Some standard content:
National Standard of the People's Republic of China
Speech Articulation Testing Method
Acoustics-Speech Articulation Testing Nethod1Subject Content and Scope of Application
GB/T 15508—1995
This standard specifies the method of using speech signals to directly test and evaluate the working quality of various speech transmission systems. This standard is applicable to the quality evaluation of speech communication systems, the quality evaluation of hall sound quality and the quality evaluation of human-machine speech communication systems, and can also be used for speech listening tests.
2 Terminology
2.1 Speech intelligibility The ratio of a language unit (sentence, word or syllable) uttered by a speaker to be correctly recognized by a listener through a speech transmission system. 2.2 Speech intelligibility test The ratio of one or more listeners to correctly record one or more incoherent syllables uttered by a speaker to quantitatively measure the quality of a speech transmission system. 2.3 Articulation score The ratio of one or more listeners to correctly record one or more incoherent syllables uttered by a speaker in a speech intelligibility test, measured in white fraction.
2.4 Speech transmission system The medium or electroacoustic system or a combination of the two that transmits speech signals from a speaker to a listener. 2.5 Testing jury
A group of trained participants in a speech intelligibility test consisting of an appropriate proportion of male and female speakers and listeners. 2.6 Voice level level
The sound level measured 1 m in front of the speaker when he/she is speaking. 2.7 Carrier phrase
In order to make the pronunciation natural and smooth, the short phrase with the syllable to be tested is inserted in it. 3 Test method
3.1 Test material
The language intelligibility test uses the Chinese intelligibility test syllable table - KXY table. There are 10 KXY tables in a set (see Appendix A1 for KXY1 to KXY10). Each table has 75 syllables. When reading aloud, it is randomly combined and divided into 25 groups. Each group has three syllables and is preceded by a carrier phrase. K艾Y The table is a phonetically balanced syllable table, and each table has good equivalence. 3.1.1 Usage
Each time the test is performed, the syllables in the syllable table should be randomly recombined into 25 two-syllable words. Each group of syllables should not form a word. At least two syllable tables are used for each test, one for odd numbers and one for even numbers. After all ten tables have been used, the syllable combinations can be rearranged and used again. 3.2 Testers
State Administration of Technical Supervision 1995-03-15 Approved for implementation on December 1, 1995
3.2. 1 Speakers
GB/T 15508—1995
Speakers should be able to speak correct Mandarin and have language defects. The number, gender and age of speakers shall be determined according to the specific test situation. Generally, there shall be at least one male and one female speaker.
3.2.2 Listeners
Listeners should have normal hearing (pure tone audiometry, hearing loss below 4000H shall not exceed 10dB), be able to speak correct Mandarin and be able to dictate proficiently. Generally, there shall be no less than 10 listeners. 3.2.3 Training
Before the formal test, the test team shall undergo necessary training. During the training, the significance and method of the intelligibility test shall be explained to the team members first. The male and female speakers shall each read Table KXY1 and Table KXY10 once, and the listeners shall make records under normal listening conditions. The records shall be recorded and the listeners shall correct each other. The training work is 10 minutes for each reading of two test sheets. If the clarity score still does not reach a stable value, the training time is extended. Through the training process, the test personnel are screened as necessary. 3.2.4 Special communication system test team
When conducting speech clarity test on the special communication system, it is advisable to use the users of the system as test team members. 3.3 Pronunciation
3.3.1 Speaking sound level
The standard speaking sound pressure level is 65±3dB at 1m in front of the speaker, using the C-weighted and slow response monitoring sound pressure level of the sound level meter. The sound level meter should comply with the national standard GB3785 "Electroacoustic performance and test methods of sound level meters". 3.3.2 Speaking speed
The speaker pronounces naturally and smoothly at a normal speaking speed (4 syllables per second). There is a pause of 5 to 10 seconds between each group of syllables for the listener to complete the recording.
3.3.3 Guide Sentences
When the speaker reads the syllable table, every three syllables form a group, and a guide sentence is added in front: "The ×th meaning group is", for example, it is read as: \The eleventh group is 布, 且, 多".
