object does not belong to any of the cases of correct answer, omission, and confusion. The speech evaluation result is represented by "×".

[0160] Taking the initial consonant "b" in Mandarin as an example, the test material contains 3 monosyllabic words with the initial consonant "b": ba (eight), bi (pen), bu (cloth). Each word has a matching picture, and the evaluation object is required to pronounce each sound three times. The pronunciation time of the whole syllable and the interval between syllables are both about 1 second; to induce spontaneous speech, the system can adopt forms such as questioning, prompting, or imitation, and require the evaluation object to say the word expressed by the picture, evaluate the speech of the initial consonant "b" of the autistic evaluation object, and obtain the speech evaluation result;

[0161] The three words have no sequential relationship and can be randomly selected. For example, when the first word "eight" is tested, it can be decided whether to conduct subsequent tests according to the capabilities of the evaluation subject. If obstacle characteristics appear during the test of the first word "eight", then the first word "eight" shall prevail; if the first word "eight" is correct while obstacle characteristics appear in the second word "pen", then the second word "pen" shall prevail; if the first word "eight" and the second word "pen" are correct while obstacle characteristics appear in the third word "cloth", then the third word "cloth" shall prevail.

[0162] Or,

[0163] Taking the initial consonant "m" in Mandarin as an example, the test materials include 3 monosyllabic words with the initial consonant "m": ma (mother), mi (rice), mu (wood), and each word has a matching picture. The evaluation subject is required to pronounce each sound three times. The pronunciation time of the entire syllable and the interval between syllables are both approximately 1 second; to induce spontaneous speech, the system can use forms such as asking questions, giving prompts, or imitating to require the evaluation subject to say the word expressed by the picture, evaluate the pronunciation of the initial consonant "m" of the autistic evaluation subject, and obtain the voice evaluation result.

[0164] The three words have no sequential relationship and can be randomly selected. For example, when the first word "mother" is tested, it can be decided whether to conduct subsequent tests according to the capabilities of the evaluation subject. If obstacle characteristics appear during the test of the first word "mother", then the first word "mother" shall prevail; if the first word "mother" is correct while obstacle characteristics appear in the second word "rice", then the second word "rice" shall prevail; if the first word "mother" and the second word "rice" are correct while obstacle characteristics appear in the third word "wood", then the third word "wood" shall prevail.

[0165] If the evaluation result is distortion or omission, it means that the evaluation subject has deficiencies in both the pronunciation method and pronunciation position of this initial consonant.

[0166] If the evaluation result is confusion, refer to the confusion sound table to determine the deficiencies in the pronunciation method and pronunciation position of this initial consonant.

[0167] The evaluation subject pronounces three times during the test. Among the three pronunciations of the evaluation subject,

[0168] If all three are correct answers, relevant consolidation exercise F needs to be pushed in the push plan.

[0169] If there is a deficiency in the pronunciation method or pronunciation position in one of the three tests, it means that the evaluation subject has learned this sound but it is not stable enough, and targeted practice and consolidation are required. The pushed practice method is: D → E → F.

[0170] If there are deficiencies in pronunciation or articulation position in two of the three tests, it means that the object of evaluation has not yet stably acquired this sound and needs to be strengthened with the following practice: C→D→E;

[0171] If the pronunciation method or pronunciation position is insufficient in all three tests, it means that the subject has problems with the pronunciation position and / or pronunciation method. In this case, basic training will be conducted on the pronunciation position and / or pronunciation method. The training method is as follows: if only the pronunciation position is insufficient, B→C→D will be pushed; if only the pronunciation method is insufficient, A→C→D will be pushed; if both the pronunciation position and pronunciation method are insufficient, A→B→C→D will be pushed;

[0172] After obtaining the speech of the autism assessment subject, the planning system records the patient's speech and selects to analyze and evaluate the speech, recording the evaluation results as correct answers, misinterpretations "×", omissions, etc. Confusion: Real pronunciation;

[0173] Table 7 below is the phoneme acquisition assessment record for "initial consonant b":

[0174] Table 7 "Initial consonant b" phoneme acquisition assessment record

[0175]

[0176]

[0177] Taking the “eight” sound as an example, the process of obtaining the speech evaluation result is as follows: Figure 4 shown.

[0178] According to the system's assessment results of articulation, the autistic assessment subjects may form two dimensions of speech confusion in terms of the place of articulation or the way of articulation when speaking. Among them, since the b sound is a bilabial sound, the autistic assessment subjects may form a confusion in the place of articulation, forming the tip-middle sound d or the root-of-the-tongue sound g; at the same time, since the b sound is an unaspirated stop consonant, there will be a confusion in the way of articulation, forming an aspirated stop consonant p or a nasal sound m or a fricative f, such as Figure 5 shown.

[0179] Table 8 below is the phoneme acquisition assessment record of “initial consonant m”:

[0180] Table 8 "Initial consonant m" phoneme acquisition assessment record

[0181]

[0182] Taking the pronunciation of “mama” as an example, the process of obtaining the speech evaluation result is as follows: Figure 8 shown.

[0183] According to the evaluation results of the system on the articulation situation, children with autism may have phonetic confusion phenomena in two dimensions of the pronunciation position or pronunciation method when vocalizing; among them, the pronunciation positions of m and b are the same, but the pronunciation methods are different. The m sound is a voiced sound, and the b sound is an unaspirated stop sound, so there will be confusion in the pronunciation method, such as Fig. 9 as shown

[0184] Table 9 Confusing Sound Table of Initial Consonants b and m

[0185] Initials Confusion (pronunciation) Confusion sound (place of articulation) Confusion sound (other) b m,p d,g f m b

[0186] In step A, the articulation disorder characteristic coding is represented by ; among them,

[0187] α represents the age in months of the evaluation object, α = 0 - 999; if the age in months represented by α is less than 3 digits, 0s are filled on the left to make it 3 digits;

[0188] the represents that the target sound of the test belongs to the initial consonants of Mandarin Chinese;

[0189] The initial consonants of Mandarin Chinese include: b, m, p, f, d, t, n, l, g, k, h, j, q, x, z, c, s, zh, ch, sh, r;

[0190] β represents the target sound, that is, the target initial consonant selected for testing among the initial consonants of Mandarin Chinese;

[0191] γ represents the evaluation result of the evaluation object's speech, which is represented by ε, η, λ, and specific confusing sounds; ε represents that the evaluation result of the evaluation object's speech is correct; η represents that the evaluation result of the evaluation object's speech is distorted; λ represents that the evaluation result of the evaluation object's speech is missing; the specific confusing sound represents that the evaluation result of the evaluation object's speech is the specific confusing sound related to the pronunciation method or pronunciation position of the target sound;

[0192] Δ represents the number of occurrences of the speech evaluation result corresponding to the coding γ, and the value range of the number is 1 - 3.

