CN108985278A - A kind of construction method of the gait function assessment models based on svm - Google Patents
A kind of construction method of the gait function assessment models based on svm Download PDFInfo
- Publication number
- CN108985278A CN108985278A CN201810974711.7A CN201810974711A CN108985278A CN 108985278 A CN108985278 A CN 108985278A CN 201810974711 A CN201810974711 A CN 201810974711A CN 108985278 A CN108985278 A CN 108985278A
- Authority
- CN
- China
- Prior art keywords
- gait
- svm
- matrix
- assessment models
- principal component
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/20—Movements or behaviour, e.g. gesture recognition
- G06V40/23—Recognition of whole body movements, e.g. for sport training
- G06V40/25—Recognition of walking or running movements, e.g. gait recognition
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
- G06F18/2135—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on approximation criteria, e.g. principal component analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Psychiatry (AREA)
- Social Psychology (AREA)
- Human Computer Interaction (AREA)
- Multimedia (AREA)
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
Abstract
The construction method for the gait function assessment models based on svm that the invention proposes a kind of, comprising the following steps: obtain the data group of health adult and the gait feature with dyskinesia Disease;Collected gait feature data are formed into original matrix X, and determine the correlation matrix R of original matrix X;Find out the characteristic root λ j of the characteristic equation det (R- λ E)=0 of R;It determines principal component number m, then determines m corresponding unit character vectors;Principal component matrix is determined according to unit character vector and original matrix;Gait function assessment models are determined according to principal component matrix.The present invention is by converting initial data, gait function assessment models are established according to transformed principal component matrix, eliminate redundancy little parameter is influenced on prediction result and maintains the information of original gait feature data to the maximum extent, so that the model established is more acurrate, so as to it is scientific, objective, accurate, intuitively gait function is assessed.
Description
Technical field
The present invention relates to exercise data processing and assessment technology fields, and in particular to a kind of gait function based on svm is commented
Estimate the construction method of model.
Background technique
Gait is the posture of human body walking, and which includes the movement of trunk, upper limb and lower limb and cooperations.In work
Journey field, the functional assessment for gait are an important contents in gait analysis, and clinically, it can help clinical doctor
It is raw to understand patient with the presence or absence of balance dysfunction, the reason of causing dysfunction is found out, judges whether treatment means are effective;?
In daily life, the research that people assess gait function can evade the risk fallen down, and bring life threat is fallen down in reduction
And somatic damage.In addition, gait function assessment can reflect motor function it is sound whether, for Healthy People, especially old man and
The life of children and the patient of dyskinesia and safety are extremely important.
Currently, what is used in gait function appraisal procedure is usually conventional parameter (such as evaluation of single lower limb motor function
It is middle to use phase percentage, joint angles, joint angle speed, acceleration etc.) and muscular features (such as contraction of muscle degree and its power
Learn the more microcosmic muscular movement feature such as characterisitic parameter) both deliberated indexes and third class deliberated index be (such as characterization walking
The symmetrical index of the step-length of symmetric property, the symmetrical index of ground reaction force characterize pair of the lower limb movement executive condition based on fitts law
Claim index etc.), this comparison by single parameter can not be objective, accurately anti-to carry out the method for overall assessment to individual
Answer the real information of person to be tested;In addition, this appraisal procedure relied on mostly in terms of gait function analysis the experience of doctor with
And the feeling of patient itself is assessed, and lacks scientific, being as a result possible to can be serious unfounded.
Summary of the invention
In view of the deficiencies of the prior art, the present invention proposes a kind of construction method of gait function assessment models based on svm,
The too strong disadvantage of subjectivity when judging to assess gait function with personal experience in the prior art is overcome, is also avoided existing
It can not objective, accurately react person's to be tested when having in technology only by the comparison of single parameter to individual progress net assessment
The problem of real information.
