CN110827980A - Dysarthria grading evaluation method based on acoustic indexes - Google Patents
Dysarthria grading evaluation method based on acoustic indexes Download PDFInfo
- Publication number
- CN110827980A CN110827980A CN201911087835.4A CN201911087835A CN110827980A CN 110827980 A CN110827980 A CN 110827980A CN 201911087835 A CN201911087835 A CN 201911087835A CN 110827980 A CN110827980 A CN 110827980A
- Authority
- CN
- China
- Prior art keywords
- dysarthria
- intelligibility
- value
- acoustic
- voice
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 206010013887 Dysarthria Diseases 0.000 title claims abstract description 80
- 238000011156 evaluation Methods 0.000 title abstract description 20
- 238000000034 method Methods 0.000 claims abstract description 40
- 230000005856 abnormality Effects 0.000 claims abstract description 18
- 238000004458 analytical method Methods 0.000 claims abstract description 7
- 230000001575 pathological effect Effects 0.000 claims abstract description 3
- 230000035945 sensitivity Effects 0.000 claims abstract description 3
- 230000002776 aggregation Effects 0.000 claims description 9
- 238000004220 aggregation Methods 0.000 claims description 9
- 239000013598 vector Substances 0.000 claims description 9
- 238000004364 calculation method Methods 0.000 claims description 6
- 238000013139 quantization Methods 0.000 claims description 6
- 238000000638 solvent extraction Methods 0.000 claims description 6
- 238000013473 artificial intelligence Methods 0.000 claims description 5
- 201000010099 disease Diseases 0.000 claims description 4
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 claims description 4
- 238000007781 pre-processing Methods 0.000 claims description 4
- 206010042008 Stereotypy Diseases 0.000 claims description 3
- 238000005070 sampling Methods 0.000 claims description 3
- 150000001875 compounds Chemical class 0.000 claims description 2
- 238000000605 extraction Methods 0.000 claims description 2
- 238000011002 quantification Methods 0.000 claims description 2
- 208000006011 Stroke Diseases 0.000 description 19
- 238000011160 research Methods 0.000 description 9
- 206010008190 Cerebrovascular accident Diseases 0.000 description 7
- 230000002490 cerebral effect Effects 0.000 description 7
- 238000011282 treatment Methods 0.000 description 6
- 238000011161 development Methods 0.000 description 4
- 238000003745 diagnosis Methods 0.000 description 4
- 208000026106 cerebrovascular disease Diseases 0.000 description 3
- 210000003484 anatomy Anatomy 0.000 description 2
- 238000007689 inspection Methods 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000004393 prognosis Methods 0.000 description 2
- 238000006467 substitution reaction Methods 0.000 description 2
- 208000018737 Parkinson disease Diseases 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000013078 crystal Substances 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000002996 emotional effect Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000009472 formulation Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 201000006417 multiple sclerosis Diseases 0.000 description 1
- 210000000056 organ Anatomy 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 208000027765 speech disease Diseases 0.000 description 1
- 210000002105 tongue Anatomy 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/70—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for mining of medical data, e.g. analysing previous cases of other patients
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Public Health (AREA)
- Data Mining & Analysis (AREA)
- Biomedical Technology (AREA)
- Databases & Information Systems (AREA)
- Pathology (AREA)
- Epidemiology (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Medical Treatment And Welfare Office Work (AREA)
Abstract
The invention discloses a dysarthria grading evaluation method based on acoustic indexes, which comprises the following steps: collecting voice data of a patient with dysarthria after stroke; extracting objective acoustic feature intelligibility of the voice data, and collecting, extracting and analyzing acoustic indexes of a testee, wherein the objective acoustic feature intelligibility is a pathological voice analysis index with high acceptance and good sensitivity at home and abroad; determining the critical values of slight abnormality, moderate abnormality and severe abnormality of the voice by the obtained acoustic feature intelligibility based on an artificial intelligent automatic clustering algorithm K-means; the intelligibility threshold is used for clinically assessing the severity level of dysarthria of the subject. The method reduces the evaluation error caused by subjective factors such as technical level, subjective judgment, regional difference and the like of technicians, and is simple to operate and less in time consumption.
Description
Technical Field
The invention belongs to the technical field of sound processing, and particularly relates to a dysarthria grading assessment method based on acoustic indexes.
