CN106073706A - A kind of customized information towards Mini-mental Status Examination and audio data analysis method and system - Google Patents
A kind of customized information towards Mini-mental Status Examination and audio data analysis method and system Download PDFInfo
- Publication number
- CN106073706A CN106073706A CN201610382274.0A CN201610382274A CN106073706A CN 106073706 A CN106073706 A CN 106073706A CN 201610382274 A CN201610382274 A CN 201610382274A CN 106073706 A CN106073706 A CN 106073706A
- Authority
- CN
- China
- Prior art keywords
- subjects
- customized information
- mini
- feature
- mental status
- 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
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/48—Other medical applications
- A61B5/4803—Speech analysis specially adapted for diagnostic purposes
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/40—Detecting, measuring or recording for evaluating the nervous system
- A61B5/4076—Diagnosing or monitoring particular conditions of the nervous system
- A61B5/4088—Diagnosing of monitoring cognitive diseases, e.g. Alzheimer, prion diseases or dementia
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7235—Details of waveform analysis
- A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
Abstract
The invention discloses a kind of customized information towards Mini-mental Status Examination and audio data analysis method and system.By gathering the personal information of subjects, allow subjects complete MMSE scale regulation exercise question and record pronunciation audio frequency simultaneously, extract the acoustic features of speech audio for pathological characteristics, and utilize high-order statistic to characterize；Then the method utilizing feature selection reduces the redundancy of feature；The personal information of the acoustic features after dimensionality reduction with subjects is merged and obtains individualized feature；Finally utilize the pathological model of pronunciation situation and the Mini-Mental scale cognition relation obtaining data construct subjects, and utilize the method for cross validation to be analyzed.The present invention need not the treatment of any invasive, need only to analyze the condition of the most measurable subjects of relation obtained between data and pathological model, saving review time and money, the misery that minimizing subjects stands in checking process avoids doctor's subjectivity simultaneously and judges the impact on result.
Description
Technical field
The invention belongs to digital medical field, be specifically related to a kind of customized information towards Mini-mental Status Examination and
Audio data analysis method and system.
Background technology
Nervous system disorder, the similar conditions such as including senile dementia, parkinson, small vessel disease, apoplexy, it is mostly by greatly
Brain, spinal cord, cranial nerve cell sustains damage and causes the change of body.The each organ of human body is with main performance the most directly or indirectly
Being under neural regulation control, therefore nervous system sustains damage, and it mainly shows as being difficult to normally, the most effectively
The inconvenience of pronunciation and action, the aspect such as decrease of cognitive function.Wherein effective as the daily affective interaction of people of language performance
Approach, sustain damage and not only affect the health mental health of patient, and the work of strong influence people and life.Along with
Social progress, pressure constantly increases, and has millions people to suffer from the disease of psychiatric system class every year, and along with population
Progressively tending to aging, continuation is also increased by these statistical magnitudes.Therefore voice quality declines as nervous system disorder disease
Early manifestation, how the quality of life of people is had extremely important by phonetic analysis judgement nervous system disease by research
Meaning.
So far, medical circle is analyzed sound quality and is mainly made subjectivity according to the Professional knowledge of doctor with practical experience and sentence
Disconnected；Or by the laryngoscopy projects such as endoscope, not only elapsed time, add the financial burden of patient simultaneously, and give
Patient causes great misery.Along with the development of human-computer interaction technology, human-computer interaction technology progressively relates to military affairs, medical treatment,
Multiple fields such as education.Therefore the method that research computer professional assists treatment, utilizes Digital Signal Processing to extract voice
Effective acoustical signal in audio frequency, obtains simple, and quickly, the pathological analysis of Noninvasive checks system, can reduce the master of doctor
Seeing and judge with invasive instrument, the judgement of the state of an illness is checked the misery brought to patient, tool is of great significance at present.
Mini-mental Status Examination is the preliminary Screening Scale for nervous functional defects, by answering in checking process
Problem, can obtain reacting the seriality score of this subjects's nervous functional defects state.In traditional acoustical signal is analyzed, one
As carry out two-value classification based training with the relation whether suffering from disease by analyzing sound characteristic, can only draw subjects ill and
The most ill two kinds of results, and unpredictable probability and the degree suffering from nervous system disorder with judgement subjects.
Summary of the invention
The present invention is directed to the problems referred to above, propose a kind of customized information towards Mini-mental Status Examination (MMSE) and sound
Frequently data analysis method and system.The method obtains the phonetic entry of subjects by simple mike, and uses this voice
Audio extraction has the sound characteristic of resistivity and merges the customized information of subjects simultaneously and merged accordingly noise
Feature is analyzed training, it is achieved that analysis to subjects's voice data under different noisy environments, by this analysis result
Combine with the MMSE score data of subjects, it is also possible to obtain subjects's mental status detection score further.