The technical scheme of the present invention is realized as follows: a kind of construction method of the gait function assessment models based on svm,
The following steps are included:
Step 10, the data group of health adult and the gait feature with dyskinesia Disease is obtained, it is described
Gait feature include the period, stride, leg speed, cadence, left step-length, right step-length, left support phase, right support phase, counter-force, surface flesh
It is power, sufficient drift angle, hip joint buckling value, hip joint stretching, extension value, knee sprung value, knee extension value, more in joint moment
Kind is whole;
Step 20, collected gait feature data are formed into original matrixAnd determination is original
The correlation matrix R of matrix X, wherein x1j~xijIndicate the data group of one of gait feature, j=1,2,3 ... j;xi1
~xijIndicate all gait feature data of a wherein people, i=1,2 ..., i;
Step 30, the characteristic root λ of the characteristic equation det (R- λ E)=0 of R is found outj, wherein λ1≥λ2≥λ3…≥λj≥0;
Step 40, it determines principal component number m, then determines m corresponding unit character vector βm, wherein m=1,2 ...,
m;
Step 50, according to unit character vector βmAnd original matrix determines principal component matrix Z;
Step 60, determine that svm sorter model, the svm sorter model are commented for gait function according to principal component matrix Z
Estimate model.
Optionally, in step 30, the expression formula of correlation matrix R are as follows:
R=(rij)i×j, whereinrij=rji,
rii=1.
Optionally, in step 40, the number m of the principal component is determined by following formula:
Wherein α is 60%~80%.
Optionally, in step 40, βmExpression formula are as follows:
Wherein
Optionally, in step 50, the expression formula of the principal component matrix are as follows:Wherein
Zj=β1X1+β2X2+…+βmXj,
X1、X2、…Xj, respectively indicate by each gait feature
Data group formed matrix.
Optionally, using the data in principal component matrix as new samples data, new samples data are averagely divided into S one's share of expenses for a joint undertaking
Sample, S-1 one's share of expenses for a joint undertaking sample therein is as training set, remaining is as checksum set;S kind training set by being likely to form respectively
It determines the model of corresponding svm classifier, and corresponding model is verified respectively by corresponding checksum set, calculate corresponding model
Accuracy rate, using the highest model of wherein accuracy rate as gait function assessment models.
Optionally, the model of the svm classifier is f (z)=sign (w*z+b*), wherein optimal normal vector w*With it is optimal
Intercept b*Value according to the condition of optimal interface determine.
Optionally, the condition of the optimal interface are as follows:
s.t.yi(w·zi+b)≥1-ξi, i=1,2 ..., N
ξi>=0, i=1,2 ..., N
Wherein, w is normal vector, and b is intercept, and C is penalty term parameter, ξiFor the slack variable of corresponding sample point, yiBased on
The class label of component matrix the i-th data group.
Optionally, the optimal interface is solved by building Lagrangian, constructed Lagrangian are as follows:
αi≥0,μi≥0。
Wherein, αi, μiFor Lagrange multiplier.
Compared with prior art, the invention has the following advantages that the principal component matrix that the present invention is determined according to initial data,
Eliminate redundancy influences little parameter on prediction result, maintains the information of original gait feature data to the maximum extent,
Svm sorter model is finally established as gait function assessment models according to principal component matrix, accuracy is higher, and more objective,
Science can provide reference when whether the gait function for diagnosing person to be tested is abnormal for doctor or patient.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only this
Some embodiments of invention without any creative labor, may be used also for those of ordinary skill in the art
To obtain other drawings based on these drawings.
Fig. 1 is a kind of flow chart of the construction method embodiment one of the gait function assessment models based on svm of the present invention;
Fig. 2 is a kind of flow chart of the construction method embodiment two of the gait function assessment models based on svm of the present invention.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts every other
Embodiment shall fall within the protection scope of the present invention.