Background
The 3 rd cause of death retrospective sampling survey report of the Chinese medical society, which is completed in 2004-2005, shows that cerebrovascular diseases reach the first cause of death of the diseases in China. From 1990 to 2013, the number of the first-place provinces with life loss caused by stroke death in China is increased from 16 to 27. Wherein the death rate of urban stroke is 125.56/10 ten thousand, and the death rate of rural stroke is 150.17/10 ten thousand. According to the research report, about 2570 thousands of the world stroke survivors and about 650 thousands of the world stroke survivors died from the stroke in 2013. And the trend of the cerebral apoplexy patients in China is obvious, and the labor population of 40-64 years old accounts for nearly 50%. According to literature reports, the incidence rate of dysarthria in stroke patients is 30-40%. Speech intelligibility was lower in 69.6% of patients with parkinson's disease than in the normal group, with dysarthria incidence as high as 51% in patients with multiple sclerosis. The incidence rate of cerebrovascular diseases is extremely high, the trend of the cerebral apoplexy is obvious, and the incidence rate of dysarthria after cerebral apoplexy is also very high, so the research on the diagnosis, the evaluation and the treatment of the sequelae of the cerebrovascular diseases is also very important, the life quality of the population suffering from cerebral apoplexy is improved, and more complete research is provided for the diagnosis and the prognosis of the population suffering from cerebral apoplexy.
At present, domestic dysarthria is mainly evaluated in a subjective mode, and objective evaluation research is not common, because the research on the objective dysarthria evaluation lacks a method and a system for truly realizing objective evaluation. Like other types of objective assessment of speech disorders, objective assessment of the ability to mute requires objective evaluation parameters and hardware and software equipment. The objective assessment based on the voice characteristic system accords with the actual requirements of dysarthria assessment and rehabilitation application, is applied to spasmodic dysarthria cerebral apoplexy testees based on artificial intelligence, assesses whether the testees have dysarthria and grades the dysarthria, and establishes a new objective dysarthria grading method, thereby providing help for diagnosis, assessment, treatment plan and assessment curative effect of the dysarthria testees after cerebral apoplexy.
The closest tool for dysarthria grading to the invention is an assessment scale of questionnaire combined physical examination, namely a French dysarthria evaluation table, and the dysarthria severity grading method is mainly classified from the clinical aspects of anatomy, movement and the like, and is not classified by acoustic indexes, so that the objectivity is still lacked.
Disclosure of Invention
The invention mainly aims to overcome the defects and shortcomings of the prior art, provides a dysarthria grading evaluation method based on acoustic indexes, provides an objective evaluation method for dysarthria patients after stroke, relieves manpower, reduces errors caused by different manpower and technical levels, and provides a more objective theoretical basis for diagnosis, treatment, evaluation and prognosis of dysarthria.
In order to achieve the purpose, the invention adopts the following technical scheme:
a dysarthria grading evaluation method based on acoustic indexes comprises the following steps:
collecting voice data of a patient with dysarthria after stroke;
extracting objective acoustic feature intelligibility of the voice data, collecting B syllables of each person of A cases of patients clinically diagnosed as dysarthria through an interview-level recording pen, wherein the syllables comprise unit tones, compound vowels, consonants and tones, editing the syllables into B independent audio files through software, randomly extracting the independent audio files through Matlab software programming, recording according to randomly heard sounds by an audiologist, completing analysis of difference with target tones, and calculating acoustic feature intelligibility, which is a pathological voice analysis index with high acceptance and good sensitivity at home and abroad;
determining the critical values of slight abnormality, moderate abnormality and severe abnormality of the voice by the obtained acoustic feature intelligibility based on an artificial intelligent automatic clustering algorithm K-means;
the intelligibility threshold is used for clinically assessing the severity level of dysarthria of the subject.
As a preferred technical scheme, the Chinese dysarthria voice evaluation system word list is adopted for collecting voice data of a dysarthria patient after stroke, and the Chinese dysarthria voice evaluation system word list is composed of 82 target voices; the value of A is 100, and the value of B is 82.
As a preferred technical scheme, when voice data of a patient with dysarthria after stroke is collected, a SonyZoomH4nPro portable digital recorder is adopted for recording, the sampling rate is 44100Hz, the precision is 16bit, double channels are adopted for recording all testees, the testees take the end seats, a person who uses a pen holds the recorder, the distance between the lips of the testees and the recorder is about 10cm, the voice speed is natural and stable, the volume is moderate, and a word list is repeatedly recorded for 2 times.
As a preferred technical scheme, the voice data is clipped and archived by using audio clipping software, then the testee clips a single vowel by using Cool Edit Pro2.1 software, and the voice data is saved and archived by using a WAV format.