[0193] Example: There is a 2.5-year-old child. When evaluating the initial consonant b for him and asking him to repeat after "eight", he has a substitution situation and reads it as "horse" twice. Then his symptom coding is:

[0194] Example: There is a 3.5-year-old child. When evaluating the initial consonant m for him and asking him to repeat after "mom", he has a substitution situation and reads it as "hair" three times. Then his symptom coding is:

[0195] Step B: According to the articulation disorder characteristic coding, query the corresponding database data to obtain the articulation disorder correction training plan;

[0196] 1) Example of “initial consonant b”

[0197] Taking the Mandarin "initial consonant b" as an example, a complete set of "initial consonant b" obstacle training program database is established based on the speech evaluation results of "initial consonant b" and the established articulation disorder feature coding; it can be used for initial consonant b phoneme correction training in the future to improve the accuracy of initial consonant phonemes and speech confusion accuracy. The basic process of obtaining the training program from the speech evaluation results and the established articulation disorder feature coding is as follows: Figure 3 shown.

[0198] Different dysarthria features are encoded based on Figure 3 Analyze the basic process in and form a training plan table, as shown in Table 10 below:

[0199] Table 10 Correspondence between the coding of the characteristics of the initial consonant b articulation disorder and the coding of the training strategy

[0200]

[0201] The training scheme in Table 10 can be Figure 7 Represented as a rule base graph.

[0202] exist Figure 7 There are 6 training targets in total, including: A pronunciation position exercises (oral movement), B pronunciation method exercises (oral movement), C pronunciation induction exercises, D phoneme training exercises, E phoneme comparison exercises, and F phonetic reinforcement exercises.

[0203] As can be seen from Table 10, the training plan table includes a total of 22 cases, involving training strategies divided into three levels:

[0204] The first level of training strategies: Type I to Type VII, each type consists of several training strategy codes in the training chain;

[0205] The second level of training strategy: training strategy codes A, B, C, D, E, F, each training strategy code consists of several specific training suggestions;

[0206] The third level of training strategy: training methods 1-13, each specific training suggestion chain contains rich and specific training methods.

[0207] The above three-layer training strategy is shown in Table 11 below:

[0208] Table 11 Initial b three-level training strategy table

[0209]

[0210] The six targets A, B, C, D, E, and F and their corresponding specific training strategies are as follows:

[0211] Target A: Articulation exercises (oral movements)

[0212] 1. Resistance method: It is used to improve the strength of the lip muscles. The therapist and the patient sit face to face. The therapist pushes the patient's upper lip upward with a tongue depressor, and asks the patient to resist downward with force, repeating several times. Then, the therapist pushes the patient's lower lip downward with a tongue depressor, and asks the patient to resist upward with force, repeating several times.

[0213] 2. Pinching method: used to increase the strength of the lip muscles. The therapist and the patient sit face to face. The therapist places the thumb and index finger on both sides of the patient's philtrum, gently pinches and stretches forward, and asks the patient to smile, repeating several times. Pay attention to smiling with the lips, not tilting the head back to smile.

[0214] 3. Kiss with sound: Close your lips tightly, then separate them, make a kissing sound, and repeat several times.

[0215] 4. Hold the tongue depressor: Hold the tongue depressor with your lips, hold for 5 seconds, and repeat several times.

[0216] Target B: Articulation exercises (oral movements)

[0217] 5. (If b is confused with m), oral resonance and nasal reduction training can be used to correct hypernasal function. Use the monosyllabic words "fish, pounce" and "clothing, adobe" for oral resonance training; read the following words aloud while extending the vowel part of each word: "aunt, sister, uncle, knowledge, watermelon, chicken, carp, flower, tooth, mouth, running, sweeping, sprinkling, watering, washing clothes".

[0218] 5. (If the b sound is confused with p, g, f, d, etc.), learn the pronunciation of b: close the lips tightly to form a blockage, and a small amount of airflow breaks through the blockage of the lips to make a sound;

[0219] Target C: Pronunciation induction exercises

[0220] 6. b pronunciation induction: the lips are tightly closed, the soft palate is lifted, the nasal passage is blocked, the airflow breaks through the obstruction of the lips, the vocal cords do not vibrate, and the airflow is weak;

[0221] Target D: Phoneme training exercises

[0222] 7. Monosyllabic words:

[0223] a: eight, white, hold, leopard, spotted, tie, back, run, jump.

[0224] i: pen, turtle, watch, braid, ice.

[0225] u: cloth, neck.

[0226] 8. Disyllabic words:

[0227] Disyllabic (front):

[0228] a: Dad, bus, swing, white paper, newspaper, gem, zebra, wrench, shell, vest, bandage.

[0229] i: gecko, nose, javelin, hotel, braid, bat, villa, pin.

[0230] u: walk, wave.

[0231] Disyllabic (later):

[0232] a: broom, pine and cypress, bread, school bag, castle, blackboard, sketchbook, water pump, wings, shoulders, popsicle, teacup.

[0233] i: pencil, wall, farewell, watch, champagne, bullwhip, soldier.

[0234] u: arms, running.

[0235] 9. Three-syllable words:

[0236] Three syllables (front):

[0237] a: ballet, radish, blizzard, bowling, safe, office, polar bear, baby on back, shoe tying.

[0238] i: Magic tricks and ice cream.

[0239] u: glass, unhappy.

[0240] Three syllables (medium):

[0241] a: baby, one hundred years old, small stool, playing baseball, covered with quilt, jumping.

[0242] i: runny nose, pencil case, watch shop, setting off firecrackers, refrigerator.

[0243] u: long neck, big strides.

[0244] Three syllables (later):

[0245] a: Big dam, hail, mopping the floor, seesaw, honor roll, gold trophy, notebook.

[0246] i: Watercolor pen, darts, roadside, scallion pancake.

[0247] u: carrot.

[0248] Mark E: Phoneme contrast exercises

[0249] 10(m), bm: bag / cat, dad / scold, cloth / wood, board / man, braid / mian.

[0250] 10(p), bp: white / card, move / climb, neck / mother-in-law, bag / throw, nose / skin.

[0251] 10(d), bd: report / theft, castle / island, dad / big, eight / pair, nose / flute.

[0252] 10(g), bg: wrap / cake, board / drive, bind / port, run / root, and supplement / bone.

[0253] 10(f), bf: quilt / lung, cloth / father, dad / hair, neck / Buddha, mix / rice.

[0254] Target F: Phonetic intensive practice

[0255] 11. b Voice repetition:

[0256] The baby is carrying a schoolbag, the dad is pulling radishes, the uncle is shaking hands with the dad, the uncle is carrying a watermelon on his back, and the white pen is in Bobo's backpack.

[0257] 12. b Voice switching: Cousin likes zebras, aunt is next to cousin, there is bread on the quilt, and the brush is next to the wooden stick.

[0258] Bai Ping is peeling loquats, firecrackers are exploding, the artillerymen are running, and the coach has a whip, a stopwatch and a briefcase.

[0259] My uncle has no white hair, a white-haired person’s baggage, the chalk is next to the darts, and the canvas shoes are placed under the darts.

[0260] There were spots on the sheets, a bench was placed on the floor, a knife handle was on the floor, and there was a big scar on the face.

[0261] My cousin is eating a popsicle, the white dove is next to the starling, the class leader has a pen, and there are sticks by the wall.