3.3.4 Misreading
To avoid misreading, the host should give pronunciation in advance for the polyphones that appear on the test table. When there is a misreading in the pronunciation, it is not stressed or corrected. When correcting and grading, the deletion of misread syllables will not be counted. Each test table should not have more than 10 misread syllables, otherwise the table will be invalid. 3.3.5 On-site test
The speaker must read the items from the head of the clarity test record table, and the listener should record them one by one. When there are many test items, a list should be compiled. Test number. After the header items are read, the test section can be read aloud. 3.3.6 Recording Test
When it is not convenient for the speaker to test on site, a high-fidelity recording of a pre-made clarity test table can also be used for testing. Note that the recording material should not introduce reverberation, noise and various distortions. The reverberation time and its frequency characteristics should comply with GYI26--86 "Acoustic Design Specifications and Technical Requirements for Cable Broadcasting and Recording Broadcasters and Technical Rooms", and the allowable noise should comply with the speech recording studio standards in GYJ42--89 "Permissible Noise Standards for Technical Rooms in Radio and Television Centers"; the recording equipment should comply with the national standard "General Technical Conditions and Performance Requirements for Mixing Consoles", GB177879 & Broadcast Recorders 3, GB 1779—79&I/Class A standard for broadcast recording tapes. 3.4 Recording
3.4.7 Recording method
The listener generally uses Chinese characters or Pinyin to write down the (syllable) pronunciation heard on a special speech clarity test record sheet. When conditions permit, the keyboard can be used to input into the computer. 3.4.2 Recording requirements
The listener should concentrate on faithfully recording every sound heard. The handwriting should be clear when recording. 3-4-3 Record verification
In order to facilitate the verification of the speaker's actual pronunciation, the speaker's pronunciation can be recorded on the spot during the on-site test. 3.4.4 Speech clarity test record sheet
GB/T15508—1995
During the test, the listener should use a special The speech intelligibility test record sheet is shown in Appendix A2. Each sheet is for one listener to record one syllable table at a time.
3.4.5 Recording content
The listener should fill in the items one by one from the beginning of the table following the pronunciation of the speaker (except for the scoring items). Only two syllables to be tested should be written in each line of the record column, and the introductory sentence should not be recorded.
3.5 Scoring
3.5.1 Scoring personnel
Special scoring can be organized, or the listeners can exchange record sheets with each other for scoring. No personal scoring is required. When the record is input by keyboard, the computer can score automatically.
3.5.2 Scoring method
Provide the standard answers to the test syllable table to the scorer, and the scorer will record the wrong records and correct records in the true and false column. Blank records are counted as errors.
3.5.3 Calculation of scores
Calculate the number of incorrect syllables E for each recorded syllable table, then the number of correct responses N=75-E. Therefore, the syllable clarity score of the recorded table is:
For the conversion table, see Appendix A3.
If the test is performed with a competent listener, the average syllable clarity score obtained is: 2s./
Test standard deviation:
3.5.4 Screening
(2)
-(3)
When the difference between the score of a recorded syllable table and the average syllable clarity score of the whole table is greater than one standard deviation, the recorded table is discarded and the average syllable clarity score is recalculated according to formula (2). 3.6 Diagnosis of speech transmission quality
In addition to providing syllable clarity scores as a quality evaluation of the transmission system, speech perception confusion analysis methods can also be used to further diagnose quality problems of the transmission system (see Appendix B, References). 3.7 Word intelligibility
The test process of word intelligibility is more complicated than that of syllable intelligibility.Listeners need to be fully trained, and it is generally not suitable to be conducted on site. It can be converted through the statistical relationship between the syllable clarity score and monotonous intelligibility obtained in experiments under various conditions (see Appendix, Reference).