As a preferred technical scheme, judging the objective acoustic feature intelligibility, adopting two women and one male to respectively judge the objective acoustic feature intelligibility, then carrying out consistency check on intelligibility results of 3 persons, and then taking an average value of the intelligibility results to finally obtain the objective acoustic feature intelligibility of the testee; and (4) carrying out consistency check on the results of the intelligibility of 3 people, namely two women and one male, and objectively judging the accuracy of the intelligibility result of the acoustic feature.
The method comprises the following steps of preprocessing objective acoustic feature intelligibility, obtaining continuous audio files through sound recording, observing whether clipped syllables are moderate in intensity through voice software, enabling the amplitude of machine vibration to be 1/3-2/3 at the height of an acquisition frame without obvious noise or interference, observing and determining the starting point of F1 as the starting point through Praat3.0 software, enabling the ending point of F2 as the ending point, clipping 82 syllable samples of each patient by using Cool Edit Pro2.1, enabling two sound recording samples to be selected for each syllable before observation and clipping, and enabling the second sample only when the first sample cannot be used.
As a preferred technical solution, the extracting the objective acoustic feature intelligibility of the speech data is specifically implemented by:
b voice utterances of each person of the A cases of sound-forming patients are judged and recorded by the audiences in a random extraction mode, and 2 minutes is recorded if the recorded voice after the audiences understand the sounds is completely consistent with the target voice; each listener records 1 point if the recorded sound is inconsistent with the target sound (vowel, consonant and tone) after listening and understanding; each listener records 0 points if the recorded sound after understanding is completely inconsistent with the target sound or is a stereotypy nonsense sound, and if 82 syllables are completely accurate, the intelligibility quantization value 164(82 × 2) is completely incorrect, and the intelligibility quantization value is 0(82 × 0).
As a preferred technical scheme, the obtained acoustic feature intelligibility is determined by using an artificial intelligent automatic clustering algorithm K-means to determine the critical values of mild abnormality, moderate abnormality and severe abnormality of the speech, specifically:
8.1, N independent speech audios per patient with quantitative intelligibility values between 0-164;
8.2, each patient had a quantification value, X ═ X, to form a data set1,X2,X3,...,Xm}
8.3, determining the optimal classification number through the sum of squared errors of the squared errors, namely finally hopeing to divide dysarthria into several grades of severity degrees through a K-means method;
wherein Ci is the ith cluster, p is a sample point in Ci, mi is a centroid of Ci, the mean value of all samples in Ci, and SSE is a clustering error of all samples, in the calculation, if the sample division is finer with the increase of a clustering number k, the aggregation degree of each cluster is higher and higher, and the square sum of the error is gradually reduced, so that a k value is debugged from small to large in the calculation process, when k is smaller than a real clustering number, the decrease range of SSE is very large because the increase of k can greatly increase the aggregation degree of each cluster, and when k reaches the real clustering number, the return of the aggregation degree obtained by increasing k is rapidly reduced, at this time, the decrease range of SSE is rapidly reduced, and then the decrease tends to be gentle with the continuous increase of the k value, and the turning point which tends to be gentle is the k value which is finally determined, that is the classification number;
8.4, input the data set in Matlab2013b, set the value of k, determine the maximum iteration number N by automatic convergence, and output isCluster division C ═ { C1,C2,...Ck}
1) Randomly select k samples from dataset X as initial k centroid vectors:
2) n for N1, 2
a) Initializing cluster partitioning C to
b) For i 1,2.. m, sample X is calculatediAnd each centroid vector muj(j ═ 1, 2.):
x is to beiMinimum mark is dijClass λ corresponding to the minimumiAt this time, update
c) For j 1,2, k, pair CjRecalculate new centroid for all sample points in the image
e) If all k centroid vectors have not changed, go to step 3)
3) Output cluster partitioning C ═ C1,C2,...Ck}
And 8.5, calculating the value of K according to the step 8.4, namely the data set, the centroid value and the boundary value of the mild, moderate and severe dysarthria, and the boundary value of the mild and moderate dysarthria, wherein the boundary value of the severe and moderate dysarthria is the determined critical value.
As a preferred technical scheme, the optimal k value is calculated to be 3, which accords with the clinical habit of classifying the severity of diseases.