[0262] 13. b Voice rotation: Dad is buying puffs, Mom is making instant noodles, the Dalmatian is rolling around, and Aunt is wearing a white coat.

[0263] 2) “Take the initial consonant m as an example”

[0264] Taking the Mandarin word “m” as an example, a complete database of training programs for “m” disorders is established based on the evaluation results of “m” and the established articulation disorder characteristic coding. This database can be used to conduct phoneme correction training for m to improve the accuracy of initial phonemes and speech confusion. The basic process of obtaining training programs based on the evaluation results and the established articulation disorder characteristic coding is as follows: Figure 3 shown.

[0265] Different dysarthria features are encoded based on Figure 3 Analyze the basic process in and form a training plan table, as shown in Table 12 below:

[0266] Table 12 Correspondence between the coding of the initial m articulation disorder characteristics and the coding of the training strategy

[0267]

[0268] The training scheme in Table 12 can be Fig.11 Represented as a rule base graph.

[0269] exist Fig.11 There are 6 training targets in total, including: A pronunciation position exercises (oral movement), B pronunciation method exercises (oral movement), C pronunciation induction exercises, D phoneme training exercises, E phoneme comparison exercises, and F phonetic reinforcement exercises.

[0270] As can be seen from Table 12, the training plan table includes a total of 10 cases, involving training strategies divided into three levels:

[0271] The first level of training strategies: Type I to Type IV, each type consists of several training strategy codes in the training chain;

[0272] The second level of training strategy: training strategy codes A, B, C, D, E, F, each training strategy code consists of several specific training suggestions;

[0273] The third level of training strategy: training methods 1-13, each specific training suggestion chain contains rich and specific training methods.

[0274] The above three-layer training strategy is shown in Table 13 below:

[0275] Table 13 Initial consonant m three-level training strategy table

[0276]

[0277]

[0278] The six targets A, B, C, D, E, and F and their corresponding specific training strategies are as follows:

[0279] Target A: Articulation exercises (oral movements)

[0280] 1. Resistance method: It is used to improve the strength of the lip muscles. The therapist and the patient sit face to face. The therapist pushes the patient's upper lip upward with a tongue depressor, and asks the patient to resist downward with force, repeating several times. Then, the therapist pushes the patient's lower lip downward with a tongue depressor, and asks the patient to resist upward with force, repeating several times.

[0281] 2. Pinching method: used to increase the strength of the lip muscles. The therapist and the patient sit face to face. The therapist places the thumb and index finger on both sides of the patient's philtrum, gently pinches and stretches forward, and asks the patient to smile, repeating several times. Pay attention to smiling with the lips, not tilting the head back to smile.

[0282] 3. Kiss with sound: Close your lips tightly, then separate them, make a kissing sound, and repeat several times.

[0283] 4. Hold the tongue depressor: Hold the tongue depressor with your lips, hold for 5 seconds, and repeat several times.

[0284] Target B: Articulation exercises (oral movements)

[0285] 5. Oral resonance and nasal reduction training can be used to correct hypernasal function. Use the monosyllabic words "fish, puff" and "clothing, adobe" for oral resonance training; read the following words aloud while extending the vowel part of each word: "aunt, sister, uncle, knowledge, watermelon, chicken, carp, flower, tooth, mouth, running, sweeping, sprinkling, watering, washing clothes".

[0286] Target C: Pronunciation induction exercises

[0287] 6. Induction of m pronunciation: close the lips tightly, lower the soft palate, open the nasal passage, vibrate the vocal cords, and send air out of the nasal cavity;

[0288] Target D: Phoneme training exercises

[0289] 7. Monosyllabic words:

[0290] a: mom, horse, buy, cat, coal, full, door, busy, dream;

[0291] i: rice, meow, meow, noodles;

[0292] u: touch, wood, moo;

[0293] 8. Disyllabic words:

[0294] Disyllabic (front):

[0295] a: ant, wheat, towel, hat, rose, jogging, eel, bolt, python.

[0296] i: rice, bee, stopwatch, agility, bread, list.

[0297] u: mushroom, Trojan horse.

[0298] Disyllabic (later):

[0299] a: mother, mountains, eyebrows, strawberry, romance, door, lemon, light.

[0300] i: corn, blow out, temple, farmer, ramen, signature.

[0301] u: touch, building blocks, letters, eyes.

[0302] 9. Three-syllable words:

[0303] Three syllables (front):

[0304] a: Microphone, owl, woolen sweater, mermaid, comic book, doorknob.

[0305] i: Camouflage clothing, bleating, meowing, baker, postcard.

[0306] u: Motorcycle, shepherd.

[0307] Three syllables (middle):

[0308] a: Little ant, barley tea, caterpillar, knitting a sweater, thrush, full, goalkeeper, guide dog, lemon juice.

[0309] i: Hami melon, sketchbook, applying a facial mask, sandwich.

[0310] u: Flower mushroom, thumb, small log cabin.

[0311] Three syllables (back):

[0312] a: Toad, doing business, giant panda, little cat, little sister, walking slowly, guarding the gate,白茫茫 (vast and white), daydream.

[0313] i: Collecting nectar, sixty seconds, instant noodles, very smart.

[0314] u: Great desert, building blocks.

[0315] Target E: Phoneme contrast exercise

[0316] 10(b), m-b: Bag / cat, Dad / scold, Cloth / wood, Board / full, Braid / face.

[0317] Target F: Speech reinforcement exercise

[0318] 11. m sound repetition:

[0319] Mom touches the little cat, the little cat meows, sister doesn't have a towel, Mom touches Mingming's face, Mingming has noodles but no rice.

[0320] 12. m sound switching: Cousin likes zebras, Aunt is next to the cousin, there is bread on the quilt, the writing brush is next to the wooden stick.

[0321] 13. m sound alternation: Dad buys puffs, Mom is making instant noodles, Mom buys lemons, the herdsman drinks mare's milk.

[0322] Example 1

[0323] Taking 22 autism assessment objects in the following table as an example, evaluate their initial consonant b by asking them to repeat after "eight", evaluate their articulation ability and plan corresponding training suggestions:

[0324]

[0325] Evaluation object 1: Pronunciation is correct. It reads as "eight" after three repetitions. According to the evaluation results and other data, the dysarthria feature code obtained is:

[0326] · 111, Program push: Type I (F)

[0327] That is, the pronunciation is correct three times. According to the three-layer training strategy table, the type of training strategy pushed is Type I, and the training strategy code is F. The specific training suggestions are carried out in the following order:

[0328] · Voice repetition (11)

[0329] · Voice switching (12)

[0330] · Voice alternation (13)

[0331] The target sound of this patient's test is b. Locate b in the initial consonant table, and the age of acquisition for the initial consonant read is 30. Since 30 is less than 92, the training suggestions are normally pushed.

[0332] Secondly, according to 1 - 3 of this test, based on the target sound b and the code 1 - 3, the corresponding strategy chain F for this test is obtained in the coding table.