Fig. 1 is a kind of flow chart of the construction method embodiment one of the gait function assessment models based on svm of the present invention, ginseng
Read Fig. 1, the construction method of the gait function assessment models based on svm disclosed in the present embodiment one, comprising the following steps:
Step 101, the data group of health adult and the gait feature with dyskinesia Disease is obtained, it is described
Gait feature include the period, stride, leg speed, cadence, left step-length, right step-length, left support phase, right support phase, counter-force, surface flesh
It is power, sufficient drift angle, hip joint buckling value, hip joint stretching, extension value, knee sprung value, knee extension value, more in joint moment
Kind is whole;
Wherein, instrument being captured by three-dimensional gait, data are acquired to corresponding gait feature.In addition, choosing sample
During, the age of sample should be evenly distributed between 20~70, and male to female ratio maintains 1:1 as far as possible, tested in sample
Person's number should be counted not less than 50 people.Moreover, during acquiring the data group of gait feature, acquisition 10 times every time, Cong Zhongxuan
It is best primary to select acquisition, to improve the accuracy of assessment models.
Further, the number of samples taken should be not less than 1500, and the number of samples the high, the model established more
Accurately.Moreover, the accounting with dyskinesia Disease should be not less than 20%, the accounting of health adult should be not less than
20%, avoid the positive and negative uneven accuracy rate for influencing model of sample.
Step 102, collected gait feature data are formed into original matrixAnd determination is original
The correlation matrix R of matrix X;
Wherein, x1j~xijIndicate the data group of one of gait feature, j=1,2,3 ... j;For example, x11~xi1It indicates
The data group of this gait feature of period, x12~xi2Indicate the data group of this gait feature of stride, x13~xi3Indicate leg speed this
The data group ... ... of one gait feature, and so on.
xi1~xijIndicate all gait feature data of a wherein people, i=1,2 ..., i;For example, x11~x1jIndicate health
All gait feature data, including period, stride, leg speed, cadence, left step-length, right step-length of adult first etc.;x21~x2jIt indicates
All gait feature data of health adult's second also include period, stride, leg speed, cadence, left step-length, right step-length etc.;x21~
x2jIndicate suffer from dyskinesia Disease third all gait feature data, also include the period, stride, leg speed, cadence,
Left step-length, right step-length etc.;..., and so on.
In the embodiment of the present invention, the expression formula of correlation matrix R are as follows:
R=(rij)i×j, whereinrij=rji,
rii=1.
Step 103, the characteristic root λ of the characteristic equation det (R- λ E)=0 of R is found outj, wherein λ1≥λ2≥λ3…≥λj≥0;
In the embodiment of the present invention, the characteristic root found out is arranged from big to small, is respectively defined as λ1, λ2, λ3..., λj, use
Principal component number m is determined according to each characteristic root later.
Step 104, it determines principal component number m, then determines m corresponding unit character vector β1、β2、…、βm, wherein m
=1,2 ..., m;
It, can basis in the embodiment of the present inventionDetermine the principal component number m of α, wherein α is 60%~80%.
In addition, unit character vector βmForWherein i=1,2 ..., i, β1m、β2m、…βimFor unit feature
Vector βmComponent.Such as
WhereinSuch as And so on.ximFor the corresponding numerical value in original matrix X.
Step 105, according to unit character vector βmAnd original matrix determines principal component matrix Z;
Principal component matrix in the embodiment of the present invention is by submatrix Z1、Z2、…ZjIt is formed, i.e.,Its
In:
Zj=β1X1+β2X2+…+βmXj, X1、X2、…XjFor the submatrix of original matrix Z, i.e., respectively by each gait feature
Data group formed matrix.
In addition, in principal component matrix Z, Zi1~ZijClassification and original matrix in x1j~xijClassification it is corresponding.
Step, 106, svm sorter model is determined according to principal component matrix Z;
In the embodiment of the present invention, the data in principal component matrix are determined into svm sorter model as new samples data.
