CN109949932A - A kind of judgment method of the essential tremor extent based on machine learning - Google Patents
A kind of judgment method of the essential tremor extent based on machine learning Download PDFInfo
- Publication number
- CN109949932A CN109949932A CN201910244938.0A CN201910244938A CN109949932A CN 109949932 A CN109949932 A CN 109949932A CN 201910244938 A CN201910244938 A CN 201910244938A CN 109949932 A CN109949932 A CN 109949932A
- Authority
- CN
- China
- Prior art keywords
- data
- essential tremor
- pen tip
- electronic
- machine learning
- 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.)
- Pending
Links
Landscapes
- Medical Treatment And Welfare Office Work (AREA)
Abstract
The present invention relates to computer data processing technology fields, more particularly, to a kind of judgment method of essential tremor extent based on machine learning.The regression model of discrimination essential tremor disease degree based on machine learning, the regression model completed for training, it is only necessary to which the degree that tester suffers from essential tremor can be calculated by so that tester is write some Chinese characters on electronic handwritten plate.
Description
Technical field
The present invention relates to computer data processing technology fields, more particularly, to a kind of based on the primary of machine learning
Property is trembled the judgment method of extent.
Background technique
Essential tremor (Essential Tremor, ET) is a kind of common dyskinesia, also known as essential tremor,
Benign tremor.In patient with essential tremor, there are about 60% family history, therefore also known as hereditary tremor or familial
It trembles.Multinomial this disease of studies have shown that be averaged the onset age be 45 years old, disease incidence be 0.14% to 5.10%, 70 years old or more crowd
Disease incidence is up to 12.6%.Its clinical symptoms is to tremble, and shows as posture or kinetic tremor, posture, which is trembled, to be kept
It trembles when a certain posture most obvious, kinetic tremor patient seldom trembles when static.It trembles usually since the hand of side,
And entire upper limb and opposite side upper limb are gradually diffused to, it upwards can be to head and bottleneck throat muscle.Frequency of trembling is generally 4Hz extremely
12Hz.Although essential tremor is counted as a kind of benign tremor, but the severe tremor of some patientss can interfere hand to complete finely
Movement, laryngeal muscles are lost independent living ability by for a long time will affect pronunciation what is more.Parkinson's disease (Parkinson ' s
Disease, PD) with essential tremor in certain cases have similar feature, therefore sometimes diagnosis when be difficult to determine, this
When can to patient carry out SPECT-DAT sweep test, however it is this test it is very expensive.In this context, suffered from by acquisition
The system that person's behavioural information is analyzed will become first choice, because Most patients think speech analysis, hand-written analysis or drawing
No pressure is analyzed, and the cost of these technologies is very low, the requirement to acquisition equipment is not also high.In addition, in the trouble for judging patient
In terms of course of disease degree, the index quantified there are no one only provides one by the judgement of doctor at present and fuzzy describes.
Traditional notes analysis all carries out offline, because only that the stroke on paper can be analyzed, electronic hand
Hand-written data and temporal information can be collected simultaneously by writing plate, while can also collect pressure, writing pencil and the hand of equipment surface
Write the angle of plate, it might even be possible to acquire the track that writing pencil moves in the sky.The integrated analysis of these data can be very good
Diagnosis to essential tremor case.For essential tremor, medical worker diagnoses usually using hand-written task.
Summary of the invention
For this reason, it may be necessary to provide a kind of judgment method of essential tremor extent based on machine learning, it is based on machine
The regression model of the discrimination essential tremor disease degree of study, the regression model completed for training, it is only necessary to make tester
Some Chinese characters are write on electronic handwritten plate can calculate the degree that tester suffers from essential tremor.
To achieve the above object, the present invention adopts the following technical scheme:
A kind of judgment method of the essential tremor extent based on machine learning, it includes the following steps,
Step 1: tester needs to write when selecting N number of Chinese character from Hanzi font library as assessment essential tremor disease
Test word, wherein N >=1；
Step 2: acquisition Healthy People and essential tremor patient are using electronic pen when writing test word on electronic handwritten plate
Data as sample data；
Step 3: being pre-processed to collected sample data；
Step 4: sample characteristics are organized into identical structure, so that training uses；
Step 5: regression result is extent coefficient to the training of SVR regression model；Regression model after saving training
Data；
Step 6: collecting test person write on electronic handwritten plate test word data, and to the data of acquisition according to
Step 3 and step 4 are handled, and are input to the trained SVR regression model of step 5, are obtained extent coefficient.