GB/T15508—1995
Appendix A
Speech intelligibility test syllable table, record attenuation and score conversion table (supplement)
Speech intelligibility test syllable table
Speech intelligibility test syllable table
Speech intelligibility test syllable table
Speech intelligibility test syllable table
Speech intelligibility test syllable table
KXY table
chudng
Speech intelligibility test syllable table
Speech intelligibility test syllable table
Speech intelligibility test syllable table
GB/T 15508
zhuing
shuang
Speech clarity test syllable table
Speech clarity test syllable table
CB/T15508-1995
Speech clarity test syllable table
GB/T15508-1995
Speech clarity test syllable table
-KXY10
A2Speech clarity test record table
Working conditions
Speaker
Listener
GB/T 155081995
Clarity test record table
Test number
GB/T 155081995
Number of correctly recorded syllables - syllable clarity score conversion tableNumber of correctly recorded
Score, %
Number of incorrectly recorded
Number of correctly recorded
Appendix B
Quality diagnosis of language transmission system
(reference)
Score, %
Number of incorrectly recorded
The clarity test score of a language transmission system can only indicate the overall quality level of the system, but cannot indicate where its quality problems lie. Therefore, for a language transmission system, in addition to giving its clarity index, it is also necessary to conduct quality diagnosis in order to point out the specific direction of improving quality and level. The quality diagnosis method introduced in this reference is a supplement and development of the clarity test. The clarity test score only calculates the correct response of the listener, that is, the part of the language information that is correctly transmitted by the transmission system. The language transmission quality diagnosis specifically examines the part that makes errors in the language transmission process, that is, the analysis of homophonic mishearing. B1 Data CollectionbZxz.net
B1. Registration of Mishearing
For syllables recorded incorrectly by the listener, the sent syllable and the error record should be recorded in Chinese Pinyin in the corresponding group of the "Correct or Error" threshold of the record table. The sent syllable is in front, and the error record is in the back, connected by a short dash in the middle. When there are more than two errors in the same group, they are separated by commas. B1.2 Analysis of Mishearing
To facilitate mishearing analysis, the registered correct or error records should be transferred to the speech analysis registration table. There are three types of speech analysis tables: initial consonant analysis table, final consonant analysis table, and tone analysis table. The error records are analyzed according to the three types of errors, namely initial consonant, final consonant, and tone, and filled in the corresponding analysis tables respectively. In general, only initial consonant analysis can be performed. The first column on the left side of the analysis table indicates the sent initial consonant, final consonant, or tone. The horizontal line of the analysis table indicates the speech that the listener misheard when sending the speech marked on the left side of the line. If there is one mishearing (one wrong speech record of one listener), one stroke is recorded in the corresponding box, which can be accumulated by marking "correct\". Obviously, in the speech analysis table, each position on the diagonal from the upper left to the lower right should be blank, which indicates the position of the correct record. According to the statistics of wrong records and the number of speech occurrences in the syllable table and the number of listeners, the number of correct records can be calculated and filled in the speech analysis tables of all male speakers and all female speakers, and one set of speech analysis tables for male and female speakers respectively. Finally, from the following, in general, the error of initial consonants is the main reason for the reduction of clarity, so only the initial consonants can be analyzed for mishearing. It is necessary to mark the number of mishearings registered in the above analysis table with numbers, and write the accumulated results in the corresponding box. GB/T 15508—1995
Start with the speech feature with the lowest clarity score, and examine it step by step. According to the distribution law of this feature, find out the defects of the speech transmission system. If necessary, you can also find out the main problems of the system from some individual initials with very low scores. In order to make an accurate diagnosis of the transmission system, it is necessary to enter the number of speech occurrences in the report section table, see B3~B5. Then combine the analysis results of male and female speakers into a set of speech analysis tables. According to the data registered in the speech registration table, further speech analysis is carried out. The speech analysis table should be filled in separately for each speaker. The initial consonants of Mandarin Chinese can be divided into 4 categories according to the pronunciation method, which are: According to the above classification, calculate the speech characteristics of each pronunciation method and each pronunciation part separately, and the same analysis can be carried out on the finals and tones. b. Aspirated stops and affricates: p,,k,y,ch,cB1. 3 Correct and False Statistics
According to the pronunciation position, the sound sequence can be divided into three categories, which are: Table B1
h.+, d, t, nl, j+g, x+zh, ch+sh.rB2 Phonetic Analysis
Nasal: m, II
Unvoiced Fricative: fh, x, sh, s
B2.1 Phonetic Classification
Fricative: 1
Lateral:
c. Solitary:
Front r: bp, m, f, zc, s
Actual Number of Occurrences
B2.2 Phonetic Feature Clarity
e, After: g, k, h
B2.3 Diagnosis and Gradient
B3KXY Table of Initial Consonants
Use a larger number of syllables.