It specifically includes following sub-step:
Step 1061, new samples data are averagely divided into S one's share of expenses for a joint undertaking sample first, S-1 one's share of expenses for a joint undertaking sample therein is as instruction
Practice collection, remaining determines the model of svm classifier by training set as checksum set, and carries out by checksum set to the model
Verifying, calculates the accuracy rate of the model;
Step 1062, it would be possible to which the S kind sorter model of appearance is trained according to step 402, with accurate in S assessment
The highest model of rate is as gait function assessment models.In the embodiment of the present invention, the model of svm classifier is f (x)=sign
(w*z+b*), optimal normal vector w can determine according to the condition of optimal interface*With optimal intercept b*Value.
Further, since all data are not so clean, it is impossible to 100% linear separability, therefore the present invention
Embodiment introduces slack variable ξ in optimal interface conditioni, some data points is allowed to may be at the one of mistake of divisional plane
Side, convenient for finding out optimal normal vector w*With optimal intercept b*Value, then the condition of optimal interface at this time are as follows:
s.t.yi(w·zi+b)≥1-ξi, i=1,2 ..., N
ξi>=0, i=1,2 ..., N
Wherein, w is normal vector, and b is intercept, and constant C is penalty term parameter, ξiFor the slack variable of corresponding sample, ziTable
Show the i-th data group in new samples data, yiFor the class label of the i-th data group in new samples data.It is drawn for this purpose, introducing
Ge Lang multiplier αi, optimal interface, therefore the Lagrange constructed according to the condition of optimal interface are solved with conditional extremum
Function are as follows:
In the Lagrangian, μiIt also is Lagrange multiplier, αi≥0,μi≥0。
Wherein, the dual problem of the condition at interface is optimized are as follows:
C-αi-μi=0
αi≥0
μi>=0, i=1,2 ..., N
Since the dual problem is convex quadratic programming problem, therefore solution meets KKT condition, therefore can obtain:
IfFor the solution of dual problem, then optimal normal vector w can be obtained by above-mentioned equation*Most
Excellent intercept b*Value are as follows:
Therefore, separating hyperplance also can be obtained are as follows:Then categorised decision function is
Due in dual problem, either objective function or decision function all relate only to input example and example it
Between inner product, the inner product (z in the objective function of dual problemi·zj) kernel function K (z can be usedi, zj)=(zi·zj) carry out generation
It replaces, at this time the objective function of dual problem are as follows:
0≤αi≤ C, i=1,2 ..., N
At this point, the model of svm classifier can convert are as follows:As long as therefore really
Determine Lagrange multiplierIt can determine the model of svm classifier.
Due to Lagrange multiplier αiMeet KKT condition, therefore sets α1、α2For variable, α3, α4,…,αNBe it is quantitative, then by
Constraint equation in objective functionKnown toWork as α2It can determine, then know α1Also therewith really
It is fixed.Therefore, the objective function of dual problem can be converted are as follows:
0≤αi≤ C, i=1,2
Wherein Kij=zi·zj, such as K11=z1·z2, K12=z1·z2... etc., if the initial solution of above-mentioned equation isOptimal solution isAnd assume constraint to it is not cropped when α2Optimal solution be
Due toNeed to meet 0≤α of inequality constraints conditioni≤ C, i=1,2, soThe value model of optimal value
It encloses are as follows:So that
If y1=y2, then have
If y1≠y2, then have
Updating α1, α2During, following equation should be met:
Wherein,
η=K11+K22—2K12
Therefore, by constantly adjustingIt is set to meet above-mentioned equation, even if αiMeet KKT condition, finally asks
α out1,α2Optimal solution, can similarly find out α3, α4,…,αNOptimal solution, thus finally determine Lagrange multiplierMost
Excellent solution.
For the present invention by converting to initial data, eliminate redundancy influences little parameter on prediction result, becomes
It changes and only needs a small amount of new variables that can explain most of variation in initial data afterwards, to improve the accuracy of model.This hair
It is bright by choosing really to valuable information attribute is predicted, exclude interfering with each other between multiple correlation attribute, make prediction result
It is more acurrate.