The technical program further optimizes, and the sample data acquired in step 2 includes electronic pen on electronic handwritten plate
Not in contact with the time of handwriting pad, electronic pen contacts the time of handwriting pad, pen tip coordinate X and Y, pen tip and water on electronic handwritten plate
The angle O of plane, the included angle A of pen tip and vertical plane, pressure of the pen tip to electronic handwritten plate.
The technical program further optimizes, and sample data pre-processes in step 3, specifically includes and checks collected number
According to whether wrong, wrong data is abandoned if having.For example it is interrupted when being write for tester, or occur when record bright
The aobvious data for not meeting common sense, directly abandon this group of data if having.Temporal characteristics are normalized, by like numbers
According to the decimal being mapped between -1 to 1, following formula can be used that data are normalized:
The included angle A of the angle O of pen tip X-axis data, Y-axis data, pen tip and horizontal plane, pen tip and vertical plane is carried out in Fu
Leaf transformation, 50hz frequency domain data before obtaining；Calculate maximum value, minimum value, average value, median and the side of the pressure data of record
Difference.
The technical program further optimizes, and step 4 is specifically, to as follows by pretreated feature structureization:
X=[Tup Tdown fX fY fO fA Fmax Fmin Fave FmidσF 2]T
Wherein TupIndicate after normalized electronic pen on electronic handwritten plate not in contact with the time of handwriting pad, TdownTable
Show that after normalized electronic pen contacts the time of handwriting pad, f on electronic handwritten plateXAnd fYIndicate pen tip X-axis and Y-axis number
According to frequency vector, fOAnd fAIndicate the angle O of pen tip and horizontal plane, the frequency vector of the included angle A of pen tip and vertical plane, Fmax,
Fmin,Fave,Fmid,σF 2Respectively indicate maximum value, minimum value, average value, median and the variance of pressure data.In addition it is also necessary to
Doctor provides the extent of patient.
The technical program further optimizes, SVR regression model in step 5 are as follows:
Establish the optimization aim of SVR regression model are as follows:
ξi ∨≥0,ξi ∧≥0
It is different from the prior art, above-mentioned technical proposal has the advantages that for traditional detection method, to tester's
It examines and needs to judge the Chinese character that it is write by experienced doctor, and the present invention is returned by collecting a sample set
Return the training of model, finally tester is predicted using regression model, does not need experienced doctor and instructed on side, simultaneously
The assessment about conditions of patients can be obtained, corresponding economic expense will also greatly reduce, and the present invention diagnoses a test
The speed of the state of an illness of person will also improve much than Artificial Diagnosis.
Detailed description of the invention
Fig. 1 is the judgment method execution flow chart of the essential tremor extent based on machine learning.
Specific embodiment
Technology contents, construction feature, the objects and the effects for detailed description technical solution, below in conjunction with specific reality
It applies example and attached drawing is cooperated to be explained in detail.
Refering to Figure 1, a kind of essential tremor extent based on machine learning of a present invention preferably embodiment
Judgment method, include the following steps:
Step 1: tester needs to write when selecting 10 Chinese characters from Hanzi font library as assessment essential tremor disease
Test word.
Step 2: Healthy People and the Chinese character of essential tremor patient writing step 1 on electronic handwritten plate and selecting are collected
When data.These data include: electronic pen on electronic handwritten plate not in contact with the time of handwriting pad, electronic pen is in electronic handwritten
The time of handwriting pad, pen tip coordinate X and Y, the angle O of pen tip and horizontal plane, the included angle A of pen tip and vertical plane, pen are contacted on plate
Pressure of the point to electronic handwritten plate.
Step 3: the data being collected into are pre-processed.Check whether collected data have apparent error, such as right
It is interrupted when tester writes, or occurs not meeting the data of common sense significantly when record, directly abandoning if having should
Group data.Temporal characteristics are normalized, decimal homogeneous data being mapped between 0 to 1.Following formula logarithm can be used
According to being normalized:
Wherein Tup/downIndicate electronic pen on electronic handwritten plate not in contact with handwriting pad or contact handwriting pad time,
Tup/down' indicate the time data after normalization.