Compared with the prior art, the invention has the following advantages and beneficial effects:
1. compared with the existing dysarthria assessment method, the assessment error caused by subjective factors such as technical level, subjective judgment, regional difference and the like of technicians is reduced, the operation is simple, the consumed time is less, and about 5 minutes per person is achieved, so that the patient coordination degree is high, and a series of errors caused by factors such as patient noncoordination and the current emotional state are reduced.
2. The voice material storage can be extracted, analyzed and compared at any time, so that the analysis before and after treatment of a testee can be visualized and visualized, and the voice conditions of a plurality of different testees can be compared in parallel; the invention discloses an objective dysarthria assessment method based on artificial intelligence, which is closely combined with the technical level of the era, develops and innovates an objective assessment method suitable for modern people, is applied to modern technology and follows the development trend of the era.
3. The closest tool for dysarthria grading is an evaluation scale of questionnaire combined physical examination, namely a French dysarthria evaluation table, the dysarthria severity grading method is mainly classified from the clinical angles of anatomy, movement and the like, and acoustic indexes are not adopted for classification, but the invention is classified from acoustic parameters boundary values obtained by pure acoustic characteristics and parameters based on a machine learning algorithm, so that the method is more objective, convenient and fast, and the reliability of re-measurement is high.
Drawings
FIG. 1 is a flow chart of the present invention.
Detailed Description
The present invention will be described in further detail with reference to examples and drawings, but the present invention is not limited thereto.
Examples
In an indoor environment without obvious interference, a testee takes an end seat, a technician holds the SONY ZOOM H4NPRO professional recorder by hand, the distance between the lips of the testee and the recorder is about 10cm, before recording, the professional technician performs demonstration description on the testee, and the testee is reminded to keep the natural and stable speech speed during recording; after the recording is started, a subject reads 82 target sounds in the Chinese dysarthria voice evaluation system word list, the recording is carried out by using a professional recording device, and each target sound is repeatedly recorded for 2 times and is stored; 82 target voices are cut out respectively by using the Cool Edit Pro2.1 software, named and classified and filed.
And extracting target voice, and analyzing and calculating the intelligibility of objective acoustic features of the testee. The specific method is that a testee cuts archived voices by using Cool Edit Pro2.1 software, two female and male professional technicians are adopted to respectively judge the intelligibility of objective acoustic features of voice materials of the testee, each voice is adopted to completely read and count 0, partial errors (such as tone and consonant) count 1, the scoring standards for counting 2 are completely read, the total scores of the 3 technical personnel for auditory identification are respectively calculated, then the results of the total scores of the 3 technical personnel for auditory identification are subjected to consistency check to prove the accuracy, finally the total scores of the 3 technical personnel for auditory identification are averaged, the average value is obtained, the error is reduced, and the final total score of the auditory identification of the testee is obtained to be archived, and finally the intelligibility of the objective acoustic features of the testee is obtained.
In this embodiment, 82 speech utterances of each of 100 sound-making patients are randomly extracted to be judged and recorded by the audiences, and the sound recorded by each audiences after understanding is completely consistent with the target sound and recorded as 2 points, for example, if the target sound randomly emitted by a computer speaker is "ba" and the audiences hear "ba", the "ba" is recorded to prompt that the intelligibility is intact; each dialer records 1 point if the recorded sound is inconsistent with the target sound (vowel, consonant and tone) after listening and understanding, for example, a computer loudspeaker emits a target sound 'bi', and the dialer hears 'pi', records 'pi', and prompts that the intelligibility part is damaged; the recorded sound after each listener understands is completely inconsistent with the target sound or is recorded as 0 minutes for stereotypy meaningless pronunciation, for example, the target sound "du" is emitted by a computer loudspeaker, and the listener hears "e" and records "e" to indicate that the intelligibility is completely damaged. That is, if 82 syllables are completely accurate, the intelligibility quantization value 164(82 × 2) is completely incorrect, and the intelligibility quantization value is 0(82 × 0).
The method is adopted to obtain the objective acoustic feature intelligibility of 100 dysarthric patients after stroke, and the objective acoustic feature intelligibility of the tested person is centrally preprocessed to obtain continuous audio files through recording, wherein the preprocessing is to observe whether the clipped syllables accord with proper intensity through voice software, the amplitude of the mechanical vibration wave is 1/3-2/3 at the height of a collecting frame on average, no obvious noise or interference exists, the Praat3.0 software observes and determines the starting point of F1 as the starting point, the ending point of F2 is the end point, 82 syllable samples of each patient are clipped by adopting Cool Edit Pro2.1, each syllable has two recording samples before observation and clipping, and the second sample is started only when the first sample cannot be used.
And taking the preprocessed intelligibility as a dependent variable of an automatic clustering algorithm, determining critical values of slight abnormality, moderate abnormality and severe abnormality of the voice, and objectively evaluating the severity grade of dysarthria of each tested person by using the critical values so as to obtain the grade of the severity degree of dysarthria of each tested person. If (119, 129), greater than 129 is mild dysarthria, 119-129 is moderate dysarthria, and less than 119 is severe dysarthria.
The method can objectively evaluate whether the subject has dysarthria and the severity grade of the dysarthria, collects real and effective objective data by using the intelligibility of objective acoustic features, obtains a threshold value of the dysarthria voice severity by a k-means algorithm, intuitively reflects the severity of the dysarthria of the subject, reduces errors caused by subjective factors, and reduces result differences caused by the difference of technical levels of technical staff. The method is simple to operate and consumes less time, so that the matching degree of the user is high, the objective condition of the current dysarthria severity of the user can be reflected better, meanwhile, whether the general conditions of damage to sound-making organs such as lips and tongues of the user are limited or not and the limited specific direction and the damage severity are judged according to the condition of reading the target sound of the user and by combining the inspection of the acoustic index, a more accurate theoretical basis is provided for the evaluation of the user, an accurate direction is provided for the formulation of a rehabilitation treatment scheme of the user, and more individualized, objective and accurate rehabilitation treatment is provided for the user.
The invention researches a new dysarthria assessment method, is different from the traditional dysarthria assessment method, adopts an objective dysarthria assessment method, adopts an artificial intelligence method based on acoustic index inspection, researches from a new angle and a new direction, explains dysarthria patient classification from a voice angle and gives a boundary value, and is an objective and intelligent dysarthria assessment new method.
Determining the critical values of mild abnormity, moderate abnormity and severe abnormity of the voice by the obtained acoustic feature intelligibility based on an artificial intelligent automatic clustering algorithm K-means, and specifically comprising the following steps:
s1, 82 independent voice audios of each patient, wherein the quantitative intelligibility value of each voice audio is between 0 and 164;
s2, 100 patients each had M quantified values, making up a data set, X ═ X1,X2,X3,...,Xm}
S3, determining the optimal classification number through the sum of squared errors of the squared errors, namely, finally hope to divide dysarthria into several grades of severity degrees through a K-means method;
wherein Ci is the ith cluster, p is a sample point in Ci, mi is a centroid of Ci, the mean value of all samples in Ci, and SSE is a clustering error of all samples, in the calculation, if the sample division is finer with the increase of a clustering number k, the aggregation degree of each cluster is higher and higher, and the square sum of the error is gradually reduced, so that a k value is debugged from small to large in the calculation process, when k is smaller than a real clustering number, the decrease range of SSE is very large because the increase of k can greatly increase the aggregation degree of each cluster, and when k reaches the real clustering number, the return of the aggregation degree obtained by increasing k is rapidly reduced, at this time, the decrease range of SSE is rapidly reduced, and then the decrease tends to be gentle with the continuous increase of the k value, and the turning point which tends to be gentle is the k value which is finally determined, that is the classification number;
s4, inputting the data set in Matlab2013b, and setting the k valueThe maximum iteration number N is determined by automatic convergence, and the output is the cluster division C ═ C1,C2,...Ck}
1) Randomly select k samples from dataset X as initial k centroid vectors:
2) n for N1, 2
a) Initializing cluster partitioning C to
b) For i 1,2.. m, sample X is calculatediAnd each centroid vector mujDistance of (j ═ 1,2.. k):
x is to beiMinimum mark is dijClass λ corresponding to the minimumiAt this time, update
c) For j 1,2, k, pair CjRecalculate new centroid for all sample points in the image
e) If all k centroid vectors have not changed, go to step 3)
3) Output cluster partitioning C ═ C1,C2,...Ck}
And S5, calculating the value of K according to the step S4, namely the data set, the centroid value and the boundary value of the mild, moderate and severe dysarthria, and the boundary value of the mild and moderate dysarthria, wherein the boundary value of the severe and moderate dysarthria is the determined critical value.
In this example, the optimal k value is calculated to be 3, which is in accordance with clinical practice for classifying disease severity.
The objective method aiming at dysarthria assessment in China is few, and an objective assessment method is researched on the basis of the intelligibility of objective acoustic features, so that dysarthria assessment tends to be objective, a more standard and objective assessment method is researched, the difference of assessment results of dysarthria caused by subjective assessment is reduced, and the final assessment results tend to be objective and standardized.
The assessment method based on artificial intelligence frees manpower, depends on the intelligent development, is the crystal of the intelligent era, is combined with the intelligent development, is the result of era progress and scientific development, combines the assessment of dysarthria with the intelligence, enables the assessment result to be more objective and convenient for front-back comparison, intelligent analysis, convenient storage and research, reduces the difference of the assessment results of different institutions, and further researches the standardized dysarthria assessment method.
The above embodiments are preferred embodiments of the present invention, but the present invention is not limited to the above embodiments, and any other changes, modifications, substitutions, combinations, and simplifications which do not depart from the spirit and principle of the present invention should be construed as equivalents thereof, and all such changes, modifications, substitutions, combinations, and simplifications are intended to be included in the scope of the present invention.
Claims (9)
1. A dysarthria grading assessment method based on acoustic indexes is characterized by comprising the following steps:
collecting voice data of a patient with dysarthria after stroke;
extracting objective acoustic feature intelligibility of the voice data, collecting B syllables of each person of A cases of patients clinically diagnosed as dysarthria through an interview-level recording pen, wherein the syllables comprise unit tones, compound vowels, consonants and tones, editing the syllables into B independent audio files through software, randomly extracting the independent audio files through Matlab software programming, recording according to randomly heard sounds by an audiologist, completing analysis of difference with target tones, and calculating acoustic feature intelligibility, which is a pathological voice analysis index with high acceptance and good sensitivity at home and abroad;
determining the critical values of slight abnormality, moderate abnormality and severe abnormality of the voice by the obtained acoustic feature intelligibility based on an artificial intelligent automatic clustering algorithm K-means;
the intelligibility threshold is used for clinically assessing the severity level of dysarthria of the subject.
2. The method for graded assessment of dysarthria based on acoustic indicators as claimed in claim 1, wherein the voice data of dysarthria patients after stroke is collected from the "Chinese dysarthria voice assessment system vocabulary", which is composed of 82 target voices; the value of A is 100, and the value of B is 82.
3. The method for assessing the sound-enough obstacle level based on the acoustic indexes as claimed in claim 1, wherein when voice data of a patient with dysarthria after stroke is collected, a Sony Zoom H4nPro portable digital recorder is adopted for recording, the sampling rate is 44100Hz, the precision is 16bit, the double channels are adopted for recording all testees, the testees take end seats, a writer holds the recorder, the distance between the lips of the testees and the recorder is about 10cm, the voice speed is natural and stable, the volume is moderate, and the word list is recorded repeatedly for 2 times.
4. The method for assessing the sound-enough obstacle level based on the acoustic index as claimed in claim 3, wherein the voice data is edited and archived by using audio editing software, and then the examinee is edited by using Cool Edit pro2.1 software to perform individual vowel editing and archived by using WAV format.
5. The method for evaluating the level of sound obstruction based on the acoustic indexes according to claim 1, wherein objective acoustic feature intelligibility is judged, two women and one male are respectively adopted to judge the objective acoustic feature intelligibility, and then the intelligibility results of 3 persons are subjected to consistency check and then averaged to finally obtain the objective acoustic feature intelligibility of the testee; and (4) carrying out consistency check on the results of the intelligibility of 3 people, namely two women and one male, and objectively judging the accuracy of the intelligibility result of the acoustic feature.
6. The dysarthria grading assessment method based on acoustic indicators according to claim 1, further comprising preprocessing the objective acoustic feature intelligibility, obtaining continuous audio files by recording, wherein the preprocessing is to observe whether the clipped syllables are of moderate intensity by the speech software, the amplitude of the machine vibration is 1/3-2/3 with no apparent noise or interference at the height of the acquisition box, observe and determine the beginning of F1 as the starting point by the praat3.0 software, the end of F2 as the ending point, and 82 syllable samples are clipped for each patient by using CoolEdit pro2.1, each syllable has two recording samples before observation and clipping, and the second sample is enabled only when the first sample cannot be used.