Fig. 2 is a kind of flow chart of the construction method embodiment two of the gait function assessment models based on svm of the present invention, ginseng
Read Fig. 2, the construction method of the gait function assessment models based on svm disclosed in the present embodiment two, comprising the following steps:
Step 201, the data group of health adult and the gait feature with dyskinesia Disease is obtained;Its
In, the gait feature include the period, stride, leg speed, cadence, left step-length, right step-length, left support phase, right support phase, it is anti-
Power, surface muscular strength, sufficient drift angle, hip joint buckling value, hip joint stretching, extension value, knee sprung value, knee extension value, joint power
It is a variety of or whole in square;
For the collected sample of institute in table 1:
Table 1
Gait feature collected is respectively period, stride, leg speed, cadence, left step-length, right step-length, left branch in the sample
Support phase and right support phase, if X1,X2,X3,X4,X5,X6,X7,X8Respectively indicate period, stride, leg speed, cadence, left step-length, right step
Length, left support phase, the gait feature data group of right support phase, xijA data are arranged for the i-th row jth in upper table.Such as the step of serial number 3
Speed is 112, i.e. x33It is expressed as=112.
Step 202, the data group of the gait feature is pre-processed, the pretreatment includes filling up for missing data
With the rejecting of abnormal data;
When there are null value, average value can be used and fill up.In table 1, x57For null value, average value replacement can be used, i.e.,
It when there are abnormal data, is rejected, i.e., is found in record and the sample data using nearest neighbor interpolation method
The attribute value of the sample of best shortcut carries out interpolation.In table 1, the period is all in 1s or so, and the period of serial number 4 is 2.2,
In the case where the parameter of the parameters such as leg speed, cadence and other serial numbers is roughly the same, so big difference should not be will appear, pushed away
It surveys estimation to be likely to be measurement process or say the error of generation in recording process, for the abnormal data, be sent out by analysis
Existing, each parameter of serial number 3 and each parameter of serial number 4 are almost the same, then can be replaced with the cycle value 1.04 of serial number 3
The cycle value of serial number 4.
Further, the step 202 in the embodiment of the present invention is the more feelings of data volume and/or exceptional value in missing
It is used under condition, if the data volume and/or exceptional value of missing are smaller, step 202 can not be used.
Step 203, collected gait feature data are formed into original matrixAnd determination is original
The correlation matrix R of matrix X, wherein x1j~xijIndicate the data group of one of gait feature, j=1,2,3 ... j;xi1
~xijIndicate all gait feature data of a wherein people, i=1,2 ..., i;
In table 1, matrix can be obtainedX in the matrix1j~xij
It is collected corresponding gait feature data group, x respectively11~xi1Value be respectively 1.2,1.02,1.04,2.2,1.1, for this
The data group in period, such as x in sample12~xi2Value be respectively 113,137,117,118,115, be this sample in stride number
According to group ... ... etc..xi1~xijFor all gait feature data of a wherein people, such as x11~x1jCollected first of institute in table 1
The gait feature data of people, respectively 1.2,113,109,100,57,61,59,59, and the artificial health adult;Such as x31~x3j
In table 1 collected third individual gait feature data, respectively 1.04,117,112,115,62,65,60,60, and should
Artificially suffer from dyskinesia Disease.
Step 204, the characteristic root λ of the characteristic equation det (R- λ E)=0 of R is found outj, wherein λ1≥λ2≥λ3…≥λj≥0;
In the embodiment of the present invention, characteristic root λjDetermination be the same as example 1.
Step 205, it determines principal component number m, then determines m corresponding unit character vector βm, wherein m=1,
2,…,m;
It, can basis in the embodiment of the present inventionDetermine the principal component number m of α, wherein α is 60%~80%.
In addition, unit character vector βmForWherein i=1,2 ..., i, β1m、β2m、…βimFor unit feature
Vector βmComponent.Such as
WhereinSuch as And so on.ximFor the corresponding numerical value in original matrix X.