The included angle A of the angle O of pen tip X-axis data, Y-axis data, pen tip and horizontal plane, pen tip and vertical plane is carried out in Fu
Leaf transformation, 50hz frequency domain data before obtaining.Fourier transformation be it is a kind of analyze signal method, it can analyze signal at
Point, by Data Representation in domain space, data can be transformed into domain space from time domain space according to the following formula:
For SVR regression model, if training data directly uses time-domain information, due to the concussion pole of time-domain information
Greatly, and the difference between similar sample is also very big, and the model over-fitting probability trained is very big, and training data is converted into frequency domain
After information, while data volume tails off, feature difference also becomes obvious, it is easier to be fitted.Calculate the pressure data of record
Maximum value, minimum value, average value, median and variance.In addition it is also necessary to which doctor provides the extent of patient, for making
Label when for regression training.Extent and corresponding quantized data are as follows:
| Extent | Quantized data |
| Health | 0.0 |
| Slightly | 0.25 |
| It is medium | 0.5 |
| It is more serious | 0.75 |
| Seriously | 1.0 |
Step 4: being organized into identical structure for sample characteristics, so that training uses.To by pretreated feature structure
It is as follows:
X=[Tup Tdown fX fY fO fA Fmax Fmin Fave FmidσF 2]T
Wherein TupIndicate after normalized electronic pen on electronic handwritten plate not in contact with the time of handwriting pad, TdownTable
Show that after normalized electronic pen contacts the time of handwriting pad, f on electronic handwritten plateXAnd fYIndicate pen tip X-axis and Y-axis number
According to frequency vector, fOAnd fAIndicate the angle O of pen tip and horizontal plane, the frequency vector of the included angle A of pen tip and vertical plane, Fmax,
Fmin,Fave,Fmid,σF 2Respectively indicate maximum value, minimum value, average value, median and the variance of pressure data.The feature of generation
Vector will input to SVR regression model as training sample, can learn the good SVR of effect out by these feature vectors
Regression model.
Step 5: to the training of SVR regression model, regression result is extent coefficient.Regression model after saving training
Data.SVR regression model are as follows:
Establish the optimization aim of SVR regression model are as follows:
s.t.|yi-ypredict|≤ε
The as feature vector of the sample of step 3 acquisition, yiFor the corresponding illness quantization parameter of sample.In order to improve
The generalization ability of model introduces slack variableThen SVR problem can be converted into
Following formula optimization aim:
Wherein C indicates penalty coefficient, i.e., to the tolerance of error, C is higher, and model is lower to the tolerance of error, more holds
Easily there is over-fitting situation, C is excessive or the too small generalization ability that can all lead to model is deteriorated.But due to the pact of the optimization aim
Beam condition is not convex function, therefore does further conversion:
ξi ∨≥0,ξi ∧≥0
Wherein, ξi ∧Indicate the loss above susceptibility isolation strip, ξi ∨Indicate the loss below susceptibility isolation strip.
Consider constraint condition, introduces Lagrangian α∧,α∨,β∧,β∨, dual problem is converted by optimization problem:
Former formula is brought into above formula derivation, and by result, introduces kernel functionAfter can obtain:
0≤αi ∧≤C,0≤αi ∨≤C
Process above needs to meet KKT condition, b can be derived with supporting according to the complementary slackness condition in KKT condition
The equivalence relation of vector, and then obtain the form of entire model.Complementary slackness condition is as follows:
By complementary slackness condition, (α is obtainedi ∧-αi ∨)=0, therefore supporting vector is (αi ∧-αi ∨The sample point of) ≠ 0.It asks
:
Then choose a 0 < α of satisfactioni ∨The sample of < C, and b is solved by following formula.
The Selection of kernel function radial direction base core of SVR.
SVR regression model, which trains, to be saved backup, and user's electronic hand in 10 texts that hand-written step 1 is selected is acquired
The data that plate is collected into are write, then data are pre-processed using step 3, pretreated data are input to SVR and are returned
It is returned in model, provides prediction result, extent coefficient is referred to for doctor.
It should be noted that, in this document, relational terms such as first and second and the like are used merely to a reality
Body or operation are distinguished with another entity or operation, are deposited without necessarily requiring or implying between these entities or operation
In any actual relationship or order or sequence.Moreover, the terms "include", "comprise" or its any other variant are intended to
Non-exclusive inclusion, so that the process, method, article or the terminal device that include a series of elements not only include those
Element, but also including other elements that are not explicitly listed, or further include for this process, method, article or end
The intrinsic element of end equipment.In the absence of more restrictions, being limited by sentence " including ... " or " including ... "
Element, it is not excluded that there is also other elements in process, method, article or the terminal device for including the element.This
Outside, herein, " being greater than ", " being less than ", " being more than " etc. are interpreted as not including this number；" more than ", " following ", " within " etc. understand
Being includes this number.