7. The dysarthria grading assessment method based on acoustic indicators according to claim 1, wherein the objective acoustic feature intelligibility of the speech data is extracted by:
b voice utterances of each person of the A cases of sound-forming patients are judged and recorded by the audiences in a random extraction mode, and 2 minutes is recorded if the recorded voice after the audiences understand the sounds is completely consistent with the target voice; each listener records 1 point if the recorded sound is inconsistent with the target sound (vowel, consonant and tone) after listening and understanding; each listener records 0 points if the recorded sound after understanding is completely inconsistent with the target sound or is a stereotypy nonsense sound, and if 82 syllables are completely accurate, the intelligibility quantization value 164(82 × 2) is completely incorrect, and the intelligibility quantization value is 0(82 × 0).
8. The dysarthria grading assessment method based on acoustic indexes according to claim 1, wherein the obtained acoustic feature intelligibility is determined by using an artificial intelligence based automatic clustering algorithm K-means to determine the critical values of mild abnormality, moderate abnormality and severe abnormality of the speech, specifically:
8.1, N independent speech audios per patient with quantitative intelligibility values between 0-164;
8.2 Each patient had A quantification valuesData set, X ═ X1,X2,X3,...,Xm}
8.3, determining the optimal classification number through the sum of squared errors of the squared errors, namely finally hopeing to divide dysarthria into several grades of severity degrees through a K-means method;
wherein Ci is the ith cluster, p is a sample point in Ci, mi is a centroid of Ci, the mean value of all samples in Ci, and SSE is a clustering error of all samples, in the calculation, if the sample division is finer with the increase of a clustering number k, the aggregation degree of each cluster is higher and higher, and the square sum of the error is gradually reduced, so that a k value is debugged from small to large in the calculation process, when k is smaller than a real clustering number, the decrease range of SSE is very large because the increase of k can greatly increase the aggregation degree of each cluster, and when k reaches the real clustering number, the return of the aggregation degree obtained by increasing k is rapidly reduced, at this time, the decrease range of SSE is rapidly reduced, and then the decrease tends to be gentle with the continuous increase of the k value, and the turning point which tends to be gentle is the k value which is finally determined, that is the classification number;
8.4, the data set is input to Matlab2013b, k is set, and the maximum iteration number N is determined by the automatic convergence to output the cluster division C ═ C1,C2,...Ck}
1) Randomly select k samples from dataset X as initial k centroid vectors:
2) n for N1, 2
a) Initializing cluster partitioning C to
b) For i 1,2.. m, sample X is calculatediAnd each centroid vector mujDistance of (j ═ 1,2.. k):
x is to beiMinimum mark is dijClass λ corresponding to the minimumiAt this time, update
c) For j 1,2, k, pair CjRecalculate new centroid for all sample points in the image
e) If all k centroid vectors have not changed, go to step 3)
3) Output cluster partitioning C ═ C1,C2,...Ck}
And 8.5, calculating the value of K according to the step 8.4, namely the data set, the centroid value and the boundary value of the mild, moderate and severe dysarthria, and the boundary value of the mild and moderate dysarthria, wherein the boundary value of the severe and moderate dysarthria is the determined critical value.
9. The method of claim 8, wherein the optimal k value is calculated to be 3, which is suitable for clinical practice in classifying the severity of the disease.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911087835.4A CN110827980B (en) | 2019-11-08 | 2019-11-08 | Dysarthria grading evaluation method based on acoustic indexes |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911087835.4A CN110827980B (en) | 2019-11-08 | 2019-11-08 | Dysarthria grading evaluation method based on acoustic indexes |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110827980A true CN110827980A (en) | 2020-02-21 |
| CN110827980B CN110827980B (en) | 2022-08-26 |
Family
ID=69553643
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911087835.4A Active CN110827980B (en) | 2019-11-08 | 2019-11-08 | Dysarthria grading evaluation method based on acoustic indexes |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110827980B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111599347A (en) * | 2020-05-27 | 2020-08-28 | 广州科慧健远医疗科技有限公司 | Standardized sampling method for extracting pathological voice MFCC (Mel frequency cepstrum coefficient) features for artificial intelligence analysis |
| CN112798102A (en) * | 2021-04-13 | 2021-05-14 | 盛瑞传动股份有限公司 | Sound quality evaluation method, device, system and medium |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104598758A (en) * | 2015-02-12 | 2015-05-06 | 上海市徐汇区中心医院 | System and method for evaluating hearing-speech rehabilitation training and curative effect of patients with post-stroke dysarthria |