[0333] According to the target sound b of the test and the strategy chain F, query the strategy link list, and the corresponding strategy for this test obtained in the strategy link list is 11 → 12 → 13.

[0334] Finally, according to the target sound b and the strategies 11, 12, 13, query the strategy table to obtain the specific training plan for this test.

[0335] Evaluation object 2: The patient misses the pronunciation all three times, then its symptom code is:

[0336] · 000, Program push: Type II (A → B → C → D)

[0337] That is, the pronunciation is missed all three times. According to the three-layer training strategy table, the type of training strategy pushed is Type II, and the training strategy code is A → B → C → D. The specific training suggestions are carried out in the following order:

[0338] · Learning of pronunciation position: Improving the muscle strength of the lips - resistance method (1), pinching method (2); Promoting lip closure -出声 kiss (3), clamping the tongue depressor (4);

[0339] · Learning of the pronunciation method of b (5): Close the lips tightly to form an obstruction, and a small amount of air breaks through the obstruction of the lips to make a sound;

[0340] b. Pronunciation induction (6): The lips are tightly closed, the soft palate is raised, the nasal passage is blocked, the airflow breaks through the obstruction of the lips, the vocal cords do not vibrate, and the airflow is weak;

[0341] Phoneme acquisition: monosyllabic words (7), disyllabic words (8), trisyllabic words (9)

[0342] Subject 3: There were two missed pronunciations. Based on the evaluation results and other data, the dysarthria feature code was obtained as

[0343] ·001 / 010 / 100, solution push: Type III (C→D→E)

[0344] That is, two of the three pronunciations are missing. According to the three-layer training strategy table, the training strategy type pushed is type III, and the training strategy code is C→D→E. The specific training suggestions are carried out in the following order:

[0345] b. Pronunciation induction (6): The lips are tightly closed, the soft palate is raised, the nasal passage is blocked, the airflow breaks through the obstruction of the lips, the vocal cords do not vibrate, and the airflow is weak;

[0346] Phoneme acquisition: monosyllabic words (7), disyllabic words (8), trisyllabic words (9);

[0347] Voice confusion: bm, bp, bd, bg, bf (10);

[0348] Subject 4: One pronunciation was omitted. Based on the assessment results and other data, the dysarthria characteristic code was 085-

[0349] ·011 / 101 / 110, solution push: Type IV (D→E→F)

[0350] That is, one of the three pronunciations is missed. According to the three-layer training strategy table, the pushed training strategy type is type IV, and the training strategy code is D→E→F. The specific training suggestions are carried out in the following order:

[0351] Phoneme acquisition: monosyllabic words (7), disyllabic words (8), trisyllabic words (9);

[0352] Voice confusion: bm, bp, bd, bg, bf (10);

[0353] Voice training: voice repetition (11), voice switching (12), voice rotation (13);

[0354] Assessment object 5: Three times of pronunciation distortion, based on the assessment results and other data, the dysarthria feature code is obtained as

[0355] 000, solution push: Type II (A→B→C→D)

[0356] That is, the three pronunciations are all misinterpreted. According to the three-layer training strategy table, the training strategy type pushed is II, and the training strategy code is A→B→C→D. The specific training suggestions are carried out in the following order:

[0357] · Articulation learning: improving the strength of the lips - resistance method (1), pinching method (2); promoting lip closure - kissing (3), clamping the tongue depressor (4);

[0358] b. Pronunciation learning (5): The lips are tightly closed to form a blockage, and a small amount of air flows through the blockage of the lips to produce sound;

[0359] b. Pronunciation induction (6): The lips are tightly closed, the soft palate is raised, the nasal passage is blocked, the airflow breaks through the obstruction of the lips, the vocal cords do not vibrate, and the airflow is weak;

[0360] Phoneme acquisition: monosyllabic words (7), disyllabic words (8), trisyllabic words (9)

[0361] Assessment object 6: Second-pass pronunciation distortion. Based on the assessment results and other data, the dysarthria feature code is obtained as

[0362] ·001 / 010 / 100, solution push: Type III (C→D→E)

[0363] That is, two of the three pronunciations are misinterpreted. According to the three-layer training strategy table, the training strategy type pushed is type III, and the training strategy code is C→D→E. The specific training suggestions are carried out in the following order:

[0364] b. Pronunciation induction (6): The lips are tightly closed, the soft palate is raised, the nasal passage is blocked, the airflow breaks through the obstruction of the lips, the vocal cords do not vibrate, and the airflow is weak;

[0365] Phoneme acquisition: monosyllabic words (7), disyllabic words (8), trisyllabic words (9);

[0366] Voice confusion: bm, bp, bd, bg, bf (10);

[0367] Assessment object 7: One pronunciation distortion, based on the assessment results and other data, the dysarthria feature code is obtained as

[0368] ·011 / 101 / 110, solution push: Type IV (D→E→F)

[0369] That is, one of the three pronunciations is misinterpreted. According to the three-layer training strategy table, the training strategy type pushed is type IV, and the training strategy code is D→E→F. The specific training suggestions are carried out in the following order:

[0370] Phoneme acquisition: monosyllabic words (7), disyllabic words (8), trisyllabic words (9);

[0371] Voice confusion: bm, bp, bd, bg, bf (10);

[0372] Voice training: voice repetition (11), voice switching (12), voice rotation (13);

[0373] Assessment object 8: The three pronunciations were confused as m. Based on the assessment results and other data, the dysarthria feature code was obtained as

[0374] 000, solution push: type V (B→C→D)

[0375] That is, the three pronunciations are all confused as the m sound. According to the three-layer training strategy table, the pushed training strategy type is type V, and the training strategy code is B→C→D. The specific training suggestions are carried out in the following order:

[0376] b. Learning pronunciation (5m): Oral resonance and nasal reduction training can be used to correct hypernasal function. Use the monosyllabic words "fish, puff" and "clothing, adobe" for oral resonance training; read the following words aloud while extending the vowel part of each word: "aunt, sister, uncle, knowledge, watermelon, chicken, carp, flower, tooth, mouth, running, sweeping, sprinkling water, watering flowers, washing clothes".