Step 206, according to unit character vector βmAnd original matrix determines principal component matrix Z;
Principal component matrix in the embodiment of the present invention is by submatrix Z1、Z2、…ZjIt is formed, i.e.,Its
In:
Zj=β1X1+β2X2+…+βmXj, X1、X2、…XjFor the submatrix of original matrix Z, i.e., respectively by each gait feature
Data group formed matrix, in table 1,
Also, in principal component matrix Z, Zi1~ZijClassification and original matrix in x1j~xijClassification it is corresponding, such as table 1
In, x11~x1jThe classification of corresponding people is health adult, then Z11~Z1jCorresponding classification is also health adult, such as such
It pushes away.
Step 207, determine that svm sorter model, the svm sorter model are gait function according to principal component matrix Z
Assessment models.
In the embodiment of the present invention, using the data in principal component matrix as new samples data, new samples data are averagely drawn
It is divided into S one's share of expenses for a joint undertaking sample, S-1 one's share of expenses for a joint undertaking sample therein is as training set, remaining is as checksum set;Respectively by being likely to form
S kind training set determines the model of corresponding svm classifier, and is verified respectively to corresponding model by corresponding checksum set, calculates
The accuracy rate of corresponding model out, using the highest model of wherein accuracy rate as gait function assessment models.
For example, S-1 one's share of expenses for a joint undertaking sample is as training set and 1 part of checksum set, it is possible to will appear the model of S kind svm classifier.
Such as work as S=7, i.e., the data of the gait feature are equally divided into 7 small samples, number is respectively 1,2,3,4,5,6,7, is taken wherein
6 parts be used as training set, 1 part be used as checksum set, then can have 7 kinds of combinations.For example, the small sample of number 1 is as checksum set,
The small sample that remaining number is 2,3,4,5,6 is as training set;The small sample of number 2 as checksum set, remaining number is 1,3,
4,5,6 small sample is as training set, and so on, be then for the 7th time using the small sample of number 7 as checksum set, number 1,
2,3,4,5,6 small sample is as training set.Therefore, the present invention is trained and verifies to this seven kinds of combinations respectively, choosing
Take wherein the highest sorter model of accuracy rate can effectively improve the accuracy of assessment models as gait function assessment models.
In the embodiment of the present invention, model f (x)=sign (w of svm classifier*z+b*) determination process and one phase of embodiment
Together, so that
Wherein,For Lagrange multiplier, yiFor the class label of the i-th data in new samples data, ziFor new samples
The i-th data in data, zjFor the i-th column data in new samples data.By taking table 2, table 3 as an example, z1Represent the number of serial number 1
According to (1.11,39,35,118,74,82,58,60), y at this time1Represent the class label 0 (i.e. not illness) of serial number 1.
Lagrange multiplierDetermination process be the same as example 1, just no longer repeat one by one here.
After gait function assessment models based on svm establish, gait function assessment is carried out, helps clinician to understand and suffers from
Person whether there is balance dysfunction, provide reference for doctor.
The present invention establishes gait function assessment mould by converting to initial data, according to transformed principal component matrix
Type eliminates influencing little parameter to prediction result and maintaining the letter of original gait feature data to the maximum extent for redundancy
Breath avoids the problem of personal subjectivity judgement is with single parametric judgement so that the model accuracy established is higher, objective
And science.
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the invention, all in essence of the invention
Within mind and principle, any modification, equivalent replacement, improvement and so on be should all be included in the protection scope of the present invention.
Claims (9)
1. a kind of construction method of the gait function assessment models based on svm, which comprises the following steps:
Step 10, the data group of health adult and the gait feature with dyskinesia Disease, the gait are obtained
Feature include the period, stride, leg speed, cadence, left step-length, right step-length, left support phase, right support phase, counter-force, surface muscular strength,
Sufficient drift angle, hip joint buckling value, hip joint stretching, extension value, knee sprung value, knee extension value, a variety of in joint moment or
All;
Step 20, collected gait feature data are formed into original matrixAnd determine original matrix X
Correlation matrix R, wherein x1j~xijIndicate the data group of one of gait feature, j=1,2,3 ... j;xi1~xijTable
Show all gait feature data of a wherein people, i=1,2 ..., i;
Step 30, the characteristic root λ of the characteristic equation det (R- λ E)=0 of R is found outj, wherein λ1≥λ2≥λ3…≥λj≥0;
Step 40, it determines principal component number m, then determines m corresponding unit character vector βm, wherein m=1,2 ..., m;
Step 50, according to unit character vector βmAnd original matrix determines principal component matrix Z;
Step 60, determine that svm sorter model, the svm sorter model are that gait function assesses mould according to principal component matrix Z
Type.