The technical solution used in the present invention is as follows:
A kind of customized information towards Mini-mental Status Examination (MMSE) and audio data analysis method, specifically include
Following steps:
(1) customized information and the pronunciation voice data of subjects are obtained；
(2) from the pronunciation voice data obtained, the effectively section of recording is extracted；
(3) from effective recording section, extract audio frequency characteristics, and merge with the customized information of subjects and obtain personalization
Fusion feature；
(4) it is trained the personalized fusion feature obtained analyzing, obtains the analysis result of subjects's voice data.
Further, the customized information of described subjects includes: the age of subjects, sex, education degree, the most sick
Histories etc. react the information of its personal characteristics.
Further, in order to avoid subjects is due to psychentonia, the reason impacts on appraisal result such as emotion is impatient, make
Subjects keeps naturalness the most as far as possible, does not select confined space, and is selected at common place and completes Mini-Mental shape
State scale sets problem；In order to get rid of educational level equality aspect to understanding difference produced by exercise question, by medical practitioner to quilt
Examination person guides, and helps subjects to understand done problem, completes as requested to arrange exercise question, according to subjects just
Often the performance under state is marked.
Further, in order to keep the universality of recorded voice, recording conditions does not select recording studio etc. to be applicable to reality
The recording location tested, simply completes audio recording, in recording process, in order to just embody subjects while completing scale
Normal physiological status, it is desirable to subjects, in the case of comfortable sitting posture, keeps the distance between recording equipment and lip, at nature
Intonation and loudness of a sound pronunciation in the case of, i.e. deliberately do not improve or force down tone.Before recording, doctor carries out demonstration and sends out
Sound, and allow subjects carry out exercise pronunciation, treat that it, in the case of loosening, starts recording.
Further, step (2) including: the pronunciation voice data obtained is carried out end-point detection or artificial speech is cut
Point, intercept the recording section that can reflect user voice feature, and reject sky, bad waits the recording affecting model training quality
Audio frequency, in order to verify training method usable condition under daily recording conditions, adds the height of different signal to noise ratio to recorded audio
This noise, obtains the recording audio of different signal to noise ratio.
Further, in step (3), according to the pathological characters of nervous system disorder, from effective recording section, extract sound
Frequently feature, and customized information with subjects merges again after characterizing with high-order statistic.
Further, the pathological characters of nervous system disorder mainly shows as intensity of phonation or breath is more weak；Pronunciation is mingled with
The aspects such as breathing noise；Some slight changes of phonatory organ.Therefore it is the versatility testing this algorithm flow further, pin
To above three pathology aspect, extract conventional traditional characteristic and represent above-mentioned pathological characters.
Further, for reducing the external factor impacts on feature description effect such as environmental noise, select to be usually used in sound
The high-order statistic of the suppressed influence of noise in detection of activity field describes the statement of the sound property extracted for pathology.
Further, step (3) also includes that the audio frequency characteristics characterizing high-order statistic carries out feature selection, defeated to reduce
Enter the information redundancy between audio frequency characteristics.Be there is the key character of discrimination, i.e. between class distance variance within clusters greatly by selection
Little feature, reduces the complexity of training, improves the precision of model；Also can carry out feature selection by the method for cross validation, carry
The high stability selecting feature.Feature as extracted the most enough is simplified, and this step also can be omitted.
Further, be generally directed to pathology sound judge disease the most only to extract can preferably state sound pronunciation this
The feature of body, and have ignored the relatedness between sound and speaker.Here it is considered that the personal traits of subjects, such as educational level
Flat, the factor such as age, brain cell nerve is controlled the impact of expressive faculty and phonatory organ motor capacity, its body can be reacted
Body state and educational level equality customized information are extracted, and blend with general sound characteristic and obtain more preferably reacting its body
Body and the feature representation of mental status.
Further, the customized information expression of results after merging trains regression model as final feature input,
For ensureing the vigorousness of training pattern, we randomly select training sample structure state of an illness model of fit and avoid occurring the feelings of over-fitting
Condition.
The matching training pattern obtained based on audio frequency and personalized fusion feature, the cognitive regression analysis of subjects and detection
Method is specific as follows:
1) method utilizing cross validation.
2) counting statistics amount or result is carried out statistical analysis.
Further, in order to verify the stability of model, we utilize the method for cross validation to comment fitting effect
Valency.