Although the various embodiments described above are described, once a person skilled in the art knows basic wounds
The property made concept, then additional changes and modifications can be made to these embodiments, so the above description is only an embodiment of the present invention,
It is not intended to limit scope of patent protection of the invention, it is all to utilize equivalent structure made by description of the invention and accompanying drawing content
Or equivalent process transformation, being applied directly or indirectly in other relevant technical fields, similarly includes in patent of the invention
Within protection scope.
Claims (5)
1. a kind of judgment method of the essential tremor extent based on machine learning, it is characterised in that: it includes following step
Suddenly,
Step 1: the survey that tester needs to write when selecting N number of Chinese character from Hanzi font library as assessment essential tremor disease
Word is tried, wherein N >=1；
Step 2: the number of acquisition Healthy People and essential tremor patient using electronic pen when writing test word on electronic handwritten plate
According to as sample data；
Step 3: being pre-processed to collected sample data；
Step 4: sample characteristics are organized into identical structure, so that training uses；
Step 5: regression result is extent coefficient to the training of SVR regression model；Regression model data after saving training；
Step 6: the data for the test word that collecting test person writes on electronic handwritten plate, and to the data of acquisition according to step
Three and step 4 handled, be input to the trained SVR regression model of step 5, obtain extent coefficient.
2. the judgment method of the essential tremor extent based on machine learning as described in claim 1, it is characterised in that:
The sample data acquired in the step 2 include electronic pen on electronic handwritten plate not in contact with the time of handwriting pad, electronic pen exists
Contact the time of handwriting pad on electronic handwritten plate, pen tip coordinate X and Y, the angle O of pen tip and horizontal plane, pen tip and vertical plane
Included angle A, pressure of the pen tip to electronic handwritten plate.
3. the judgment method of the essential tremor extent based on machine learning as described in claim 1, it is characterised in that:
Sample data pre-processes in the step 3, specifically includes and checks whether collected data are wrong, and mistake is abandoned if having
Data；Temporal characteristics are normalized, homogeneous data is mapped to the decimal between -1 to 1, following formula logarithm can be used
According to being normalized:
Fourier's change is carried out to the included angle A of the angle O of pen tip X-axis data, Y-axis data, pen tip and horizontal plane, pen tip and vertical plane
It changes, 50hz frequency domain data before obtaining；Calculate maximum value, minimum value, average value, median and the variance of the pressure data of record.
4. the judgment method of the essential tremor extent based on machine learning as described in claim 1, it is characterised in that:
The step 4 is specifically, to as follows by pretreated feature structureization:
X=[Tup Tdown fX fY fO fA Fmax Fmin Fave Fmid σF 2]T
Wherein TupIndicate after normalized electronic pen on electronic handwritten plate not in contact with the time of handwriting pad, TdownIndicate warp
Electronic pen contacts the time of handwriting pad, f on electronic handwritten plate after normalizedXAnd fYIndicate pen tip X-axis and Y-axis data
Frequency vector, fOAnd fAIndicate the angle O of pen tip and horizontal plane, the frequency vector of the included angle A of pen tip and vertical plane, Fmax,Fmin,
Fave,Fmid,σF 2Respectively indicate maximum value, minimum value, average value, median and the variance of pressure data.
5. the judgment method of the essential tremor extent based on machine learning as described in claim 1, it is characterised in that:
SVR regression model in the step 5 are as follows:
Establish the optimization aim of SVR regression model are as follows:
s.t.-ε-ξi ∨≤ωgxi+b-yi≤ξi ∧+ε
ξi ∨≥0,ξi ∧≥0。
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910244938.0A CN109949932A (en) | 2019-03-28 | 2019-03-28 | A kind of judgment method of the essential tremor extent based on machine learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910244938.0A CN109949932A (en) | 2019-03-28 | 2019-03-28 | A kind of judgment method of the essential tremor extent based on machine learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109949932A true CN109949932A (en) | 2019-06-28 |
Family
ID=67012377
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910244938.0A Pending CN109949932A (en) | 2019-03-28 | 2019-03-28 | A kind of judgment method of the essential tremor extent based on machine learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109949932A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110544410A (en) * | 2019-07-15 | 2019-12-06 | 山东第一医科大学(山东省医学科学院) | handwriting capability quantitative evaluation system and method |