Should Occur Number| |tt||GB/T15508-1995
Continued Table B1
Actual number of occurrences
Expected number of occurrences
Number of occurrences of finals in KXY table
Actual number of occurrences
Expected number of occurrences
GB/T15508-1995
Continued Table B2
Actual number of occurrences
Imaginary number of occurrences
Number of occurrences of tones in KXY table
Actual number of occurrences
Expected number of occurrences
Appendix C
Statistical relationship between monotonous intelligibility and syllable clarity score (reference)
The statistical relationship curve between word intelligibility and syllable clarity score given in Figure C1 can be used to find out the corresponding word intelligibility from the syllable clarity. The following analytical formula can also be used to convert the syllable clarity into the monosyllabic intelligibility W1, disyllabic intelligibility W and trisyllabic intelligibility W. The occurrence rate of tetrasyllabic words is less than 1%, so this result is not considered.In addition to providing its clarity index, quality diagnosis is also required to point out the specific direction of improving quality and level. The quality diagnosis method introduced in this reference is a supplement and development of the clarity test. The clarity test score only calculates the correct response of the listener, that is, the part of the language information that is correctly transmitted by the transmission system. The language transmission quality diagnosis is specifically to examine the part of the language transmission process that is wrong, that is, to conduct homophonic mishearing analysis. B1 Data Collection
B1. Mishearing Registration
For the syllables recorded by the listener, the syllables and the error records should be recorded in the corresponding group of the "correct and error" threshold of the record table. The sent syllables are in front and the error records are in the back, connected by a short dash in the middle. When there are more than two errors in the same group, they are separated by commas. B1.2 Mishearing Analysis
To facilitate mishearing analysis, the registered correct and error records should be transferred to the speech analysis registration table. There are three types of speech analysis tables: initial consonant analysis table, final vowel analysis table, and tone analysis table. The error records are analyzed according to the three types of errors: initial consonants, final consonants, and tones, and filled in the corresponding analysis tables. In general, only initial consonant analysis can be performed. The first column on the left side of the analysis table indicates the initial consonant, final consonant, or tone sent. The horizontal line of the analysis table indicates the speech that the listener misheard when sending the speech marked on the left side of the line. If there is one mishearing (one wrong speech record of one listener), one stroke is recorded in the corresponding box, which can be accumulated by marking "correct\". Obviously, in the speech analysis table, each position on the diagonal from the upper left to the lower right should be blank, which indicates the position of the correct record. According to the statistics of wrong records and the number of speech occurrences in the syllable table and the number of listeners, the number of correct records can be calculated and filled in the speech analysis tables of all male speakers and all female speakers, and one set of speech analysis tables for male and female speakers respectively. Finally, from the following, in general, the error of initial consonants is the main reason for the reduction of clarity, so only the initial consonants can be analyzed for mishearing. It is necessary to mark the number of mishearings registered in the above analysis table with numbers, and write the accumulated results in the corresponding box. GB/T 15508—1995
Start with the speech feature with the lowest clarity score, and examine it step by step. According to the distribution law of this feature, find out the defects of the speech transmission system. If necessary, you can also find out the main problems of the system from some individual initials with very low scores. In order to make an accurate diagnosis of the transmission system, it is necessary to enter the number of speech occurrences in the report section table, see B3~B5. Then combine the analysis results of male and female speakers into a set of speech analysis tables. According to the data registered in the speech registration table, further speech analysis is carried out. The speech analysis table should be filled in separately for each speaker. The initial consonants of Mandarin Chinese can be divided into 4 categories according to the pronunciation method, which are: According to the above classification, calculate the speech characteristics of each pronunciation method and each pronunciation part separately, and the same analysis can be carried out on the finals and tones. b. Aspirated stops and affricates: p,,k,y,ch,cB1. 3 Correct and False Statistics
According to the pronunciation position, the sound sequence can be divided into three categories, which are: Table B1
h.+, d, t, nl, j+g, x+zh, ch+sh.rB2 Phonetic Analysis
Nasal: m, II
Unvoiced Fricative: fh, x, sh, s
B2.1 Phonetic Classification
Fricative: 1
Lateral:
c. Solitary:
Front r: bp, m, f, zc, s
Actual Number of Occurrences
B2.2 Phonetic Feature Clarity
e, After: g, k, h
B2.3 Diagnosis and Gradient
B3KXY Table of Initial Consonants
Use a larger number of syllables.