It is possible to further by point analysis method such as fitting result and actual value contrast obtain the accuracy of training pattern with
Vigorousness.
A kind of customized information towards Mini-mental Status Examination (MMSE) and audio data analysis system, including:
Data acquisition module, for gathering the customized information of subjects, Mini-mental Status Examination score data and
Pronunciation voice data；
Data preprocessing module, for carrying out pretreatment to the pronunciation voice data of subjects, to extract effective recording
Section；
Pathology identification model construction module, for building pathology identification model according to effective recording section；Include again:
Feature extraction submodule, for extracting audio frequency characteristics from effective recording section；
Individualized feature fusant module, for melting the audio frequency characteristics of extraction with the customized information of subjects
Close, build pathology identification model；
Pathological data regression analysis module, for according to the Mini-mental Status Examination score data gathered and structure
Pathology identification model carries out regression analysis to pathological data, obtains the mental status scoring of subjects.
Further, described feature extraction submodule is recorded section from effective according to the pathological characters of nervous system disorder
Extract audio frequency characteristics and characterize with high-order statistic.
Further, above-mentioned pathology identification model construction module also includes feature selection submodule, for subjects
Customized information merge before the audio frequency characteristics that high-order statistic characterized by dimensionality reduction select.
Compared with prior art, the present invention has the advantage that as follows with good effect:
1) The present invention gives by training acoustic characteristic and Mini-Mental scale regression relation model analysis subjects's sound
The computer-aid method of frequency evidence.
System based on the method can detect whether subjects suffers from nervous disorder, it is to avoid passes through in patient checking process
The misery that intrusive mood apparatus measures is brought, saves patient's checking process simultaneously and waits time energy spent in result.
2) present invention carries out feature extraction according to the pathological characteristics that nervous disorder may cause, and calculates the high-order system of feature
Metering is as final input feature vector.
Validity feature for pathological reaction extraction audio frequency can preferably embody the feature of this type of disease, makes extraction feature more
Add comprehensive and reliable, suppress the noise impact on feature analysis by high-order statistic, even if recording audio is noisy at noise
Also can obtain under environment and extract good feature.
3) personal information of the audio frequency characteristics of extraction with subjects is merged by the present invention, obtains personalized fusion spy
Levy.
The high-order statistic of the information and traditional feature that add other passages merges, and obtains reflecting subjects from various dimensions
The feature of information, is more beneficial for analyzing the relation between subjects and nervous system disorder cognition.
Accompanying drawing explanation
Fig. 1 is scene operation schematic diagram of the present invention.
Fig. 2 is area of computer aided schematic flow sheet of the present invention.
Fig. 3 is the detailed description of the invention schematic diagram of the present invention.
Fig. 4 is cross validation feature selection system of selection schematic diagram of the present invention.
Detailed description of the invention
For making those skilled in the art be better understood from the present invention, hereafter by specific embodiment, and combine accompanying drawing, do
Detailed description, but be not construed as limiting the invention.
The present invention can pass through such as Fig. 1, and 2 understand operation scenario of the present invention and frameworks, mainly include gathering training data, data
Pretreatment, builds pathology identification model and data simulation and analysis four part, and wherein subjects only need to gather training data, its
He can predict whether suffer from mental sickness and degree by area of computer aided.Its algorithm is under conditions of matlab7.10.0
Process and training data is tested, as follows:
1) gathering training data, the work of this part is the basis of experiment, prepares with training for follow-up process, its tool
The instrument of body is as follows with parameter:
(1) subjects's information gathering and MMSE scale check part, first pass through and inquire the information of subjects as standby
Case obtains the record of personalization.This scale is carried out under the guiding of medical practitioner, by the exchange and interdynamic of doctor Yu subjects, note
Record subjects provides phase reserved portion to the performance level of problem according to its professional judgement.When exercise question is understood difficult by patient or produces
During raw ambiguity, doctor provides reasonably explanation and helps subjects to understand problem with demonstration.Unlike inspection before, this
Check and support that user is carried out on computers, interactive result is obtained real-time electronical record and preserves, checking process is more accelerated
Victory, convenient, effectively.The present invention is the highest to the requirement of computer equipment, meets normal viewing, alternately, gathers and use.
(2) sound collection part, in order to allow subjects not fettered by equipment, employs the external Mike of Sen Haisaier
Wind, for keep admission data consistent, it is desirable to when subjects gathers voice with mike distance be about 10cm, monophonic, sampling
Rate is 44110 hertz.The speech analysis of present stage mainly includes that continuous speech analysis and single vowel pronunciation are analyzed, due to continuously
There is the linguistics problems such as dysarthria in voice, therefore we select continuous speech analysis.The pronunciation of its medial vowel/ah/ and its
His vowel effect is roughly the same, and therefore application claims subjects is in the case of comfortable, keep as far as possible normal tone and
Volume sends vowel/ah/, it is desirable to every subjects records 3 times.