[0377] b. Pronunciation induction (6): The lips are tightly closed, the soft palate is raised, the nasal passage is blocked, the airflow breaks through the obstruction of the lips, the vocal cords do not vibrate, and the airflow is weak;

[0378] Phoneme acquisition: monosyllabic words (7), disyllabic words (8), trisyllabic words (9)

[0379] Assessment object 9: The second pronunciation was confused as m. Based on the assessment results and other data, the dysarthria feature code was obtained as

[0380] ·001 / 010 / 100, solution push: Type III (C→D→E)

[0381] That is, two of the three pronunciations are confused as m sounds. According to the three-layer training strategy table, the training strategy type pushed is type III, and the training strategy code is C→D→E. The specific training suggestions are carried out in the following order:

[0382] b. Pronunciation induction (6): The lips are tightly closed, the soft palate is raised, the nasal passage is blocked, the airflow breaks through the obstruction of the lips, the vocal cords do not vibrate, and the airflow is weak;

[0383] Phoneme acquisition: monosyllabic words (7), disyllabic words (8), trisyllabic words (9);

[0384] Voice confusion: bm(10);

[0385] Subject 10: The first pronunciation was confused as m. Based on the evaluation results and other data, the dysarthria feature code was obtained as

[0386] ·011 / 101 / 110, solution push: Type IV (D→E→F)

[0387] That is, one of the three pronunciations is confused as the m sound. According to the three-layer training strategy table, the pushed training strategy type is type IV, and the training strategy code is D→E→F. The specific training suggestions are carried out in the following order:

[0388] Phoneme acquisition: monosyllabic words (7), disyllabic words (8), trisyllabic words (9);

[0389] Voice confusion: bm(10);

[0390] Voice training: voice repetition (11), voice switching (12), voice rotation (13);

[0391] Subject 11: The pronunciation was confused as p three times. Based on the evaluation results and other data, the dysarthria feature code was obtained as

[0392] 000, solution push: type V (B→C→D)

[0393] That is, the three pronunciations are all confused as the sound p. According to the three-layer training strategy table, the training strategy type pushed is type V, and the training strategy code is B→C→D. The specific training suggestions are carried out in the following order:

[0394] b pronunciation learning (5dpfg): close the lips tightly to form a blockage, and a small amount of airflow breaks through the blockage of the lips to produce sound;

[0395] b. Pronunciation induction (6): The lips are tightly closed, the soft palate is raised, the nasal passage is blocked, the airflow breaks through the obstruction of the lips, the vocal cords do not vibrate, and the airflow is weak;

[0396] Phoneme acquisition: monosyllabic words (7), disyllabic words (8), trisyllabic words (9)

[0397] Subject 12: The second pronunciation was confused as the p sound. Based on the evaluation results and other data, the dysarthria feature code was obtained as

[0398] ·001 / 010 / 100, solution push: Type III (C→D→E)

[0399] That is, two of the three pronunciations are confused as p sounds. According to the three-layer training strategy table, the training strategy type pushed is type III, and the training strategy code is C→D→E. The specific training suggestions are carried out in the following order:

[0400] b. Pronunciation induction (6): The lips are tightly closed, the soft palate is raised, the nasal passage is blocked, the airflow breaks through the obstruction of the lips, the vocal cords do not vibrate, and the airflow is weak;

[0401] Phoneme acquisition: monosyllabic words (7), disyllabic words (8), trisyllabic words (9);

[0402] Voice confusion: bp(10);

[0403] Subject 13: The pronunciation was confused as p. Based on the evaluation results and other data, the dysarthria feature code was obtained as

[0404] ·011 / 101 / 110, solution push: Type IV (D→E→F)

[0405] That is, one of the three pronunciations is confused as the p sound. According to the three-layer training strategy table, the pushed training strategy type is type IV, and the training strategy code is D→E→F. The specific training suggestions are carried out in the following order:

[0406] Phoneme acquisition: monosyllabic words (7), disyllabic words (8), trisyllabic words (9);

[0407] Voice confusion: bp(10);

[0408] Voice training: voice repetition (11), voice switching (12), voice rotation (13);

[0409] Subject 14: The pronunciation of the d sound was confused three times. Based on the evaluation results and other data, the dysarthria feature code was obtained as

[0410] 000, solution push: type VI (A→C→D)

[0411] That is, the three pronunciations are all confused as d. According to the three-layer training strategy table, the training strategy type pushed is type VI, and the training strategy code is A→C→D. The specific training suggestions are carried out in the following order:

[0412] · Articulation learning: improving the strength of the lips - resistance method (1), pinching method (2); promoting lip closure - kissing (3), clamping the tongue depressor (4);

[0413] b. Pronunciation induction (6): The lips are tightly closed, the soft palate is raised, the nasal passage is blocked, the airflow breaks through the obstruction of the lips, the vocal cords do not vibrate, and the airflow is weak;

[0414] Phoneme acquisition: monosyllabic words (7), disyllabic words (8), trisyllabic words (9);

[0415] Subject 15: The second pronunciation was confused as d. Based on the evaluation results and other data, the dysarthria feature code was obtained as

[0416] ·001 / 010 / 100, solution push: Type III (C→D→E)

[0417] That is, two of the three pronunciations are confused as d. According to the three-layer training strategy table, the training strategy type pushed is type III, and the training strategy code is C→D→E. The specific training suggestions are carried out in the following order:

[0418] b. Pronunciation induction (6): The lips are tightly closed, the soft palate is raised, the nasal passage is blocked, the airflow breaks through the obstruction of the lips, the vocal cords do not vibrate, and the airflow is weak;

[0419] Phoneme acquisition: monosyllabic words (7), disyllabic words (8), trisyllabic words (9);

[0420] Voice confusion: bd(10);

[0421] Subject 16: The first pronunciation was confused as d. Based on the evaluation results and other data, the dysarthria feature code was obtained as

[0422] ·011 / 101 / 110, solution push: Type IV (D→E→F)

[0423] That is, one of the three pronunciations is confused as the d sound. According to the three-layer training strategy table, the training strategy type pushed is type IV, and the training strategy code is D→E→F. The specific training suggestions are carried out in the following order:

[0424] Phoneme acquisition: monosyllabic words (7), disyllabic words (8), trisyllabic words (9);

[0425] Voice confusion: bd(10);

[0426] Voice training: voice repetition (11), voice switching (12), voice rotation (13)

[0427] Subject 17: The pronunciation of the g sound was confused three times. Based on the evaluation results and other data, the dysarthria feature code was obtained as

[0428] 000, solution push: type VI (A→C→D)

[0429] That is, the three pronunciations are all confused as g. According to the three-layer training strategy table, the training strategy type pushed is type VI, and the training strategy code is A→C→D. The specific training suggestions are carried out in the following order:

[0430] · Articulation learning: improving the strength of the lips - resistance method (1), pinching method (2); promoting lip closure - kissing (3), clamping the tongue depressor (4);

[0431] b. Pronunciation induction (6): The lips are tightly closed, the soft palate is raised, the nasal passage is blocked, the airflow breaks through the obstruction of the lips, the vocal cords do not vibrate, and the airflow is weak;

[0432] Phoneme acquisition: monosyllabic words (7), disyllabic words (8), trisyllabic words (9);

[0433] Subject 18: The second pronunciation was confused as g. Based on the evaluation results and other data, the dysarthria feature code was obtained as

[0434] ·001 / 010 / 100, solution push: Type III (C→D→E)

[0435] That is, two of the three pronunciations are confused as g. According to the three-layer training strategy table, the training strategy type pushed is type III, and the training strategy code is C→D→E. The specific training suggestions are carried out in the following order:

[0436] b. Pronunciation induction (6): The lips are tightly closed, the soft palate is raised, the nasal passage is blocked, the airflow breaks through the obstruction of the lips, the vocal cords do not vibrate, and the airflow is weak;

[0437] Phoneme acquisition: monosyllabic words (7), disyllabic words (8), trisyllabic words (9);

[0438] Voice confusion: bg(10);

