CN102488963A - Functional electrical stimulation knee joint angle control method - Google Patents

Abstract

The invention discloses a functional electrical stimulation knee joint angle control method, which comprises the following steps: acquiring a first knee joint angle and an ultraphonic muscle image of a target muscle under functional electrical stimulation; frequency-domain filtering the ultraphonic muscle image of the target muscle, and extracting a thickness parameter of the target muscle by utilizing a crosscorrelation method; extracting a texture parameter of the target muscle through a gray level co-occurrence matrix method; obtaining a second knee joint angle according to stimulus intensity, the thickness parameter of the target muscle and the texture parameter of the target muscle; obtaining a relative root-mean-square error, a standard root-mean-square error and a relative coefficient according to the second knee joint angle and the first knee joint angle; and obtaining a compound mode of the stimulus intensity, the thickness parameter of the target muscle and the texture parameter of the target muscle according to the relative root-mean-square error, the standard root-mean-square error and the relative coefficient. According to the method, the accuracy of control signals is high, and the input and output parameters can be adjusted according to the functional status of the current target muscle.

Description

A kind of functional electric stimulation knee joint angle control method

Technical field

The present invention relates to the disability rehabilitation and treat the control field of instructing, particularly a kind of functional electric stimulation knee joint angle control method.

Background technology

In recent years, the sickness rate that cerebrovascular such as spinal cord injury and apoplexy cause paralysing is remarkable ascendant trend, all brings bigger burden not only for individual and family, also becomes heavy day by day social problem.The Executive Meeting of the State Council that China in 2011 holds is pointed out, strives 2015, China people with disability is lived totally attain the well-off standard, and participates in and state of development is significantly improved, and tentatively realizes people with disability's " everybody enjoys rehabilitation service " target.

It is emphasis and the difficult point of paying close attention to when paralytic patient is carried out rehabilitation that limb function is rebuild, and this is related to the raising problem of its daily life active ability and quality of life.Aspect the recovery of paralytic patient locomotor activity, FES (Functional electrical stimulation, functional electric stimulation) is generally believed it is a kind of more effective clinical tool at present.FES utilizes the excitable muscle of certain particular current (voltage) signal stimulus, tissue or organ, to improve its muscle performance, recovery or to rebuild the technology of the limb activity function of being lost by nerve injury.The sixties in 20th century, Liberson successfully utilizes the electricity irritation peroneal nerve to correct the gait of hemiplegic patient's drop foot first, has started the new way that functional electric stimulation is used to move and Sensory rehabilitation is treated.

In FES; Utilize neurocyte the response of electricity irritation to be transmitted the Artificial Control signal that adds; Effect through extrinsic current; Neurocyte can produce one with excite the similar neural impulse of action potential that causes naturally, the meat fiber of its domination produce to be shunk, thereby obtains the effect of motion.Although deepen continuously along with what understand; FES has been applied to many fields of rehabilitation; But compare with its wide application prospect, a lot of new FES technology also are confined to laboratory stage, and the FES stimulus modelity of clinical practice and the effect that reaches are all very limited.FES wants to give full play to its effect at rehabilitation field, just must set up the perfect FES system of a cover.

Traditional lower limb function electric stimulation can be divided into according to control signal: limbs control formula functional electric stimulation system, bio electricity control formula functional electric stimulation system:

(a) limbs control formula functional electric stimulation system

Limbs control formula FES system can be divided into again: foot-control type FES system, manual mode FES system etc.

Foot-control type is the control mode of functional electrical stimulato the earliest, and major advantage is simple and convenient, easy operating.But it has many limitation, as: receive environmental limitations such as light foot use; The triggering functional lability of sole switch; The scope of application is little, and only be applicable to slight paralysis, remain the patient of part lower limb function, be not ideal rehabilitation scheme for the patient of severe lower extremity paralysis.

Manual mode functional electric stimulation advantage is that other control modes are easy to operate relatively; Than the foot-control type functional electric stimulation system, more initiative is arranged in patient's use, accepted by the patient easily.Its limitation is that the patient need concentrate one's energy to operate in the use, and maloperation might take place, and causes the patient in use to fall down the generation of phenomenon because of losing balance, thereby causes patient's secondary damage.

Limbs formula control utilizes action or the information Control functional electric stimulation of remaining function at other positions of health in addition except that foot-control type and manual mode, stretch in the wrong, head rotation, breathing and voice etc. like shoulder.At present, these control modes still are in laboratory stage, as yet not working control lower limb walking; The control source of these modes all is indirect with the control relation between objects simultaneously, and learning training is relatively complicated, is unfavorable for the convenient, flexible application of system.

(b) bio electricity control formula FES system

The FES of electromyographic signal control commonly system of bio electricity control formula FES system, electroneurographic signal control FES system and EEG signals control FES system etc.

Electromyographic signal control FES system utilizes myoelectric information as control signal, controls corresponding position with the movable function that recovers and reconstruction is lost by spinal cord injury.But in practice; Electromyographic signal possibly fallen into oblivion by FES system stimulating current signal; Electromyographic signal also possibly receive noise influences such as power frequency, electrocardio, motion artifacts simultaneously; And muscle fatigue is a complicated problems to the influence of musculation ability under the FES, causes electromyographic signal that functional stimulating system control ability is descended, and this control mode still is in laboratory stage.

Electroneurographic signal control FES system utilizes the neural snap ring electrode that is looped around around the target nerve to gather electroneurographic signal control functional electrical stimulato to stimulate the related muscles piece to make it produce corresponding motion.But because higher to the material requirements of neural snap ring electrode, and the extraction rate of electroneurographic signal is slower; Moreover this control mode need belong to the control mode that wound is arranged with in the electrode implant into body, so should technology still be in the primary research stage.

EEG signals control FES system utilizes the current control mode of certain characteristic control FES system of EEG signals.But EEG signals control FES system also faces very big difficulty, and is slower like brain-computer interface speed; And EEG signals are also had a defective as control signal is that the pattern of brain electricity is simple; The brain electricity is more weak, gathers and handles and all compare difficulty, seldom can walk out laboratory so far.

Shortcoming and defect below the inventor finds to exist at least in the prior art in realizing process of the present invention:

The shortcoming separately of traditional limbs control formula and bio electricity control formula FES system seriously limits the extensive use of FES in the paralytic patient rehabilitation training; Sum up; Main limiting factor is exactly that the input signal that is used for PREDICTIVE CONTROL can not in real time, accurately reflect the functional activity state (like muscle fatigue) of target muscle, thereby makes the FES system have control accuracy problem and system stability problem.

Summary of the invention

The invention provides a kind of functional electric stimulation knee joint angle control method, this method has realized in real time, accurately reflecting the functional activity state of target muscle, has improved precision and stability, sees hereinafter for details and describes:

A kind of functional electric stimulation knee joint angle control method said method comprising the steps of:

(1) the ultrasonic muscle image of first knee joint angle under the electricity irritation of acquisition function property and target muscle;

(2) the ultrasonic muscle image to said target muscle carries out frequency domain filtering, utilizes the cross correlation method to extract the muscle thickness parameter of target muscle;

(3) pass through the grain of meat parameter that the gray level co-occurrence matrixes method is extracted target muscle;

(4) according to the muscle thickness parameter of stimulus intensity, said target muscle and grain of meat parameter acquiring second knee joint angle of said target muscle;

(5) obtain relative root-mean-square error, standard root-mean-square error and relative coefficient through said second knee joint angle and said first knee joint angle, choose the muscle thickness parameter of said stimulus intensity, said target muscle and the grain of meat combinations of parameters mode of said target muscle according to said relative root-mean-square error, said standard root-mean-square error and said relative coefficient.

Said ultrasonic muscle image to target muscle carries out frequency domain filtering, utilizes the cross correlation method to extract the muscle thickness parameter and is specially:

1) the ultrasonic muscle image of all said target muscle of gained is carried out frequency domain filtering and handle, obtain the ultrasonic muscle image that several handle back target muscle;

2) manual three picture element matrixs selecting first width of cloth muscle image in the ultrasonic muscle image of said several processing back target muscle;

3) through the cross correlation method three picture element matrixs of t (t＞1) width of cloth muscle image and three picture element matrixs of said first width of cloth muscle image are done relevantly, obtained the maximal correlation zone on three borders of the t width of cloth and first width of cloth;

4) obtain said muscle thickness parameter according to said three maximal correlations zone.

The beneficial effect of technical scheme provided by the invention is:

The invention provides a kind of functional electric stimulation knee joint angle control method; The knee joint angle under the present invention's acquisition function property electricity irritation simultaneously and the ultrasonic muscle image of target muscle; Adopt ultrasonoscopy treatment technologies such as filtering, zone coupling tracking to extract target muscle thickness information; Adopt gray level co-occurrence matrixes to extract target grain of meat information; As input, adopt SVMs and Artificial Neural Network model to obtain second knee joint angle muscle thickness information, texture information and stimulus intensity; Control signal precision of the present invention is high, because image information is with electromyographic signal and disturb and directly do not get in touch, so the interference of image information and noise are comparatively speaking, much smaller than signals of telecommunication such as the electric myoelectricities of brain; Ultra sonic imaging has superior accuracy and specificity in tissue characterization and bio-measurement; Can adjust input according to the functional status of current goal muscle (the following degradation of muscle activity under the FES that causes like muscle fatigue).Muscle B ultrasonic image information comprises muscle thickness and grain of meat; Existing research shows that the increase of muscle thickness can reflect muscle fatigue; Angle second moment, contrast, homogeneity and the entropy that comprises in the grain of meat can reflect the uniformity and the complexity of grain of meat under the FES; So the muscle image can fully reflect the state of muscle, for the accurate stimulation of FES provides assurance.

Description of drawings

Fig. 1 is the structural representation of experiment provided by the invention;

Fig. 2 is the flow chart of a kind of functional electric stimulation knee joint angle control method provided by the invention;

Fig. 3 is the sketch map of 3 10 * 40 picture element matrixs on first width of cloth image provided by the invention;

Fig. 4 is the sketch map of t width of cloth image cross-correlation result provided by the invention;

Fig. 5 is the sketch map of co-occurrence matrix generative process provided by the invention;

Fig. 6 is the sketch map of knee joint angle under the no-load condition provided by the invention;

Fig. 7 is the sketch map that knee joint angle under the loading condition is arranged provided by the invention.

The specific embodiment

For making the object of the invention, technical scheme and advantage clearer, will combine accompanying drawing that embodiment of the present invention is done to describe in detail further below.

In order in real time, accurately to reflect the functional activity state of target muscle, improve precision and stability, referring to Fig. 1 and Fig. 2, the embodiment of the invention provides a kind of functional electric stimulation knee joint angle control method, sees hereinafter for details and describes:

101: the ultrasonic muscle image of first knee joint angle under the electricity irritation of acquisition function property and target muscle;

Experimental subjects is sitting on the experiment chair, and hip joint becomes 90 ° with chair, and shank is sagging naturally, exposes lower limb quadriceps femoris muscle group to be measured position, about 25 ℃ of room temperatures.Angle probe is fixed in knee joint.The FES electrode is fixed in two sections tendon places of rectus femoris.Ultrasonic coupling agent is smeared at the place at the rectus femoris belly of muscle, and ultrasonic probe is fixed in rectus femoris belly of muscle place, keeps the probe plane radially vertical with lower limb.The experiment of muscle form and active studies comprises two parts under the FES:

Non-loaded experiment: stimulus intensity increases from 1 grade gradually, and each intensity continues 4s, till knee joint is stretched.The intensity that knee joint is begun to transport is designated as lower critical intensity, and the intensity when knee joint is stretched is designated as critical intensity, with down-on-variation of lower critical intensity is as a FES cycle.From first knee joint angle of the 1st grade of each grade of opening entry, and begin to gather the muscle image in first lower critical intensity.Gather ultrasonic muscle image at the 3s of each intensity, continue 8 cycles.

The load experiment is arranged: before the experiment beginning, will be that the counterweight ofhealth body weight 1/150 quality is fixed in experimenter's ankle place.Then, experimentation is identical with non-loaded experiment.Because rectus femoris has certain overlapping property with vastus intermedius in the quadriceps femoris system, thus test can gather simultaneously rectus femoris (rectus femoris, RF) and vastus intermedius (vastus intermedius, image information VI).

102: the ultrasonic muscle image to target muscle carries out frequency domain filtering, utilizes the cross correlation method to extract the muscle thickness parameter of target muscle;

Wherein, Because the boundary profile of muscle belongs to low-frequency component, and noise and details belong to radio-frequency component, so the muscle image of autonomous contraction is carried out frequency domain filtering; The low frequency profile information that only keeps the autonomous muscle image that shrinks is for subsequent extracted muscle thickness parameter is given security and prerequisite.

Wherein, the muscle thickness parameter extracting method is the cross correlation method.In nature or human society,, then claim relevant (correlation) between the variable if having the relation of the change that accompanies between the variable.For one-dimensional signal, the cross correlation method is the standard method that is used to estimate the dependency of two columns.Suppose to exist two columns x and y, i=0,1,2...N-1, the value of N is a positive integer, then two columns are about postponing for shown in the following formula of the correlation coefficient r of d (d):

$r\left(d\right)=\frac{{\mathrm{\Σ}}_{i=0}^{N-1}\left[(x\left(i\right)-\mathrm{mx})(y(i-d)-\mathrm{my})\right]}{\sqrt{{\mathrm{\Σ}}_{i=0}^{N-1}{(x\left(i\right)-\mathrm{mx})}^2}\sqrt{{\mathrm{\Σ}}_{i=0}^N{(y(i-d)-\mathrm{my})}^2}}$

Wherein, mx, my are respectively the average of two columns, and the span of r (d) is [1,1].R (d)=0 representes that two columns are uncorrelated; R (d)=-1, it is relevant to represent that two columns are maximum negative; R (d)=1 representes that two columns are maximum positive correlation.

The cross correlation method of one-dimensional signal is expanded in two dimensional image, can be used for the identification and the extraction of image boundary characteristic.Get a borderline region picture element matrix mark1 of first width of cloth image earlier, size is m * n.In second width of cloth image, find borderline region matrix mark2 according to the position of first width of cloth image the inside boundary region matrix mark1 earlier; Then according to vector (m; N) borderline region matrix mark2 is moved, then mark1 and mark2 are made cross correlation, shown in the following formula:

$r(m,n)=\frac{{\mathrm{\Σ}}_{i=0}^{M-1}{\mathrm{\Σ}}_{j=0}^{N-1}\left[(x(i,j)-\mathrm{mx})(y(i-m,j-n)-\mathrm{my})\right]}{\sqrt{{\mathrm{\Σ}}_{i=0}^{M-1}{\mathrm{\Σ}}_{j=0}^{N-1}{(x(i,j)-\mathrm{mx})}^2}\sqrt{{\mathrm{\Σ}}_{i=0}^{M-1}{\mathrm{\Σ}}_{i=0}^{N-1}{(y(i-d)-\mathrm{my})}^2}}$

(m n) is cross-correlation result to the r that obtains, and the position at mark2 place is the position after mark1 moves when cross-correlation result is maximum, and the value of M and N is a positive integer.

Wherein, the ultrasonic muscle image of target muscle is carried out frequency domain filtering, utilizes the cross correlation method to extract the muscle thickness parameter and be specially:

1) the ultrasonic muscle image of all target muscle of gained is carried out frequency domain filtering and handle, obtain the ultrasonic muscle image that several handle back target muscle;

2) manual three picture element matrixs selecting first width of cloth muscle image in the ultrasonic muscle image of several processing back target muscle;

Wherein, Referring to Fig. 3 and Fig. 4; This step is specially: after the experimenter passes through frequency domain filtering pretreatment to all muscle ultrasonoscopys of once testing gained; In first image, manually select 3 10 * 40 picture element matrixs, lay respectively at skin and rectus femoris border, rectus femoris and vastus intermedius border and vastus intermedius and femur border.

3) through the cross correlation method three picture element matrixs of t (t＞1) width of cloth muscle image and three picture element matrixs of first width of cloth muscle image are done relevantly, obtained the maximal correlation zone on three borders of the t width of cloth and first width of cloth;

This step is specially: the 1st picture element matrix that the t width of cloth is handled the autonomous muscle image that shrinks in back on image according to vector (m₁, n₁) move m wherein₁∈ [10,9], n₁∈ [10,9] calculates 400 cross correlation score of the 1st the 1st picture element matrix and the 1st picture element matrix of t width of cloth image, seeks maximum related value r_Max1Corresponding motion-vector (m_Max1, n_Max1), m then_Max1Be t width of cloth image with respect to the 1st width of cloth image thickness amount of movement; The 2nd picture element matrix on image according to vector (m₂, n₂) move m wherein₂∈ [30,19], n₂∈ [40,39]; The 3rd picture element matrix on image according to vector (m₃, n₃) move m wherein₃∈ [20,39], n₃∈ [10,19], wherein, the 2nd picture element matrix is identical with the 1st picture element matrix with the method that the 3rd picture element matrix calculates muscle thickness.The maximal correlation of seeking the t width of cloth image and the 1st width of cloth by the cross correlation method is regional, and it is as shown in Figure 4 to obtain the result.

4) obtain the muscle thickness parameter according to three maximal correlation zones.

103: the grain of meat parameter of extracting target muscle through the gray level co-occurrence matrixes method;

Texture is the notion commonly used in the graphical analysis, is an important and inenarrable characteristic of image, is an extremely important information source understanding image, and certain of reflection color of image and gray scale changes, and this variation is relevant with the attribute of object itself.It definitely is defined as certain picture element that regularity or randomness repeat to show texture.Local irregularities in the image and macroscopic view clocklike characteristic are referred to as texture.Texture analysis is a kind of important analytical method during pattern recognition and Flame Image Process are used, and it refers to through certain image processing techniques and extracts textural characteristics, thereby obtains the processing procedure of the quantitative or qualitative description of texture.

The grain of meat information extracting method is gray level co-occurrence matrixes method (gray level co-occurrence matrixmethod).Gray level co-occurrence matrixes carries out investigation statistics to all pixels of image, holds concurrently to reflect the gray value of image and the characteristic of two aspects of intensity profile, is a kind ofly can describe the spatial characteristics of gradation of image and the method for spatial coherence simultaneously.

(x, y), and (x+a, y+b), the right gray value of this point is designated as that (g1 g2), makes that (x y) moves, and then can obtain various (g1, g2) values on whole interesting areas to depart from its another point to get any point in the image.With in this method statistical picture at a distance of (a, the associating frequency probability P that two pixel gray values b) occur (g1, g2).If the progression of gray value is L, g1 then, the combination of g2 has the L2 kind; Whole interesting areas is counted each (g1; G2) (this number of times is co-occurrence matrix f (g1, the element value of the capable g2 row of g1 g2)) to the number of times that occurs, and is arranged in a square formation f (g1; G2), promptly obtain the co-occurrence matrix of image.With occurrence number normalization obtain each point to the Probability p that occurs (g1, g2).To same image, the gray scale quantized level is different, and range difference score value a and b get different combinations, obtain different co-occurrence matrixs.The selection that the gray level co-occurrence matrixes pixel is right mainly contains four direction: 0 °, 45 °, 90 ° and 135 °.The co-occurrence matrix generative process is as shown in Figure 5.

In order to describe the texture situation with co-occurrence matrix more intuitively, derive the parameter that some reflect the matrix situations from co-occurrence matrix, typically there are following several kinds:

(a) angle second moment (energy) ASM

$\mathrm{ASM}=\underset{i}{\mathrm{\&Sigma;}}\underset{j}{\mathrm{\&Sigma;}}{\left\{p(\mathrm{<figure-callout\; label="g"\; filenames="HDA0000117517930000011.png,HDA0000117517930000023.png"\; state="\{\{state\}\}">g</figure-callout>}1,g2)\right\}}^2$

ASM is the quadratic sum of gray level co-occurrence matrixes element value, so also claim energy, has reflected gradation of image be evenly distributed degree and texture fineness degree.If all values of co-occurrence matrix equates that all then the ASM value is little; On the contrary, other value is little if the some of them value is big, and then the ASM value is big.When distributing in the element set in the co-occurrence matrix, this moment, the ASM value was big.The ASM value shows a kind of uniform, regular texture pattern that changes greatly.

(b) contrast C ON

$\mathrm{CON}=\underset{i}{\mathrm{\&Sigma;}}\underset{j}{\mathrm{\&Sigma;}}{(g1-g2)}^{2}p(\mathrm{<figure-callout\; label="g"\; filenames="HDA0000117517930000011.png,HDA0000117517930000023.png"\; state="\{\{state\}\}">g</figure-callout>}1,g2)$

The definition of image and the degree of the texture rill depth have been reflected.The texture rill is dark more, and its contrast is big more, and visual effect is clear more; Otherwise contrast is little, and then rill is shallow, and effect is fuzzy.Gray scale difference is that the big pixel of contrast is many to more, and this value is big more.Big more away from cornerwise element value in the gray level co-occurrence matrixes, CON is big more.

(c) homogeneity (unfavourable balance distance) HOM

$\mathrm{HOM}=\underset{i}{\mathrm{\&Sigma;}}\underset{\mathrm{<figure-callout\; label="j"\; filenames="HDA0000117517930000011.png,HDA0000117517930000023.png"\; state="\{\{state\}\}">j</figure-callout>}}{\mathrm{\&Sigma;}}\frac{1}{1+{(g1-g2)}^2}p(\mathrm{<figure-callout\; label="g"\; filenames="HDA0000117517930000011.png,HDA0000117517930000023.png"\; state="\{\{state\}\}">g</figure-callout>}1,g2)$

HOM reflects the homogeneity of image texture, and the image texture localized variation what are measured.Its value explains then greatly between the zones of different of image texture and lacks variation that the part is very even.

(d) entropy ENT

$\mathrm{ENT}=-\underset{i}{\mathrm{\&Sigma;}}\underset{j}{\mathrm{\&Sigma;}}p(\mathrm{<figure-callout\; label="g"\; filenames="HDA0000117517930000011.png,HDA0000117517930000023.png"\; state="\{\{state\}\}">g</figure-callout>}1,g2)\log p(\mathrm{<figure-callout\; label="g"\; filenames="HDA0000117517930000011.png,HDA0000117517930000023.png"\; state="\{\{state\}\}">g</figure-callout>}1,g2)$

ENT is the tolerance of the quantity of information that has of image, and texture information also belongs to the information of image, is the tolerance of a randomness, when all elements in the co-occurrence matrix maximum randomness is arranged, when all elements is almost equal, entropy is bigger.It has represented the non-uniform degree or the complexity of texture in the image.

104: according to the muscle thickness parameter of stimulus intensity, target muscle and grain of meat parameter acquiring second knee joint angle of target muscle;

Wherein, obtain second knee joint angle and can adopt support vector machine method or artificial log on method or other method, when specifically realizing, the embodiment of the invention does not limit this.

SVM (support vector machine; SVMs) be that the VC that is based upon Statistical Learning Theory ties up on theoretical and the structure risk minimum principle basis; Between the complexity of model and learning capacity, seek optimal compromise according to limited sample information, in the hope of obtaining best popularization ability.SVM shows many distinctive advantages in solving small sample, non-linear and higher-dimension pattern recognition, and can promote the use of in the other machines problem concerning study such as function match.

Artificial neural network is to human brain or ANN (natural neural network; The nature neutral net) the abstract and simulation of some fundamental characteristics; It is the basis with the physiological Study achievement to brain, and its purpose is to simulate some mechanism and mechanism of brain, realizes the function of certain aspect.The characteristics of the self study that ANN has, association's storage, at a high speed a large amount of computings.Especially the mechanism of problem is had little understanding and can not or the difficult system that representes with mathematical model, the best often instrument of ANN.

The embodiment of the invention adopts the Forecasting Methodology of SVM and ANN, adopts set stimulus intensity information to combine the muscle thickness parameter of lower limb target muscle and grain of meat parameter acquiring second knee joint angle of target muscle.For example: when adopting SVM as Forecasting Methodology, respectively with stimulus intensity (S), stimulus intensity and rectus femoris thickness (S-RF), stimulus intensity and vastus intermedius thickness (S-VI), stimulus intensity and gross thickness (S-RFVI), stimulus intensity and rectus femoris texture (S-RFcom), stimulus intensity and vastus intermedius texture (S-VIcom) as importing; When adopting ANN as Forecasting Methodology, respectively with stimulus intensity and all thickness parameters (S-DEP), stimulus intensity and all parametric textures (S-TEX), stimulus intensity and all grain of meat parameters (S-DEP-TEX) as importing.

105: obtain relative root-mean-square error, standard root-mean-square error and relative coefficient through second knee joint angle and first knee joint angle, choose the muscle thickness parameter of stimulus intensity, target muscle and the grain of meat combinations of parameters mode of target muscle according to relative root-mean-square error, standard root-mean-square error and relative coefficient.

Wherein, (a) relative root-mean-square error:

$\mathrm{RMSD}=\sqrt{\frac{\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{(\mathrm{ptesty}\left(i\right)-\mathrm{testy}\left(i\right))}^2}{\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{\left(\mathrm{testy}\left(i\right)\right)}^2}}$

Wherein, ptesty representes second knee joint angle;

Testy representes first knee joint angle;

N representes the number of angle points in the one-period, and the value of n is a positive integer.Identical as follows.

Root-mean-square error characterizes the relative error of second knee joint angle and first knee joint angle.

(b) standard root-mean-square error:

$\mathrm{RMSE}=\sqrt{\frac{\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{(\mathrm{ptesty}\left(i\right)-\mathrm{testy}\left(i\right))}^2}{n}}$

Root-mean-square error characterizes the absolute error of second knee joint angle and first knee joint angle.

(c) correlation coefficient:

$\mathrm{RR}=\frac{\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}(\mathrm{ptesty}\left(i\right)-{\mathrm{mean}}_1)(\mathrm{testy}\left(i\right)-{\mathrm{mean}}_2)}{\sqrt{\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{(\mathrm{ptesty}\left(i\right)-{\mathrm{mean}}_1)}^2\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{(\mathrm{testy}\left(i\right)-{\mathrm{mean}}_2)}^2}}$

Wherein, mean₁The average of representing second knee joint angle, mean₂The average of representing first knee joint angle.

Correlation coefficient characterizes the similarity degree of second knee joint angle and first knee joint angle.

According to the minimum error principle, in different input parameters, choose optimum input parameter according to relative root-mean-square error, standard root-mean-square error and relative coefficient.

Below in conjunction with concrete test the feasibility of the embodiment of the invention is described, is seen for details hereinafter and describe:

Test procedure is specially: angle probe is fixed in knee joint, and the FES electrode is fixed in two sections tendon places of rectus femoris.Ultrasonic coupling agent is smeared at the place at the rectus femoris belly of muscle, and ultrasonic probe is fixed in rectus femoris belly of muscle place, keeps the probe plane radially vertical with lower limb.The experimental procedure of muscle form and active studies is under the FES: stimulus intensity increases from 1 grade gradually, and each intensity continues 4s, till knee joint is stretched.The intensity of knee joint setting in motion is designated as lower critical intensity, and the intensity when knee joint is stretched is designated as critical intensity, with down-on-variation of lower critical intensity is as a FES cycle.From the knee joint angle of the 1st grade of each grade of opening entry, and begin to gather the muscle image in first lower critical intensity.Gather ultrasonic muscle image at the 3s of each intensity, continue 8 cycles.Because rectus femoris and vastus intermedius have certain overlapping property in the quadriceps femoris system, so the image information that experiment can be gathered RF (rectus femoris, rectus femoris) and VI (vastus intermedius, vastus intermedius) simultaneously.Referring to Fig. 6 and Fig. 7; The embodiment of the invention has provided predicting the outcome of following 8 experimenters of non-loaded contraction and has had load to shrink predicting the outcome of following 8 experimenters, and wherein, S represents stimulus intensity; RFVI represents gross thickness; DEP represents all thickness parameters, and VIcom represents the vastus intermedius texture, and TEX represents all parametric textures; SVM represents SVMs, ANN representative artificial neural networks, as can be seen from Figure 6; 8 experimenter's joint angles RMSE＜4.5, RMSD＜0.09, RR＞97%; RFcom represents the vastus intermedius texture, as can be seen from Figure 7, and 8 experimenter's joint angles RMSE＜6.0; RMSD＜0.10, RR＞95%.

The result of this test shows: no matter load collapsed mode or non-loaded collapsed mode are arranged; Unite as input signal with muscle thickness information or texture information and stimulus intensity and to be used to obtain second knee joint angle; Only be superior to stimulus intensity as input signal; Be that relative root-mean-square error, standard root-mean-square error are littler, guaranteed higher dependency simultaneously, demonstrated fully the important function of muscle information on musculation state under the reflection FES.

In sum; The embodiment of the invention provides a kind of functional electric stimulation knee joint angle control method; The knee joint angle under the embodiment of the invention acquisition function property electricity irritation simultaneously and the ultrasonic muscle image of target muscle adopt ultrasonoscopy treatment technologies such as filtering, boundary matching tracking to extract target muscle thickness information, adopt gray level co-occurrence matrixes to extract target grain of meat information; Muscle thickness information, texture information and stimulus intensity as input, are obtained second knee joint angle; The control signal precision of the embodiment of the invention is high, because image information is with electromyographic signal and disturb and directly do not get in touch, so the interference of image information and noise are comparatively speaking, much smaller than signals of telecommunication such as the electric myoelectricities of brain; Ultra sonic imaging has superior accuracy and specificity in tissue characterization and bio-measurement; Can adjust input according to the functional status of current goal muscle (the following degradation of muscle activity under the FES that causes like muscle fatigue).Muscle B ultrasonic image information comprises muscle thickness and grain of meat; Existing research shows that the increase of muscle thickness can reflect muscle fatigue; Angle second moment, contrast, homogeneity and the entropy that comprises in the grain of meat can reflect the uniformity and the complexity of grain of meat under the FES; So the muscle image can fully reflect the state of muscle, for the accurate stimulation of FES provides assurance.

It will be appreciated by those skilled in the art that accompanying drawing is the sketch map of a preferred embodiment, the invention described above embodiment sequence number is not represented the quality of embodiment just to description.

The above is merely preferred embodiment of the present invention, and is in order to restriction the present invention, not all within spirit of the present invention and principle, any modification of being done, is equal to replacement, improvement etc., all should be included within protection scope of the present invention.

Claims (2)

1. a functional electric stimulation knee joint angle control method is characterized in that, said method comprising the steps of:

(1) the ultrasonic muscle image of first knee joint angle under the electricity irritation of acquisition function property and target muscle;

(2) the ultrasonic muscle image to said target muscle carries out frequency domain filtering, utilizes the cross correlation method to extract the muscle thickness parameter of target muscle;

(3) pass through the grain of meat parameter that the gray level co-occurrence matrixes method is extracted target muscle;

(4) according to the muscle thickness parameter of stimulus intensity, said target muscle and grain of meat parameter acquiring second knee joint angle of said target muscle;

(5) obtain relative root-mean-square error, standard root-mean-square error and relative coefficient through said second knee joint angle and said first knee joint angle, choose the muscle thickness parameter of said stimulus intensity, said target muscle and the grain of meat combinations of parameters mode of said target muscle according to said relative root-mean-square error, said standard root-mean-square error and said relative coefficient.

2. a kind of functional electric stimulation knee joint angle control method according to claim 1 is characterized in that, said ultrasonic muscle image to target muscle carries out frequency domain filtering, utilizes the cross correlation method to extract the muscle thickness parameter and is specially:

1) the ultrasonic muscle image of all said target muscle of gained is carried out frequency domain filtering and handle, obtain the ultrasonic muscle image that several handle back target muscle;

2) manual three picture element matrixs selecting first width of cloth muscle image in the ultrasonic muscle image of said several processing back target muscle;

3) through the cross correlation method three picture element matrixs of t (t＞1) width of cloth muscle image and three picture element matrixs of said first width of cloth muscle image are done relevantly, obtained the maximal correlation zone on three borders of the t width of cloth and first width of cloth;

4) obtain said muscle thickness parameter according to said three maximal correlations zone.

Images