2) data prediction, the work of this part is the base component of whole work, is carried out clearly by the voice data of collection
Reason, rejects fail data, extracts the reciprocal fraction of scale simultaneously, prepare for subsequent step.
(1) Audio Processing part, the work of this part is the effective audio frequency extracting and can reacting subjects's pronunciation characteristics, its
Main step is carried out the most successively:
I. removing bad audio frequency by speech detection or the method for manual confirmation, empty audio frequency etc. records situation.
Ii. the too high bass excessively started over is removed by end-point detection or the artificial method intercepted.
Iii. the Gaussian noise adding different signal to noise ratio in existing audio frequency obtains new speech audio, simulated environment noise
The impact that sound is differentiated
(2) extracting customized information and scale total score, this part mainly extracts main information, the letter merged as personalization
Cease the criterion with fitting result to preserve respectively.
3) building pathology identification model, the work of this part is whole intermediate portions, the most effectively extracts acoustic data and enters
Row training, the grader obtaining stalwartness carries out more preferable regression analysis to pathology sound and specifically studies as follows.
(1) feature extraction.Sound characteristic is the live part representing its pronunciation situation, the most how to extract validity feature table
Self the pronunciation situation levying subjects is significant with the training of condition model.Here we carry according to pathological effects
Take sound characteristic, in order to suppress the effect of noise such as environment to use the high-order statistic of feature as final expression way.
A) nervous system disorder can produce several respects impact to patient, extracts feature according to pathology specific as follows:
I. the pronunciation situation of sound channel organ.In the case of subjects is health, the pronunciation situation of its sound channel should be the cycle
Property vibrations, in order to reflect subjects's controlling extent to phonatory organ, the present invention uses the tradition such as jitter, shimmer special
Levy the periodicity of information measurement sound channel, in order to preferably state, other periodical measurement method can be attempted from now on.
Ii. signal to noise ratio, not completely closed due to sound channel may have more pronunciation to producing substantial amounts of pathology sound
Noise, therefore has the noise that more pronounces, and how to extract and calculate effective audio frequency ratio energy effecting reaction in a large amount of noises
The health condition of subjects.
Iii. the slight change of phonatory organ, if subjected to the impact of the factors such as nervous disorder, subjects is at phonation
In, tongue, the phonatory organ such as lip have trickle change, can select mel frequency cepstral
These type of trickle changes of feature description such as coefficients (MFCC).
According to foregoing description, extract above-mentioned several respects feature statement audio frequency according to pathological characteristics.In recent years, frequency domain character,
The feature extracting methods such as wavelet transformation show preferable effect, can be as the direction of later Selecting research.
B) high-order statistic characteristic feature,
In signal processing applications, high-order statistic all has in gaussian sum nongausian process and nonlinear system are applied
Preferably effect.When processing voice signal, high-order statistic suppression Gaussian noise keeps its phase information, by voice messaging from height
Separating in this noise, the high-order statistic the most often extracting voice signal observational characteristic carries out the detection of activity of voice.Here
Carry out feature extraction for pathological characteristics, calculate corresponding high-order statistic as feature representation thus suppress environment noise etc. because of
The element impact on extracting characteristic effect.
(2) feature selection
Although there is not the linear relationship determined between feature and grader, but higher when extracting intrinsic dimensionality, super
Training speed and the accuracy of categorizing system can be reduced when going out certain scope.Actually some feature does not has or comprises few
Information, there is between feature certain repeatability simultaneously, classification results is not the most affected by it, the most how to reduce feature
Dimension, the efficiency and precision tool improving training is of great significance.We utilize the method for cross validation to carry out feature choosing
Select, as shown in Figure 4:
(1) selection of character subset and the data selecting method of cross validation are identical.As a example by 10 folding cross validations, every time
M N-dimensional audio frequency is randomly divided into 10 parts, and in turn by wherein 90% as training set, data volume is M*90%, remains 10% conduct
Test set, data volume is M*10%.
(2) by the method for certain feature selection, the N-dimensional feature of character subset is carried out dimensionality reduction, after obtaining ten dimensionality reductions
N (n < N) dimensional vector.The feature selection result of 10 secondary data collection is identical in theory, but actual and differ, therefore with to result
Add up.