Should Occur Number| |tt||GB/T15508-1995
Continued Table B1
Actual number of occurrences
Expected number of occurrences
Number of occurrences of finals in KXY table
Actual number of occurrences
Expected number of occurrences
GB/T15508-1995
Continued Table B2
Actual number of occurrences
Imaginary number of occurrences
Number of occurrences of tones in KXY table
Actual number of occurrences
Expected number of occurrences
Appendix C
Statistical relationship between monotonous intelligibility and syllable clarity score (reference)
The statistical relationship curve between word intelligibility and syllable clarity score given in Figure C1 can be used to find out the corresponding word intelligibility from the syllable clarity. The following analytical formula can also be used to convert the syllable clarity into the monosyllabic intelligibility W1, disyllabic intelligibility W and trisyllabic intelligibility W. The occurrence rate of tetrasyllabic words is less than 1%, so this result is not considered.In addition to providing its clarity index, quality diagnosis is also required to point out the specific direction of improving quality and level. The quality diagnosis method introduced in this reference is a supplement and development of the clarity test. The clarity test score only calculates the correct response of the listener, that is, the part of the language information that is correctly transmitted by the transmission system. The language transmission quality diagnosis is specifically to examine the part of the language transmission process that is wrong, that is, to conduct homophonic mishearing analysis. B1 Data Collection
B1. Mishearing Registration
For the syllables recorded by the listener, the syllables and the error records should be recorded in the corresponding group of the "correct and error" threshold of the record table. The sent syllables are in front and the error records are in the back, connected by a short dash in the middle. When there are more than two errors in the same group, they are separated by commas. B1.2 Mishearing Analysis
To facilitate mishearing analysis, the registered correct and error records should be transferred to the speech analysis registration table. There are three types of speech analysis tables: initial consonant analysis table, final vowel analysis table, and tone analysis table. The error records are analyzed according to the three types of errors: initial consonants, final consonants, and tones, and filled in the corresponding analysis tables. In general, only initial consonant analysis can be performed. The first column on the left side of the analysis table indicates the initial consonant, final consonant, or tone sent. The horizontal line of the analysis table indicates the speech that the listener misheard when sending the speech marked on the left side of the line. If there is one mishearing (one wrong speech record of one listener), one stroke is recorded in the corresponding box, which can be accumulated by marking "correct\". Obviously, in the speech analysis table, each position on the diagonal from the upper left to the lower right should be blank, which indicates the position of the correct record. According to the statistics of wrong records and the number of speech occurrences in the syllable table and the number of listeners, the number of correct records can be calculated and filled in the speech analysis tables of all male speakers and all female speakers, and one set of speech analysis tables for male and female speakers respectively. Finally, from the following, in general, the error of initial consonants is the main reason for the reduction of clarity, so only the initial consonants can be analyzed for mishearing. It is necessary to mark the number of mishearings registered in the above analysis table with numbers, and write the accumulated results in the corresponding box. GB/T 15508—1995
Start with the speech feature with the lowest clarity score, and examine it step by step. According to the distribution law of this feature, find out the defects of the speech transmission system. If necessary, you can also find out the main problems of the system from some individual initials with very low scores. In order to make an accurate diagnosis of the transmission system, it is necessary to enter the number of speech occurrences in the report section table, see B3~B5. Then combine the analysis results of male and female speakers into a set of speech analysis tables. According to the data registered in the speech registration table, further speech analysis is carried out. The speech analysis table should be filled in separately for each speaker. The initial consonants of Mandarin Chinese can be divided into 4 categories according to the pronunciation method, which are: According to the above classification, calculate the speech characteristics of each pronunciation method and each pronunciation part separately, and the same analysis can be carried out on the finals and tones. b. Aspirated stops and affricates: p,,k,y,ch,cB1. 3 Correct and False Statistics
According to the pronunciation position, the sound sequence can be divided into three categories, which are: Table B1
h.+, d, t, nl, j+g, x+zh, ch+sh.rB2 Phonetic Analysis
Nasal: m, II
Unvoiced Fricative: fh, x, sh, s
B2.1 Phonetic Classification
Fricative: 1
Lateral:
c. Solitary:
Front r: bp, m, f, zc, s
Actual Number of Occurrences
B2.2 Phonetic Feature Clarity
e, After: g, k, h
B2.3 Diagnosis and Gradient
B3KXY Table of Initial Consonants
Use a larger number of syllables.