2. the construction method of the gait function assessment models based on svm as described in claim 1, which is characterized in that in step 30,
The expression formula of correlation matrix R are as follows:
R=(rij)i×j, whereinrij=rji,rii=1.
3. the construction method of the gait function assessment models based on svm as described in claim 1, which is characterized in that in step 40,
The number m of the principal component is determined by following formula:
Wherein α is 60%~80%.
4. the construction method of the gait function assessment models based on svm as described in claim 1, which is characterized in that in step 40,
βmExpression formula are as follows:
Wherein
5. the construction method of the gait function assessment models based on svm as claimed in claim 4, which is characterized in that in step 50,
The expression formula of the principal component matrix are as follows:Wherein
Zj=β1X1+β2X2+…+βmXj,
X1、X2、…Xj, respectively indicate by the number of each gait feature
According to group matrix formed.
6. the construction method of the gait function assessment models based on svm as described in claim 1, which is characterized in that by principal component
New samples data are averagely divided into S one's share of expenses for a joint undertaking sample, S-1 one's share of expenses for a joint undertaking sample therein as new samples data by the data in matrix
As training set, remaining is as checksum set;The mould of corresponding SVM classifier is determined by the S kind training set being likely to form respectively
Type, and corresponding model is verified respectively by corresponding checksum set, the accuracy rate of corresponding model is calculated, with wherein accuracy rate
Highest model is as gait function assessment models.
7. the construction method of the gait function assessment models as described in claim 1~6 based on svm, which is characterized in that described
The model of SVM classifier is f (z)=sign (w*z+b*), wherein optimal normal vector w*With optimal intercept b*Value according to most optimal sorting
The condition at interface determines.
8. the construction method of the gait function assessment models based on svm as claimed in claim 7, which is characterized in that described optimal
The condition of interface are as follows:
s.t.yi(w·zi+b)≥1-ξi, i=1,2 ..., N
ξi>=0, i=1,2 ..., N
Wherein, w is normal vector, and b is intercept, and C is penalty term parameter, ξiFor the slack variable of corresponding sample point, yiFor principal component
The class label of matrix the i-th data group.
9. the construction method of the gait function assessment models based on svm as claimed in claim 8, which is characterized in that described optimal
Interface is solved by building Lagrangian, constructed Lagrangian are as follows:
αi≥0,μi≥0。
Wherein, αi, μiFor Lagrange multiplier.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810974711.7A CN108985278A (en) | 2018-08-24 | 2018-08-24 | A kind of construction method of the gait function assessment models based on svm |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810974711.7A CN108985278A (en) | 2018-08-24 | 2018-08-24 | A kind of construction method of the gait function assessment models based on svm |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108985278A true CN108985278A (en) | 2018-12-11 |
Family
ID=64546729
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810974711.7A Pending CN108985278A (en) | 2018-08-24 | 2018-08-24 | A kind of construction method of the gait function assessment models based on svm |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108985278A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110893100A (en) * | 2019-12-16 | 2020-03-20 | 广东轻工职业技术学院 | Device and method for monitoring posture change based on plantar pressure sensor |
CN114224325A (en) * | 2021-10-11 | 2022-03-25 | 南京理工大学 | Gait analysis system and method for calculating joint moment and angle by using inertial sensor |