(3) application voting mechanism, first creates the feature selection result that an empty set local is final, for the most one-dimensional
K (K is 1 ... the scalar of N), we select the frequency of occurrences the highest from 10*K characteristic element and the most nonoptional feature is put
Enter final feature selection set.
(3) individualized feature information fusion
Customized information merges detectivity and the credibility that can improve information, expands the range of information so that propose letter
Described thing can preferably be expressed by breath.Sound characteristic is only utilized at present generally to be fitted classification based training, but single-pass
The signal characteristic that road information characteristics obtains is the most coarse, is difficult to preferably express the personal considerations of subjects.If by certain
Processing method, obtains multi-faceted information from multiple passages simultaneously, information is carried out comprehensively, be that these information complement each other, completely
Embody much information characteristic and perception information acoustic information supplemented thus describe environment or subjects more accurately
The state of itself.Therefore in sound characteristic, we add the age, sex, and education degree etc. can its popular feeling of multi-faceted embodiment
In and the characteristic information of state at heart obtain final fuse information.
(4) model training
Want the state of an illness to carry out accurate matching with prediction it is necessary to train a stable grader.Here individual character will be obtained
Changing fusion feature, as we have M audio frequency, the personalized fusion feature N-dimensional of each audio frequency, using M N-dimensional feature as the most gloomy
Woods, the input feature vector of the methods such as neutral net is trained, and obtains accurately, and pathological examination is carried out by sane recurrence grader
Effectively analyze.
4) regression analysis and prediction
The method generally having traditional distribution method and cross validation is fitted training analysis, for guaranteeing the reliable of result
Property, the method for commonly used cross validation, wherein the method for 10 folding cross validations is the most commonly used, carries out classification based training and divides with returning
Analysis.For the effectiveness of the result, conventional match value and actual value error mean absolute classification
The effectiveness of the statistical result characterization results such as error (MAE).
Generally Mini-mental Status Examination is when adding up to total score, and 8 points and 9 points all by 0 point of calculating.Best result is 30 points.27-
30 are divided into normal condition, 21-26 to think suffers from slight nervous functional defects disease, and 10-20 suffers from the neuro-cognitive merit of moderate
Energy disease, 0-9 is divided into the nervous functional defects disease of severe.Divide simultaneously and whether suffer from mental sickness and have with schooling
Close, if therefore old people be illiteracy not only less than 17 points, primary school not only less than 20 points, more than middle school but also less than 24 points, be then severe
Nervous functional defects disease, therefore analytical data is compared with existing standard, can learn subjects's P with
Degree.
Being compared with existing standard by the scoring results of model prediction, if prediction score value is 24 points, wherein score value is at 21-
26 think suffer from slight nervous functional defects disease, therefore, it is determined that subjects may suffer from mild neurocognitive function problem.
The system realized based on said method, can find the pass of pararthria and function of nervous system's system by regression training
System, allows subjects by nature, exchange free of a burden in common scenarios, carries according to the universal pathological characters of nervous disorder simultaneously
Take speech audio corresponding speech audio feature to train, carry out regression analysis with scale mark, can be to the cognition of subjects
Whether function has obstacle carries out preliminary analysis and screening.And when reality is applied, patient only need to steadily pronounce, Ji Keli
It is analyzed detection with existing model.
Above example is merged by individualized feature to be analyzed only in order to the technical side of the present invention to be described sound simulation
Case rather than be limited, those of ordinary skill in the art, can be right without departing from the spirit and scope of the present invention time
Technical scheme is modified or equivalent, and protection scope of the present invention should be as the criterion with described in claim.
Claims (10)
1., towards customized information and the audio data analysis method of Mini-mental Status Examination, specifically include following step
Rapid:
(1) customized information and the pronunciation voice data of subjects are obtained；
(2) from the pronunciation voice data obtained, the effectively section of recording is extracted；
(3) from effective recording section, extract audio frequency characteristics, and merge with the customized information of subjects and obtain personalized fusion
Feature；
(4) it is trained the personalized fusion feature obtained analyzing, obtains the analysis result of subjects's voice data.
2. as claimed in claim 1 towards customized information and the audio data analysis method of Mini-mental Status Examination, its
Being characterised by, the customized information of described subjects includes: the age of subjects, sex, education degree, medical history.
3. as claimed in claim 1 towards customized information and the audio data analysis method of Mini-mental Status Examination, its
Being characterised by, step (2) including: the pronunciation voice data obtained is carried out end-point detection or artificial speech cutting, and picks
Remove the recorded audio affecting model training quality, then recorded audio is added the Gaussian noise of different signal to noise ratios, obtains effectively
Recording section.