| CN113571167A (en) * | 2021-07-28 | 2021-10-29 | 重庆橡树信息科技有限公司 | Rapid triage system based on configured scoring knowledge model |
| CN115881287A (en) * | 2023-03-03 | 2023-03-31 | 四川互慧软件有限公司 | Doctor recommendation method based on data collected by ECG (electrocardiograph) monitor |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109276255A (en) * | 2018-11-27 | 2019-01-29 | 平安科技（深圳）有限公司 | A kind of limb tremor detection method and device |
| CN109308521A (en) * | 2018-08-27 | 2019-02-05 | 广东工业大学 | A kind of quaternary SerComm degree study filtering method for eliminating Physiological tremor |
-
2019
- 2019-03-28 CN CN201910244938.0A patent/CN109949932A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109308521A (en) * | 2018-08-27 | 2019-02-05 | 广东工业大学 | A kind of quaternary SerComm degree study filtering method for eliminating Physiological tremor |
| CN109276255A (en) * | 2018-11-27 | 2019-01-29 | 平安科技（深圳）有限公司 | A kind of limb tremor detection method and device |
Non-Patent Citations (2)
| Title |
|---|
| 张子涛: "气田集气站安全监测数据异常值识别研究", 《油气田地面工程》 * |
| 李芳: "特定群体手写运动定量分析研究", 《中国博士学位论文全文数据库》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110544410A (en) * | 2019-07-15 | 2019-12-06 | 山东第一医科大学(山东省医学科学院) | handwriting capability quantitative evaluation system and method |
| CN110544410B (en) * | 2019-07-15 | 2021-06-22 | 山东第一医科大学(山东省医学科学院) | Handwriting capability quantitative evaluation system and method |
| CN113571167A (en) * | 2021-07-28 | 2021-10-29 | 重庆橡树信息科技有限公司 | Rapid triage system based on configured scoring knowledge model |
| CN113571167B (en) * | 2021-07-28 | 2024-04-19 | 重庆橡树信息科技有限公司 | Rapid triage system based on configuration type grading knowledge model |
| CN115881287A (en) * | 2023-03-03 | 2023-03-31 | 四川互慧软件有限公司 | Doctor recommendation method based on data collected by ECG (electrocardiograph) monitor |
| CN115881287B (en) * | 2023-03-03 | 2023-05-23 | 四川互慧软件有限公司 | Doctor recommendation method based on electrocardiograph monitor acquisition data |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Gaur et al. | Knowledge-aware assessment of severity of suicide risk for early intervention | |
| Chaudhari et al. | Survey on handwriting-based personality trait identification | |
| Impedovo et al. | A handwriting-based protocol for assessing neurodegenerative dementia | |
| Drotár et al. | Decision support framework for Parkinson’s disease based on novel handwriting markers | |
| CN109949932A (en) | A kind of judgment method of the essential tremor extent based on machine learning | |
| CN109949938B (en) | Method and device for standardizing medical non-standard names | |
| Gupta et al. | Indian sign language recognition using wearable sensors and multi-label classification | |
| Castrillón et al. | Characterization of the Handwriting Skills as a Biomarker for Parkinson’s Disease | |
| Bhardwaj et al. | Anveshan: a framework for analysis of multiple annotators’ labeling behavior | |
| Klerke et al. | At a glance: The impact of gaze aggregation views on syntactic tagging | |
| Alonso-Martinez et al. | A comparative study of in-air trajectories at short and long distances in online handwriting | |
| Ambica et al. | An efficient expert system for diabetes by naive Bayesian classifier | |
| Badarna et al. | The importance of pen motion pattern groups for semi-automatic classification of handwriting into mental workload classes | |
| CN115019964A (en) | Cognitive ability evaluation system and method based on digital biomarker | |
| CN111210912A (en) | Parkinson prediction method and device | |
| Kamijo et al. | Personality estimation from Japanese text | |
| Jinnuo et al. | Implementation of Artificial Intelligence Image Emotion Detection Mechanism Based on Python Architecture for Industry 4.0 | |
| CN117152514A (en) | Hand tremor recognition system based on improved MobileViT | |
| Deharab et al. | Parkinson's disease detection using dynamic writing traces warping | |
| Liu et al. | Comprehensive analysis of resting tremor based on acceleration signals of patients with Parkinson’s disease | |
| Azmi et al. | Unveiling Emotions through Handwriting: A Data Analysis Approach | |
| Zhou et al. | Predicting mental disorder from noisy questionnaires: an anomaly detection approach based on keyword extraction and machine learning techniques | |
| Xu et al. | Review of Handwriting Analysis for Predicting Personality Traits | |
| Khawar et al. | Feature relevance analysis for handwriting based identification of parkinson’s disease | |
| Xie | Rural Folk Dance Movement Recognition Based on an Improved MCM‐SVM Model in Wireless Sensing Environment |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190628 |
|
| RJ01 | Rejection of invention patent application after publication |