Should Occur Number| |tt||GB/T15508-1995
Continued Table B1
Actual number of occurrences
Expected number of occurrences
Number of occurrences of finals in KXY table
Actual number of occurrences
Expected number of occurrences
GB/T15508-1995
Continued Table B2
Actual number of occurrences
Imaginary number of occurrences
Number of occurrences of tones in KXY table
Actual number of occurrences
Expected number of occurrences
Appendix C
Statistical relationship between monotonous intelligibility and syllable clarity score (reference)
The statistical relationship curve between word intelligibility and syllable clarity score given in Figure C1 can be used to find out the corresponding word intelligibility from the syllable clarity. The following analytical formula can also be used to convert the syllable clarity into the monosyllabic intelligibility W1, disyllabic intelligibility W and trisyllabic intelligibility W. The occurrence rate of tetrasyllabic words is less than 1%, so this result is not considered.3症凉
B3KXY table shows the number of initial consonants appearing
Use a syllable table with a larger number.
Expected number of occurrences
GB/T15508-1995
Continued Table B1
Actual number of occurrences
Expected number of occurrences
Number of occurrences of finals in KXY table
Actual number of occurrences
Expected number of occurrences
GB/T15508-1995
Continued Table B2
Actual number of occurrences
Imaginary number of occurrences
Number of occurrences of tones in KXY table
Actual number of occurrences
Expected number of occurrences
Appendix C
Statistical relationship between monotonous intelligibility and syllable clarity score (reference)
The statistical relationship curve between word intelligibility and syllable clarity score given in Figure C1 can be used to find out the corresponding word intelligibility from the syllable clarity. The following analytical formula can also be used to convert the syllable clarity into the monosyllabic intelligibility W1, disyllabic intelligibility W and trisyllabic intelligibility W. The occurrence rate of tetrasyllabic words is less than 1%, so this result is not considered.3症凉
B3KXY table shows the number of initial consonants appearing
Use a syllable table with a larger number.
Expected number of occurrences
GB/T15508-1995
Continued Table B1
Actual number of occurrences
Expected number of occurrences
Number of occurrences of finals in KXY table
Actual number of occurrences
Expected number of occurrences
GB/T15508-1995
Continued Table B2
Actual number of occurrences
Imaginary number of occurrences
Number of occurrences of tones in KXY table
Actual number of occurrences
Expected number of occurrences
Appendix C
Statistical relationship between monotonous intelligibility and syllable clarity score (reference)
The statistical relationship curve between word intelligibility and syllable clarity score given in Figure C1 can be used to find out the corresponding word intelligibility from the syllable clarity. The following analytical formula can also be used to convert the syllable clarity into the monosyllabic intelligibility W1, disyllabic intelligibility W and trisyllabic intelligibility W. The occurrence rate of tetrasyllabic words is less than 1%, so this result is not considered.
Tip: This standard content only shows part of the intercepted content of the complete standard. If you need the complete standard, please go to the top to download the complete standard document for free.