4. as claimed in claim 1 towards customized information and the audio data analysis method of Mini-mental Status Examination, its
It is characterised by, in step (3), according to the pathological characters of nervous system disorder, from effective recording section, extracts audio frequency characteristics, and
Merge with the customized information of subjects again after characterizing with high-order statistic.
5. as claimed in claim 4 towards customized information and the audio data analysis method of Mini-mental Status Examination, its
Being characterised by, step (3) also includes: the audio frequency characteristics characterizing high-order statistic carries out feature selection.
6. as claimed in claim 5 towards customized information and the audio data analysis method of Mini-mental Status Examination, its
Being characterised by, step (3) also includes: have the feature of discrimination or the method for cross validation to higher order statistical scale by selection
The audio frequency characteristics levied carries out feature selection.
7. towards customized information and the audio data analysis system of Mini-mental Status Examination, including:
Data acquisition module, for gathering the customized information of subjects, Mini-mental Status Examination score data and pronunciation
Voice data；
Data preprocessing module, for the pronunciation voice data of subjects is carried out pretreatment, extracts the effectively section of recording；
Pathology identification model construction module, for building pathology identification model according to effective recording section, includes again:
Feature extraction submodule, for extracting audio frequency characteristics from effective recording section；
Individualized feature fusant module, for the audio frequency characteristics of extraction is merged with the customized information of subjects, structure
Build pathology identification model；
Pathological data regression analysis module, for according to the Mini-mental Status Examination score data gathered and the pathology of structure
Identify that model carries out regression analysis to pathological data, obtain the mental status scoring of subjects.
8. as claimed in claim 7 towards customized information and the audio data analysis system of Mini-mental Status Examination, its
Being characterised by, described pretreatment includes: pronunciation voice data is carried out end-point detection or artificial speech cutting, rejects and affect mould
The recorded audio of type training quality, and recorded audio is added the Gaussian noise of different signal to noise ratio.
9. as claimed in claim 7 towards customized information and the audio data analysis system of Mini-mental Status Examination, its
Being characterised by, described feature extraction submodule extracts audio frequency according to the pathological characters of nervous system disorder from effective recording section
Feature also characterizes with high-order statistic.
10. as claimed in claim 7 towards customized information and the audio data analysis system of Mini-mental Status Examination, its
Being characterised by, described pathology identification model construction module also includes feature selection submodule, in the personalization with subjects
The audio frequency characteristics that high-order statistic is characterized by dimensionality reduction before merging by information selects.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610382274.0A CN106073706B (en) | 2016-06-01 | 2016-06-01 | A kind of customized information and audio data analysis method and system towards Mini-mental Status Examination |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610382274.0A CN106073706B (en) | 2016-06-01 | 2016-06-01 | A kind of customized information and audio data analysis method and system towards Mini-mental Status Examination |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106073706A true CN106073706A (en) | 2016-11-09 |
| CN106073706B CN106073706B (en) | 2019-08-20 |
Family
ID=57446834
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610382274.0A Active CN106073706B (en) | 2016-06-01 | 2016-06-01 | A kind of customized information and audio data analysis method and system towards Mini-mental Status Examination |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106073706B (en) |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107016233A (en) * | 2017-03-14 | 2017-08-04 | 中国科学院计算技术研究所 | The association analysis method and system of motor behavior and cognitive ability |
| CN107273677A (en) * | 2017-06-08 | 2017-10-20 | 中国科学院软件研究所 | A kind of multi-channel nerve function quantitative evaluation system |