CN117133402A (en) * | 2023-08-30 | 2023-11-28 | 广东省中医院(广州中医药大学第二附属医院、广州中医药大学第二临床医学院、广东省中医药科学院) | Method, device, equipment and readable storage medium for dynamically supervising patient rehabilitation |
-
2018
- 2018-08-24 CN CN201810974711.7A patent/CN108985278A/en active Pending
Non-Patent Citations (3)
Title |
---|
姚志明: "基于步态触觉信息的身份识别研究", 《中国博士学位论文全文数据库信息科技辑》 * |
张书涛: "用于功能评估的步态运动测量与特征提取研究", 《中国博士学位论文全文数据库医药卫生科技辑》 * |
蒋三庚: "《北京商务中心区(CBD)发展指数研究 北京市哲学社会科学CBD发展研究基地2015年度报告》", 31 October 2016 * |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110893100A (en) * | 2019-12-16 | 2020-03-20 | 广东轻工职业技术学院 | Device and method for monitoring posture change based on plantar pressure sensor |
CN114224325A (en) * | 2021-10-11 | 2022-03-25 | 南京理工大学 | Gait analysis system and method for calculating joint moment and angle by using inertial sensor |
CN117133402A (en) * | 2023-08-30 | 2023-11-28 | 广东省中医院(广州中医药大学第二附属医院、广州中医药大学第二临床医学院、广东省中医药科学院) | Method, device, equipment and readable storage medium for dynamically supervising patient rehabilitation |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107753026B (en) | Intelligent shoe self-adaptive monitoring method for spinal leg health | |
CN107194433B (en) | A kind of Radar range profile's target identification method based on depth autoencoder network | |
CN104997523A (en) | Upper limb rehabilitation robot rehabilitation training motion function assessment method | |
CN108985278A (en) | A kind of construction method of the gait function assessment models based on svm | |
CN109544518A (en) | A kind of method and its system applied to the assessment of skeletal maturation degree | |
CN108681685A (en) | A kind of body work intension recognizing method based on human body surface myoelectric signal | |
CN114822763A (en) | Personalized exercise prescription recommendation method driven by exercise data | |
CN110570946A (en) | Lower limb rehabilitation robot rehabilitation training motor function rehabilitation evaluation method | |
Heazlewood et al. | A comparison of classification accuracy for gender using neural networks multilayer perceptron (MLP), radial basis function (RBF) procedures compared to discriminant function analysis and logistic regression based on nine sports psychological constructs to measure motivations to participate in masters sports competing at the 2009 world masters games | |
CN109009148A (en) | A kind of gait function appraisal procedure | |
CN114512239A (en) | Cerebral apoplexy risk prediction method and system based on transfer learning | |
Rein | Measurement methods to analyze changes in coordination during motor learning from a non-linear perspective | |
CN103249353A (en) | Method for determining physical constitutions using integrated information | |
CN116597940A (en) | Modeling method of movement disorder symptom quantitative evaluation model | |
CN109934179B (en) | Human body action recognition method based on automatic feature selection and integrated learning algorithm | |
CN111916179A (en) | Method for carrying out 'customized' diet nourishing model based on artificial intelligence self-adaption individual physical sign | |
Bajpai et al. | AbnormNet: A neural network based suggestive tool for identifying gait abnormalities in cerebral palsy children | |
CN115147768B (en) | Fall risk assessment method and system | |
CN116502066A (en) | Exoskeleton swing period prediction system and method based on BP neural network | |
Mack et al. | Movement prototypes and their relationship in the performance of a gymnastics floor routine | |
CN108877931A (en) | Shoulder rehabilitation evaluation method, apparatus and system | |
CN110123317A (en) | Merge the knee joint contact force calculation method of artificial fish-swarm and random forests algorithm | |
CN115281662A (en) | Intelligent auxiliary diagnosis system for instable chronic ankle joints | |
Hou et al. | Prediction of EMG signals of trunk muscles in manual lifting using a neural network model | |
CN115089147A (en) | Blood pressure measuring device |
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: 20181211 |
|
RJ01 | Rejection of invention patent application after publication |