[0439] Subject 19: The pronunciation was confused as g. Based on the evaluation results and other data, the dysarthria feature code was obtained as

[0440] ·011 / 101 / 110, solution push: Type IV (D→E→F)

[0441] That is, one of the three pronunciations is confused as the g sound. According to the three-layer training strategy table, the pushed training strategy type is type IV, and the training strategy code is D→E→F. The specific training suggestions are carried out in the following order:

[0442] Phoneme acquisition: monosyllabic words (7), disyllabic words (8), trisyllabic words (9);

[0443] Voice confusion: bg(10);

[0444] Voice training: voice repetition (11), voice switching (12), voice rotation (13)

[0445] Subject 20: The pronunciation of the sound f was confused three times. Based on the evaluation results and other data, the dysarthria feature code was obtained as

[0446] 000, solution push: Type II (A→B→C→D)

[0447] That is, one of the three pronunciations is confused as the sound f. According to the three-layer training strategy table, the training strategy type pushed is type II, and the training strategy code is A→B→C→D. The specific training suggestions are carried out in the following order:

[0448] · Articulation learning: improving the strength of the lips - resistance method (1), pinching method (2); promoting lip closure - kissing (3), clamping the tongue depressor (4);

[0449] b pronunciation learning (5dpfg): close the lips tightly to form a blockage, and a small amount of airflow breaks through the blockage of the lips to produce sound;

[0450] b. Pronunciation induction (6): The lips are tightly closed, the soft palate is raised, the nasal passage is blocked, the airflow breaks through the obstruction of the lips, the vocal cords do not vibrate, and the airflow is weak;

[0451] Phoneme acquisition: monosyllabic words (7), disyllabic words (8), trisyllabic words (9);

[0452] Subject 21: The second pronunciation was confused as f. Based on the evaluation results and other data, the dysarthria feature code was obtained as

[0453] ·001 / 010 / 100, solution push: Type III (C→D→E)

[0454] That is, two of the three pronunciations are confused as f. According to the three-layer training strategy table, the training strategy type pushed is type III, and the training strategy code is C→D→E. The specific training suggestions are carried out in the following order:

[0455] b. Pronunciation induction (6): The lips are tightly closed, the soft palate is raised, the nasal passage is blocked, the airflow breaks through the obstruction of the lips, the vocal cords do not vibrate, and the airflow is weak;

[0456] Phoneme acquisition: monosyllabic words (7), disyllabic words (8), trisyllabic words (9);

[0457] Voice confusion: bf(10);

[0458] Subject 22: The first pronunciation was confused as f. Based on the evaluation results and other data, the dysarthria feature code was obtained as

[0459] ·011 / 101 / 110, solution push: Type IV (D→E→F)

[0460] That is, one of the three pronunciations is confused as the sound f. According to the three-layer training strategy table, the training strategy type pushed is type IV, and the training strategy code is D→E→F. The specific training suggestions are carried out in the following order:

[0461] Phoneme acquisition: monosyllabic words (7), disyllabic words (8), trisyllabic words (9);

[0462] Voice confusion: bf(10);

[0463] Voice training: voice repetition (11), voice switching (12), voice alternation (13).

[0464] Example 2

[0465] Take the 10 autism assessment subjects in the table below as an example, and assess their initial consonant m, ask them to repeat the pronunciation of "mama", assess their articulation ability and plan corresponding training suggestions:

[0466]

[0467] Subject 1: Pronounced correctly, reading “mama” three times. Based on the assessment results and other data, the dysarthria feature code was obtained as follows:

[0468] 111, Solution Push: Type I (F)

[0469] That is, the three pronunciations are correct. According to the three-layer training strategy table, the training strategy type pushed is type I, and the training strategy code is F. The specific training suggestions are carried out in the following order:

[0470] Voice repetition (11)

[0471] Voice switching (12)

[0472] Voice rotation (13)

[0473] The target sound for this test of the assessment subject is m. The initial consonant table locates m, and the required learning age for reading the initial consonant is 30. 30 is less than 92, so the training suggestion is pushed normally.

[0474] Secondly, according to 1-3 of this test, according to the target sound m and the codes 1-3, the strategy chain F corresponding to this test is obtained in the coding table.

[0475] According to the target tone m and the strategy chain F of the test, the strategy chain table is queried, and the strategy corresponding to the test is obtained in the strategy chain table as 11→12→13.

[0476] Finally, according to the target sound m and strategies 11, 12, and 13, the strategy table is queried to obtain the specific training plan for this test.

[0477] Subject 2: Subject missed three pronunciations, so the symptom code is:

[0478] 000, solution push: Type II (A→B→C→D)

[0479] That is, the three pronunciations are all missed. According to the three-layer training strategy table, the training strategy type pushed is type II, and the training strategy code is A→B→C→D. The specific training suggestions are carried out in the following order:

[0480] · Articulation learning: improving the strength of the lips - resistance method (1), pinching method (2); promoting lip closure - kissing (3), clamping the tongue depressor (4);

[0481] ·M pronunciation learning (5): close the lips tightly to form a blockage, and a small amount of airflow breaks through the blockage of the lips to produce sound;

[0482] m pronunciation induction (6): the lips are tightly closed, the soft palate is raised, the nasal passage is blocked, the airflow breaks through the obstruction of the lips, the vocal cords do not vibrate, and the airflow is weak;

[0483] Phoneme acquisition: monosyllabic words (7), disyllabic words (8), trisyllabic words (9)

[0484] Subject 3: There were two missed pronunciations. Based on the evaluation results and other data, the dysarthria feature code was obtained as

[0485] ·001 / 010 / 100, solution push: Type III (C→D→E)

[0486] That is, two of the three pronunciations are missing. According to the three-layer training strategy table, the training strategy type pushed is type III, and the training strategy code is C→D→E. The specific training suggestions are carried out in the following order:

[0487] m pronunciation induction (6): the lips are tightly closed, the soft palate is raised, the nasal passage is blocked, the airflow breaks through the obstruction of the lips, the vocal cords do not vibrate, and the airflow is weak;

[0488] Phoneme acquisition: monosyllabic words (7), disyllabic words (8), trisyllabic words (9);

[0489] Voice confusion: mb(10);

[0490] Assessment object 4: One pronunciation was omitted. Based on the assessment results and other data, the dysarthria feature code was obtained as

[0491] ·011 / 101 / 110, solution push: Type IV (D→E→F)

[0492] That is, one of the three pronunciations is missed. According to the three-layer training strategy table, the pushed training strategy type is type IV, and the training strategy code is D→E→F. The specific training suggestions are carried out in the following order:

[0493] Phoneme acquisition: monosyllabic words (7), disyllabic words (8), trisyllabic words (9);

[0494] Voice confusion: mb(10);

[0495] Voice training: voice repetition (11), voice switching (12), voice rotation (13);

[0496] Assessment object 5: Three times of pronunciation distortion, based on the assessment results and other data, the dysarthria feature code is obtained as

[0497] 000, solution push: Type II (A→B→C→D)

[0498] That is, the three pronunciations are all misinterpreted. According to the three-layer training strategy table, the training strategy type pushed is II, and the training strategy code is A→B→C→D. The specific training suggestions are carried out in the following order:

[0499] · Articulation learning: improving the strength of the lips - resistance method (1), pinching method (2); promoting lip closure - kissing (3), clamping the tongue depressor (4);

[0500] ·M pronunciation learning (5): close the lips tightly to form a blockage, and a small amount of airflow breaks through the blockage of the lips to produce sound;

[0501] m pronunciation induction (6): the lips are tightly closed, the soft palate is raised, the nasal passage is blocked, the airflow breaks through the obstruction of the lips, the vocal cords do not vibrate, and the airflow is weak;

[0502] Phoneme acquisition: monosyllabic words (7), disyllabic words (8), trisyllabic words (9)

[0503] Assessment object 6: Second-pass pronunciation distortion. Based on the assessment results and other data, the dysarthria feature code is obtained as

[0504] ·001 / 010 / 100, solution push: Type III (C→D→E)

[0505] That is, two of the three pronunciations are misinterpreted. According to the three-layer training strategy table, the training strategy type pushed is type III, and the training strategy code is C→D→E. The specific training suggestions are carried out in the following order:

[0506] m pronunciation induction (6): the lips are tightly closed, the soft palate is raised, the nasal passage is blocked, the airflow breaks through the obstruction of the lips, the vocal cords do not vibrate, and the airflow is weak;

[0507] Phoneme acquisition: monosyllabic words (7), disyllabic words (8), trisyllabic words (9);

[0508] Voice confusion: mb(10);

[0509] Assessment object 7: One pronunciation distortion, based on the assessment results and other data, the dysarthria feature code is obtained as

[0510] ·011 / 101 / 110, solution push: Type IV (D→E→F)

[0511] That is, one of the three pronunciations is misinterpreted. According to the three-layer training strategy table, the training strategy type pushed is type IV, and the training strategy code is D→E→F. The specific training suggestions are carried out in the following order:

[0512] Phoneme acquisition: monosyllabic words (7), disyllabic words (8), trisyllabic words (9);

[0513] Voice confusion: mb(10);

[0514] Voice training: voice repetition (11), voice switching (12), voice alternation (13);

[0515] Assessment object 8: The pronunciation was confused as b three times. Based on the assessment results and other data, the dysarthria feature code was obtained as

[0516] 000, solution push: type V (B→C→D)

[0517] That is, the three pronunciations are all confused as sound b. According to the three-layer training strategy table, the pushed training strategy type is type V, and the training strategy code is B→C→D. The specific training suggestions are carried out in the following order:

[0518] ·M pronunciation learning (5m): Oral resonance method can be used. Use the monosyllabic words "fish, pounce" and "clothing, adobe" for oral resonance training;

[0519] b. Pronunciation induction (6): the lips are tightly closed, the soft palate is lowered, the nasal passage is opened, the vocal cords vibrate, and air is sent out from the nasal cavity;

[0520] Phoneme acquisition: monosyllabic words (7), disyllabic words (8), trisyllabic words (9)

[0521] Assessment object 9: The second pronunciation was confused as b. Based on the assessment results and other data, the dysarthria feature code was obtained as

[0522] ·001 / 010 / 100, solution push: Type III (C→D→E)

[0523] That is, two of the three pronunciations are confused as b. According to the three-layer training strategy table, the training strategy type pushed is type III, and the training strategy code is C→D→E. The specific training suggestions are carried out in the following order:

[0524] b. Pronunciation induction (6): the lips are tightly closed, the soft palate is lowered, the nasal passage is opened, the vocal cords vibrate, and air is sent out from the nasal cavity;

[0525] Phoneme acquisition: monosyllabic words (7), disyllabic words (8), trisyllabic words (9);

[0526] Voice confusion: mb(10);

[0527] Assessment object 10: The pronunciation was confused as b. Based on the assessment results and other data, the dysarthria feature code was obtained as

[0528] ·011 / 101 / 110, solution push: Type IV (D→E→F)

[0529] That is, one of the three pronunciations is confused as b. According to the three-layer training strategy table, the training strategy type pushed is type IV, and the training strategy code is D→E→F. The specific training suggestions are carried out in the following order:

[0530] Phoneme acquisition: monosyllabic words (7), disyllabic words (8), trisyllabic words (9);

[0531] Voice confusion: mb(10);

[0532] Voice training: voice repetition (11), voice switching (12), voice alternation (13).

[0533] The protection content of the present invention is not limited to the above embodiments. Without departing from the spirit and scope of the present invention, changes and advantages that can be thought of by those skilled in the art are included in the present invention and are protected by the attached claims.

Claims

1. A method for intelligently pushing a Chinese pronunciation training program, characterized in that: The push method comprises the following steps: Step A: establishing a dysarthria feature code based on the speech evaluation data of the evaluation object; obtaining the evaluation data comprises the following steps: Step A1, collecting the voice of the evaluation object; Step A2: Processing and evaluating the collected speech to obtain a speech evaluation result of the evaluation object; In step A2, the speech evaluation results include: correct interpretation, misinterpretation, loss, confusion; wherein, The correct answer refers to the result of speech evaluation: the Chinese pinyin of the single word recognized by the speech of the evaluation object is consistent with the initial consonant and final vowel of the target Chinese pinyin; The missing refers to the result of speech evaluation that the Chinese pinyin of the single word obtained by the speech of the evaluation object is consistent with the final vowel of the target sound, and the initial consonant is not pronounced. express; The confusion refers to the speech evaluation result that the Chinese Pinyin initial consonant of the single word obtained by the speech of the evaluation object is different from the Chinese Pinyin initial consonant of the target sound, but is related to the Chinese Pinyin initial consonant of the target sound in pronunciation mode or pronunciation position; the speech evaluation result is recorded in the actual speech of the evaluation object; The misinterpretation refers to the speech evaluation result that the speech evaluation result of the evaluation object is not any of the following: correct answer, missing, or confused, and the speech evaluation result is represented by "×"; If the assessment result is misinterpretation or loss, the subject has deficiencies in both the pronunciation method and the place of articulation of the initial consonant; If the evaluation result is confusion, the confusion sound table is used to determine the deficiencies in the pronunciation method and articulation position of the initial consonant; The subject pronounced the word three times during the test. In each of the three pronunciations, If all three answers are correct, relevant exercises F should be pushed in the push plan to consolidate them; If there is a deficiency in pronunciation method or pronunciation position in one of the three tests, it means that the object of evaluation has learned the sound, but it is not stable enough and needs targeted practice and consolidation. The push practice method is: D→E→F; If there are deficiencies in pronunciation or articulation position in two of the three tests, it means that the object of evaluation has not yet stably acquired this sound and needs to be strengthened with the following practice: C→D→E; If the pronunciation method or pronunciation position is insufficient in all three tests, it means that the subject has problems with the pronunciation position and / or pronunciation method. In this case, basic training will be conducted on the pronunciation position and / or pronunciation method. The training method is as follows: if only the pronunciation position is insufficient, B→C→D will be pushed; if only the pronunciation method is insufficient, A→C→D will be pushed; if both the pronunciation position and pronunciation method are insufficient, A→B→C→D will be pushed; Among them, A is the articulation position exercise, B is the articulation method exercise, C is the pronunciation induction exercise, D is the phoneme training exercise, E is the phoneme contrast exercise, and F is the phonetic reinforcement exercise; The dysarthria feature coding is represented by α-T-β-γ-Δ; wherein, The α represents the age of the evaluation subject in months, α=0-999; The T indicates that the target sound of the test belongs to the initial consonant of Mandarin Chinese; The Mandarin Chinese initial consonants include: b, m, p, f, d, t, n, l, g, k, h, j, q, x, z, c, s, zh, ch, sh, r; The β represents the target sound, that is, the target initial consonant selected for testing among the Mandarin Chinese initial consonants; The γ represents the evaluation result of the speech of the evaluation object, which is represented by ε, η, λ, and specific confused sounds; the ε represents that the evaluation result of the speech of the evaluation object is correct; the η represents that the evaluation result of the speech of the evaluation object is distorted; the λ represents that the evaluation result of the speech of the evaluation object is lost; the specific confused sound represents that the evaluation result of the speech of the evaluation object is a specific confused sound related to the pronunciation method or pronunciation position of the target sound; The Δ represents the number of occurrences of the speech evaluation result corresponding to the code γ, and the numerical range of the number is 1-3; the method for establishing the articulation disorder feature code from the speech evaluation result is: Let α represent the age in months of the subject of this test, and fill in 0 on the left to make it a 3-digit number; T is used to indicate that the target sound in this test belongs to the initial consonant of Mandarin Chinese; Use β to represent the specific initial consonant of the target sound in this test; Let γ represent the evaluation result of the speech of the evaluation object in this test; Δ represents the number of occurrences of the speech evaluation results of the evaluation object in this test; The characters α, T, β, γ, and Δ are concatenated to obtain the final dysarthria feature code as α-T-β-γ-Δ; Step A3: generating speech evaluation data based on the speech evaluation result and combining the age of the evaluation object and the target sound; Step B: query the corresponding database data according to the articulation disorder feature code to obtain the Chinese articulation training program; in the data in the database, Each initial consonant has unique attributes, and the initial consonants and their corresponding attributes are summarized as a database initial consonant table in the database; the unique attributes of each initial consonant are: target sound, target sound acquisition age, target sound detailed information, target sound type, and single-, double-, and three-syllable test words corresponding to the target sound; each attribute of the initial consonant corresponds to a row of entries in the database; For each initial consonant, because of the different evaluation results, the γ-Δ composition code is also different, and each evaluation result code of each initial consonant is summarized as a database γ-Δ code-strategy chain correspondence table in the database; the database γ-Δ code-strategy chain correspondence table has the following attributes: γ-Δ code, target sound, evaluation result category, confused sound, number of errors, strategy chain; each attribute in the database γ-Δ code-strategy chain correspondence table corresponds to a row of entries in the database; For each initial consonant, the specific code in the strategy chain has a corresponding training suggestion chain, and a database training suggestion chain corresponding table is established; the database training suggestion chain corresponding table includes the following attributes: specific code in the strategy chain, target sound, and training suggestion chain; each attribute in the database training suggestion chain corresponding table corresponds to a row of entries in the database; For each training suggestion chain, the training method id in the training suggestion chain has corresponding specific content, and a database training method table is established, wherein the database training method table includes the following attributes: target sound, method id in the training suggestion chain, and specific content corresponding to the method id; each attribute in the database training method table corresponds to a column of entries in the database; Step C: pushing the Chinese pronunciation training program.

2. The intelligent push method according to claim 1, characterized in that: In step A1, the speech is tested and collected according to a test word list; the test word list is obtained by summarizing and arranging according to the Chinese phonetic combination rules in the "Basic Rules of Chinese Phonetic Orthography".

3. The intelligent push method according to claim 1, characterized in that: In step A1, the voice collection of the evaluation object is carried out through the following steps: 1) Use a microphone and ADC to connect to a computer, and the microphone collects the voice of the person being evaluated; 2) The ADC device samples at a rate of 48,000 times per second. Each sampling is cascaded through a multi-stage amplifier and comparator, and the analog signal of the voice is sampled into level signals of different voltages through the analog-to-digital conversion circuit. It is connected and transmitted to the computer via a 3.5mm interface or USB interface, and stored as a sound file for analysis.

4. The intelligent push method according to claim 1, characterized in that: In step A2, the processing of the collected speech includes: First, the audio in the sound file is filtered using the following algorithm: The state equation of the system: k =φ k，k-1 *x k-1 +Γ k-1 w k-1 , The measurement equation of the system: Z k =H k *x k +v k ; Where: w k-1 is the process white noise; x k is the system state at time k; H k are the parameters of the measurement system; φ k，k-1 is the state transition moment; v k To measure noise; Γ is the noise driving matrix; Z k Measure the output of the equation for the system; Take the optimal estimate x at time k-1 k-1 As a benchmark, predict the state change at time k At the same time, observe the state and obtain the observed variable Z k , and then use the observed value to correct the predicted value, so as to obtain the optimal state estimate x at time k k ; The filtered audio is the human voice audio with environmental noise removed, and the audio is further analyzed; The filtered human voice audio is divided into frames, 12,000 samples of 0.25 seconds are taken as one frame, and the entire audio is sliced; and the corresponding combination is found in the Chinese speech database. The algorithm involved is as follows: Assume that Q is the set of all possible hidden states and V is the set of all possible observed states, that is: Q={q1,q2,...,q N },V={v1,v2,...v M }; Where N is the number of possible hidden states and M is the number of all possible observed states; For a sequence of length T, I is the state sequence corresponding to sequence T, and O is the observation sequence corresponding to sequence T, that is: I={i1,i2,...,i T },O={o1,o2,...,o T }; Among them, any hidden state i t ∈Q, any observation state o t ∈V; The pronunciation of each Mandarin Chinese character is decomposed into a combination of multiple frames. In the Chinese speech database, there is a sequence set of Q and V containing each character; the input audio slice O is compared with V to find the corresponding character.

5. The intelligent push method according to claim 1, characterized in that: According to the articulation disorder feature code, in the process of querying the corresponding database data, When querying for specific training strategies, first locate the database initial consonant table through the target sound, and read the expected learning age of the initial consonant; if the evaluation object is younger than the expected learning age of the initial consonant, no training plan will be pushed; Secondly, according to the tested γ-Δ coding, according to the target sound and γ-Δ coding, the strategy chain corresponding to the target sound of the test is obtained in the database γ-Δ coding-strategy chain correspondence table; According to the specific code in the strategy chain corresponding to the target sound and the target sound, query the database training suggestion chain corresponding table to obtain the training suggestion chain of the specific code of the strategy chain; Finally, according to the target sound and the method ID in the training suggestion chain, query the database training method table, obtain the specific content corresponding to each method ID and combine them to obtain the complete training plan for this test.

Images