| CN108806720A (en) * | 2017-05-05 | 2018-11-13 | 京东方科技集团股份有限公司 | Microphone, data processor, monitoring system and monitoring method |
| CN108962397A (en) * | 2018-06-06 | 2018-12-07 | 中国科学院软件研究所 | A kind of multichannel multitask the nervous system disease assistant diagnosis system based on pen and voice |
| CN109448820A (en) * | 2018-10-10 | 2019-03-08 | 上海整合医学研究院有限公司 | A kind of wearable mental disease voice assisting in diagnosis and treatment equipment |
| CN109473170A (en) * | 2018-10-26 | 2019-03-15 | 首都医科大学附属北京安定医院 | Schizoid system is diagnosed using cognition index |
| CN109480864A (en) * | 2018-10-26 | 2019-03-19 | 首都医科大学附属北京安定医院 | A kind of schizophrenia automatic evaluation system based on nervous functional defects and machine learning |
| CN109616141A (en) * | 2019-01-03 | 2019-04-12 | 燕山大学 | Heterophemia detection method |
| CN109800790A (en) * | 2018-12-24 | 2019-05-24 | 厦门大学 | A kind of feature selection approach towards high dimensional data |
| CN111462773A (en) * | 2020-03-26 | 2020-07-28 | 心图熵动科技(苏州)有限责任公司 | Suicide risk prediction model generation method and prediction system |
| CN113082447A (en) * | 2021-04-02 | 2021-07-09 | 电子科技大学 | Prediction method for music modulation brain plasticity effect of fMRI brain loop |
| WO2021189903A1 (en) * | 2020-10-09 | 2021-09-30 | 平安科技（深圳）有限公司 | Audio-based user state identification method and apparatus, and electronic device and storage medium |
| CN115969331A (en) * | 2023-03-21 | 2023-04-18 | 中国医学科学院生物医学工程研究所 | Cognitive language disorder assessment system |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1839749A (en) * | 2005-04-01 | 2006-10-04 | 浙江工业大学 | Computer aided device for early identifying and predicting senile dementia |
| US20080201144A1 (en) * | 2007-02-16 | 2008-08-21 | Industrial Technology Research Institute | Method of emotion recognition |
| CN101739869A (en) * | 2008-11-19 | 2010-06-16 | 中国科学院自动化研究所 | Priori knowledge-based pronunciation evaluation and diagnosis system |
| CN103021406A (en) * | 2012-12-18 | 2013-04-03 | 台州学院 | Robust speech emotion recognition method based on compressive sensing |
| CN103778913A (en) * | 2014-01-22 | 2014-05-07 | 苏州大学 | Pathologic voice recognizing method |
| CN103956171A (en) * | 2014-04-01 | 2014-07-30 | 中国科学院软件研究所 | Multi-channel mini-mental state examination system |
| CN104900229A (en) * | 2015-05-25 | 2015-09-09 | 桂林电子科技大学信息科技学院 | Method for extracting mixed characteristic parameters of voice signals |
| CN105448291A (en) * | 2015-12-02 | 2016-03-30 | 南京邮电大学 | Parkinsonism detection method and detection system based on voice |
-
2016
- 2016-06-01 CN CN201610382274.0A patent/CN106073706B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1839749A (en) * | 2005-04-01 | 2006-10-04 | 浙江工业大学 | Computer aided device for early identifying and predicting senile dementia |
| US20080201144A1 (en) * | 2007-02-16 | 2008-08-21 | Industrial Technology Research Institute | Method of emotion recognition |
| CN101739869A (en) * | 2008-11-19 | 2010-06-16 | 中国科学院自动化研究所 | Priori knowledge-based pronunciation evaluation and diagnosis system |
| CN103021406A (en) * | 2012-12-18 | 2013-04-03 | 台州学院 | Robust speech emotion recognition method based on compressive sensing |
| CN103778913A (en) * | 2014-01-22 | 2014-05-07 | 苏州大学 | Pathologic voice recognizing method |
| CN103956171A (en) * | 2014-04-01 | 2014-07-30 | 中国科学院软件研究所 | Multi-channel mini-mental state examination system |
| CN104900229A (en) * | 2015-05-25 | 2015-09-09 | 桂林电子科技大学信息科技学院 | Method for extracting mixed characteristic parameters of voice signals |
| CN105448291A (en) * | 2015-12-02 | 2016-03-30 | 南京邮电大学 | Parkinsonism detection method and detection system based on voice |
Cited By (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107016233A (en) * | 2017-03-14 | 2017-08-04 | 中国科学院计算技术研究所 | The association analysis method and system of motor behavior and cognitive ability |
| US10499149B2 (en) | 2017-05-05 | 2019-12-03 | Boe Technology Group Co., Ltd. | Microphone, vocal training apparatus comprising microphone and vocal analyzer, vocal training method, and non-transitory tangible computer-readable storage medium |
| CN108806720A (en) * | 2017-05-05 | 2018-11-13 | 京东方科技集团股份有限公司 | Microphone, data processor, monitoring system and monitoring method |
| CN108806720B (en) * | 2017-05-05 | 2019-12-06 | 京东方科技集团股份有限公司 | Microphone, data processor, monitoring system and monitoring method |
| CN107273677A (en) * | 2017-06-08 | 2017-10-20 | 中国科学院软件研究所 | A kind of multi-channel nerve function quantitative evaluation system |
| CN108962397A (en) * | 2018-06-06 | 2018-12-07 | 中国科学院软件研究所 | A kind of multichannel multitask the nervous system disease assistant diagnosis system based on pen and voice |