| US20160351074A1 (en) * | 2004-09-16 | 2016-12-01 | Lena Foundation | Systems and methods for expressive language, developmental disorder, and emotion assessment, and contextual feedback |
| CN107456208A (en) * | 2016-06-02 | 2017-12-12 | 深圳先进技术研究院 | The verbal language dysfunction assessment system and method for Multimodal interaction |
-
2019
- 2019-11-08 CN CN201911087835.4A patent/CN110827980B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160351074A1 (en) * | 2004-09-16 | 2016-12-01 | Lena Foundation | Systems and methods for expressive language, developmental disorder, and emotion assessment, and contextual feedback |
| CN104598758A (en) * | 2015-02-12 | 2015-05-06 | 上海市徐汇区中心医院 | System and method for evaluating hearing-speech rehabilitation training and curative effect of patients with post-stroke dysarthria |
| CN107456208A (en) * | 2016-06-02 | 2017-12-12 | 深圳先进技术研究院 | The verbal language dysfunction assessment system and method for Multimodal interaction |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111599347A (en) * | 2020-05-27 | 2020-08-28 | 广州科慧健远医疗科技有限公司 | Standardized sampling method for extracting pathological voice MFCC (Mel frequency cepstrum coefficient) features for artificial intelligence analysis |
| CN111599347B (en) * | 2020-05-27 | 2024-04-16 | 广州科慧健远医疗科技有限公司 | Standardized sampling method for extracting pathological voice MFCC (functional peripheral component interconnect) characteristics for artificial intelligent analysis |
| CN112798102A (en) * | 2021-04-13 | 2021-05-14 | 盛瑞传动股份有限公司 | Sound quality evaluation method, device, system and medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110827980B (en) | 2022-08-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Al-Nasheri et al. | An investigation of multidimensional voice program parameters in three different databases for voice pathology detection and classification | |
| Mekyska et al. | Robust and complex approach of pathological speech signal analysis | |
| Matos et al. | Detection of cough signals in continuous audio recordings using hidden Markov models | |
| US11786171B2 (en) | Method and system for articulation evaluation by fusing acoustic features and articulatory movement features | |
| CN103730130B (en) | A kind of detection system of pathological voice | |
| CN107170445B (en) | Parkinsonism detection device based on voice mixed information feature collaborative optimization judgment | |
| CN110827980B (en) | Dysarthria grading evaluation method based on acoustic indexes | |
| Wallen et al. | A screening test for speech pathology assessment using objective quality measures | |
| CN111951824A (en) | Detection method for distinguishing depression based on sound | |
| Vojtech et al. | Refining algorithmic estimation of relative fundamental frequency: Accounting for sample characteristics and fundamental frequency estimation method | |
| Ding et al. | Deep connected attention (DCA) ResNet for robust voice pathology detection and classification | |
| El Emary et al. | Towards developing a voice pathologies detection system | |
| Cordella et al. | Classification-based screening of Parkinson’s disease patients through voice signal | |
| Parrot et al. | Independent and automatic evaluation of speaker-independent acoustic-to-articulatory reconstruction | |
| CN108766462B (en) | Voice signal feature learning method based on Mel frequency spectrum first-order derivative | |
| Li et al. | Improvement on speech depression recognition based on deep networks | |
| US20210090734A1 (en) | System, device and method for detection of valvular heart disorders | |
| Panek et al. | Quantification of linear and non-linear acoustic analysis applied to voice pathology detection | |
| CN108962397B (en) | Pen and voice-based cooperative task nervous system disease auxiliary diagnosis system | |
| CN116723793A (en) | Automatic physiological and pathological assessment based on speech analysis | |
| CN111028863B (en) | Method for diagnosing post-stroke dysarthria tone errors based on neural network and diagnostic device thereof | |
| Alimuradov et al. | A method to determine cepstral markers of speech signals under psychogenic disorders | |
| Mallol-Ragolta et al. | Covid-19 detection exploiting self-supervised learning representations of respiratory sounds | |
| Marczyk et al. | Optimizing linguistic materials for feature-based intelligibility assessment in speech impairments | |
| Sol et al. | Towards Automated Vocal Mode Classification in Healthy Singing Voice—An XGBoost Decision Tree-Based Machine Learning Classifier |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| PE01 | Entry into force of the registration of the contract for pledge of patent right |
Denomination of invention: A Classification and Evaluation Method for Articulation Disorders Based on Acoustic Indicators Effective date of registration: 20231024 Granted publication date: 20220826 Pledgee: CITIC Bank Co.,Ltd. Guangzhou Branch Pledgor: Guangzhou Kehui Jianyuan Medical Technology Co.,Ltd. Registration number: Y2023440000074 |
|
| PE01 | Entry into force of the registration of the contract for pledge of patent right |