| CN109448820A (en) * | 2018-10-10 | 2019-03-08 | 上海整合医学研究院有限公司 | A kind of wearable mental disease voice assisting in diagnosis and treatment equipment |
| CN109473170A (en) * | 2018-10-26 | 2019-03-15 | 首都医科大学附属北京安定医院 | Schizoid system is diagnosed using cognition index |
| CN109480864A (en) * | 2018-10-26 | 2019-03-19 | 首都医科大学附属北京安定医院 | A kind of schizophrenia automatic evaluation system based on nervous functional defects and machine learning |
| CN109800790A (en) * | 2018-12-24 | 2019-05-24 | 厦门大学 | A kind of feature selection approach towards high dimensional data |
| CN109616141A (en) * | 2019-01-03 | 2019-04-12 | 燕山大学 | Heterophemia detection method |
| CN109616141B (en) * | 2019-01-03 | 2022-01-11 | 燕山大学 | Pronunciation abnormality detection method |
| CN111462773A (en) * | 2020-03-26 | 2020-07-28 | 心图熵动科技(苏州)有限责任公司 | Suicide risk prediction model generation method and prediction system |
| WO2021189903A1 (en) * | 2020-10-09 | 2021-09-30 | 平安科技（深圳）有限公司 | Audio-based user state identification method and apparatus, and electronic device and storage medium |
| CN113082447A (en) * | 2021-04-02 | 2021-07-09 | 电子科技大学 | Prediction method for music modulation brain plasticity effect of fMRI brain loop |
| CN113082447B (en) * | 2021-04-02 | 2021-12-07 | 电子科技大学 | Prediction method for music modulation brain plasticity effect of fMRI brain loop |
| CN115969331A (en) * | 2023-03-21 | 2023-04-18 | 中国医学科学院生物医学工程研究所 | Cognitive language disorder assessment system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106073706B (en) | 2019-08-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106073706B (en) | A kind of customized information and audio data analysis method and system towards Mini-mental Status Examination | |
| Scherer et al. | Self-reported symptoms of depression and PTSD are associated with reduced vowel space in screening interviews | |
| Benba et al. | Discriminating between patients with Parkinson’s and neurological diseases using cepstral analysis | |
| CN111461176B (en) | Multi-mode fusion method, device, medium and equipment based on normalized mutual information | |
| Laganaro et al. | Sensitivity and specificity of an acoustic-and perceptual-based tool for assessing motor speech disorders in French: The MonPaGe-screening protocol | |
| Muhammad et al. | Convergence of artificial intelligence and internet of things in smart healthcare: a case study of voice pathology detection | |
| Hartelius et al. | Long-term phonatory instability in individuals with multiple sclerosis | |
| Ziegler et al. | Gauging the auditory dimensions of dysarthric impairment: Reliability and construct validity of the Bogenhausen Dysarthria Scales (BoDyS) | |
| CN109727608A (en) | A kind of ill voice appraisal procedure based on Chinese speech | |
| Callan et al. | Self-organizing map for the classification of normal and disordered female voices | |
| CN111000556A (en) | Emotion recognition method based on deep fuzzy forest | |
| Pierce et al. | A field-based approach to establish normative acoustic data for healthy female voices | |
| CN108962397B (en) | Pen and voice-based cooperative task nervous system disease auxiliary diagnosis system | |
| Hashim et al. | Analysis of timing pattern of speech as possible indicator for near-term suicidal risk and depression in male patients | |
| Nishikawa et al. | Machine learning model for discrimination of mild dementia patients using acoustic features | |
| Gu et al. | Disordered speech assessment using automatic methods based on quantitative measures | |
| Xu et al. | A novel smart depression recognition method using human-computer interaction system | |
| Nisar et al. | Speech recognition-based automated visual acuity testing with adaptive mel filter bank | |
| Guarin et al. | Video-based facial movement analysis in the assessment of bulbar amyotrophic lateral sclerosis: clinical validation | |
| Kim et al. | " How are you?" Estimation of anxiety, sleep quality, and mood using computational voice analysis | |
| Jing et al. | Speech-language pathologists' ratings of speech accuracy in children with speech sound disorders | |
| Chiu et al. | Exploring the acoustic perceptual relationship of speech in Parkinson's disease | |
| Shabber et al. | A review and classification of amyotrophic lateral sclerosis with speech as a biomarker | |
| Mijić et al. | Classification of cognitive load using voice features: A preliminary investigation | |
| Wisler et al. | The effects of symptom onset location on automatic amyotrophic lateral sclerosis detection using the correlation structure of articulatory movements |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |