Movatterモバイル変換


[0]ホーム

URL:


CN120144985A - Lower limb muscle fatigue state recognition method and system based on multimodal feature fusion - Google Patents

Lower limb muscle fatigue state recognition method and system based on multimodal feature fusion
Download PDF

Info

Publication number
CN120144985A
CN120144985ACN202510625199.5ACN202510625199ACN120144985ACN 120144985 ACN120144985 ACN 120144985ACN 202510625199 ACN202510625199 ACN 202510625199ACN 120144985 ACN120144985 ACN 120144985A
Authority
CN
China
Prior art keywords
muscle
fatigue state
force
lower limb
feature fusion
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202510625199.5A
Other languages
Chinese (zh)
Other versions
CN120144985B (en
Inventor
张虹淼
于嘉浩
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Suzhou University
Original Assignee
Suzhou University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Suzhou UniversityfiledCriticalSuzhou University
Priority to CN202510625199.5ApriorityCriticalpatent/CN120144985B/en
Publication of CN120144985ApublicationCriticalpatent/CN120144985A/en
Application grantedgrantedCritical
Publication of CN120144985BpublicationCriticalpatent/CN120144985B/en
Activelegal-statusCriticalCurrent
Anticipated expirationlegal-statusCritical

Links

Classifications

Landscapes

Abstract

The invention discloses a lower limb muscle fatigue state identification method and system based on multi-mode feature fusion, wherein the lower limb muscle fatigue state identification method based on multi-mode feature fusion comprises the following steps of S1, arranging a multi-mode sensor on an object to be tested, collecting signals through the multi-mode sensor, acquiring bioelectricity signals, muscle form information and force position information through the multi-mode sensor, and S2, acquiring fatigue states through a fatigue state identification model according to the bioelectricity signals, the muscle form information and the force position information. The invention discloses a lower limb muscle fatigue state identification method and system based on multi-mode feature fusion, and the method and system have the characteristics of high detection precision, quick dynamic response and strong anti-interference performance.

Description

Lower limb muscle fatigue state identification method and system based on multi-modal feature fusion
Technical Field
The invention relates to the technical field of multi-modal feature recognition, in particular to a lower limb muscle fatigue state recognition method and system based on multi-modal feature fusion.
Background
The detection of muscle fatigue has profound significance in the fields of sports science, rehabilitation medicine, occupational health and the like, and the real-time performance and the accuracy of the detection are vital to individual health management and sports performance optimization. The existing muscle fatigue detection method has the following defects:
1. The prior art is mostly dependent on a single biological signal (such as surface electromyography signals sEMG, heart rate and the like) to be easily interfered by noise, and the characteristic loss or the single characteristic of the joint angle can be caused by high-frequency noise, so that the coupling relation between the dynamic change of the muscle morphology and the mechanical state can not be reflected;
2. the prior art lacks real-time dynamic capturing capability for fatigue evaluation of continuous repetitive motion, does not consider the influence of muscle section deformation and muscle deformation on fatigue, and the bioelectric signal sensor is easy to suffer from motion artifact and environmental interference;
3. poor generalization of the model, and insufficient classification precision of the machine learning model on complex fatigue states caused by single feature dimension.
Disclosure of Invention
The invention overcomes the defects of the prior art, and provides a lower limb muscle fatigue state identification method and system based on multi-mode feature fusion, which have the characteristics of high detection precision, quick dynamic response and strong anti-interference performance.
In order to achieve the aim, the technical scheme adopted by the invention is that the lower limb muscle fatigue state identification method based on multi-mode feature fusion comprises the following steps:
Step S1, arranging a multi-mode sensor on an object to be detected, and collecting signals through the multi-mode sensor, wherein the signal collection comprises the steps of acquiring bioelectric signals, muscle form information and force position information through the multi-mode sensor;
and S2, obtaining a fatigue state through a fatigue state identification model according to the bioelectric signals, the muscle morphology information and the force position information.
In a preferred embodiment of the invention, the multi-mode sensor comprises a bioelectric signal sensor, a muscle morphology collector and a force position information collection device;
The bioelectric signals on the object to be detected are collected through the bioelectric signal sensor, the muscle form information on the object to be detected is collected through the muscle form collector, and the force position information on the object to be detected is collected through the force position information collecting device.
In a preferred embodiment of the present invention, the electromyographic feature extraction is performed on the bioelectrical signal by a fatigue state recognition model, and the electromyographic feature extraction step includes:
preprocessing bioelectric signals acquired by the multi-mode sensor;
and then acquiring time domain features, frequency domain features and nonlinear features according to the preprocessed bioelectric signals.
In a preferred embodiment of the invention, the time domain features comprise a root mean square RMS, the root mean square RMS comprising:
Wherein Xi is the voltage value of the ith discrete sampling point in the electromyographic signal sequence, N is the total number of discrete data points in the electromyographic signal time window, and is calculated for the surface electromyographic signal sEMG signal with the length of N;
and/or the frequency domain features comprise a median frequency MF, the median frequency MF comprising performing Fourier transform on the preprocessed signal, calculating the power spectral density of the signal;
Where ω is the signal frequency.
In a preferred embodiment of the present invention, the nonlinear characteristic comprises a sample entropy SampEn, and the sample entropy SampEn comprises:
for a sequence X (N) = { X (1), X (2), X (N), given an embedding dimension m, the original sequence forms a set of vectors Xm (i) of length N-m +1,
Xm (i) = { X (i), X (i+1),. The term, X (i+m-1) }, where 1≤i≤n-m+1; define d [ Xm(i),Xm (j) ] as the maximum distance between the corresponding vectors in the two different vector sets, N is the position index of the data point in the time series, representing the nth time point;
d [ Xm(i),Xm (j) ]=max (|x (i+k) -X (j+k) |), wherein 0≤k≤m-1, 1≤i, j≤N-m+1, i is not equal to j, given that r, r is a similarity tolerance; The ratio of the number defined as d [ Xm(i),Xm (j) ]. Ltoreq.r to the total number of vectors N-m+1 is: Wherein Bi (r) is the number of d [ Xm(i),Xm (j) ]. Ltoreq.r; Bi(r)=num{d[Xm(i),Xm (j) ]. Ltoreq.r },The average value of (2) is expressed asIs applicable to all the fields of i,;
When the embedding dimension is m+1, the ratio of the number of d [ Xm+1(i),Xm+1 (j) ]. Ltoreq.r to the total number of vectors N-m is recorded as,Wherein Ai (r) is the number of D [ Xm+1(i),Xm+1 (j) ]. Ltoreq.r, and the average value of Ai(r)=num{d[Xm+1(i),Xm+1(j)]≤r},Aim (r) is represented as Dm+1 (r), applicable to all i,The sample entropy of the time series is:
In a preferred embodiment of the present invention, the muscle morphology information is extracted by a fatigue state recognition model, and the muscle morphology feature extraction step includes:
according to the bioelectric signal sensor in the multi-mode sensor, an infrared optical motion capturing system is adopted to capture the three-dimensional coordinates of the set marking points on the object to be detected, and the radius of the muscle section circle, the surface stretching length and the triangular area change of the marking points are dynamically calculated according to the three-dimensional coordinates of the marking points.
In a preferred embodiment of the invention, the calculating step of the triangular area change of the mark point comprises the steps of calculating the triangular area A by each side length of the triangle based on the sea-renformula;
; Wherein a, b and c are three sides of a triangle;
the calculation step of the muscle section circle radius comprises the following steps of according to a calculation formula of the circumscribed circle radius RCalculating the radius of a circumscribed circle, wherein the radius of the circumscribed circle is equal to the radius of the muscle section circle;
And/or the calculation step of the surface stretching length comprises the steps of calculating a central angle theta through cosine theorem based on the radius of the muscle section circle, and calculating the arc length according to the radius and the central angle, wherein the radius of the muscle section circle is as follows: central angle: wherein R is the radius of the muscle section circle, C is the chord length and the arc length: wherein arc is arc length, and the surface stretching length is muscle surface strain obtained by multiplying the central angle radian value and the radius.
In a preferred embodiment of the present invention, the force bit information is extracted by a fatigue state recognition model, and the force bit joint feature extracting step includes:
obtaining a ground reaction force peak value Fz-max by searching the maximum ground reaction force in the reverse jump process, wherein the ground reaction force peak value Fz-max is regarded as the maximum ground reaction force peak value Fz-max by traversing the maximum ground reaction force in the reverse jump process; , wherein,As a function of the change in the ground reaction force in the vertical direction Z over time, i.e. the vertical component value of the ground reaction force at time t,The force bit information acquisition equipment of the multi-mode sensor directly reads the force bit information;
The reverse jump height is calculated through the vertical displacement of the hip joint mark points, and the method comprises the following steps:
The z-axis coordinate in the rest state is subtracted from the highest point value by traversing the highest point of the z-axis coordinate of the mark point in the reverse jump process to obtain the reverse jump height,, wherein,To mark the point z-axis coordinate highest point,A z-axis coordinate when in a resting state;
the method for calculating the joint angle comprises the following steps:
And establishing an equivalent model of the object to be measured through three marking points arranged on the object to be measured, and calculating angles in x, y and z planes between relatively movable parts according to the equivalent model on the object to be measured to obtain the joint angle of the object to be measured.
In a preferred embodiment of the invention, the fatigue state recognition model comprises a random forest algorithm, wherein the random forest algorithm comprises the steps that a base unit adopts a decision tree, and an optimal splitting characteristic and a threshold value are selected by using a coefficient of Kernin;
The fatigue state recognition method of the fatigue state recognition model comprises the steps of dividing fatigue grades according to joint angle offset and force position characteristics, and generating a label;
model performance was assessed by cross-validation and fatigue status was output.
In a preferred embodiment of the invention, the lower limb muscle fatigue state recognition system based on multi-modal feature fusion comprises a fatigue state recognition model and a multi-modal sensor for providing bioelectric signals, muscle morphology information and force position information for the fatigue state recognition model;
The multi-mode sensor comprises a bioelectric signal sensor, a muscle form collector and a force position information collection device, wherein the bioelectric signal sensor, the muscle form collector and the force position information collection device are respectively interconnected with the fatigue state recognition model through a wireless myoelectricity collection module, an infrared optical motion capture system and a force measuring platform system;
the method is used for realizing the lower limb muscle fatigue state identification method based on multi-mode feature fusion.
The invention solves the defects existing in the technical background, and has the beneficial technical effects that:
a lower limb muscle fatigue state identification method and system based on multi-mode feature fusion have the characteristics of high detection precision, quick dynamic response and strong anti-interference performance.
The multi-mode feature fusion enables the fatigue recognition accuracy to be further improved compared with a single myoelectricity method.
The real-time synchronous acquisition of the muscle section morphology and the mechanical parameters realizes the rapid fatigue state update.
The introduction of various motion features reduces the interference of artifacts on the electromyographic signals.
Drawings
The invention will be further described with reference to the drawings and examples.
FIG. 1 is a schematic illustration of a multi-modal sensor arrangement on the surface of the rectus femoris muscle in a preferred embodiment of the present invention;
FIG. 2 is a schematic representation of a multi-modal sensor arrangement in reverse-jump joints in accordance with a preferred embodiment of the present invention;
FIG. 3 is a schematic diagram of a second (back) view of the reverse jump joint arrangement multi-modal sensor in accordance with the preferred embodiment of the present invention;
FIG. 4 is a schematic diagram of a workflow of a method and system for identifying a muscular fatigue status of a lower limb based on multi-modal feature fusion in a preferred embodiment of the present invention;
FIG. 5 is a schematic diagram of a workflow of a method and system for identifying fatigue status of lower limb muscles based on multi-modal feature fusion using a random forest fatigue identification model according to a preferred embodiment of the present invention;
FIG. 6 is a graph showing dynamic radius calculation in the method and system for identifying the fatigue state of lower limb muscles based on multi-modal feature fusion according to the preferred embodiment of the present invention;
FIG. 7 is a schematic diagram of dynamic radius calculation in the method and system for identifying the fatigue state of lower limb muscles based on multi-modal feature fusion according to the preferred embodiment of the invention;
FIG. 8 is a schematic diagram of the myoelectricity feature (root mean square for example) calculation in the method and system for identifying the fatigue state of the lower limb muscle based on the multi-modal feature fusion according to the preferred embodiment of the invention;
FIG. 9 is a schematic diagram of a calculation flow of the radius of the circle, the arc length and the triangular area of the section of the muscle in the method and the system for identifying the fatigue state of the muscle of the lower limb based on the multi-mode feature fusion according to the preferred embodiment of the invention;
FIG. 10 is a schematic diagram of maximum ground reaction force acquisition in a method and system for identifying a lower limb muscle fatigue state based on multi-modal feature fusion according to a preferred embodiment of the present invention;
FIG. 11 is a diagram showing the maximum jump height acquisition in the method and system for identifying the fatigue state of lower limb muscles based on multi-modal feature fusion according to the preferred embodiment of the present invention;
FIG. 12 is a schematic diagram showing the acquisition of the variation of the joint angle in the method and system for identifying the fatigue state of the lower limb muscle based on the multi-mode feature fusion according to the preferred embodiment of the invention;
fig. 13 is a schematic flow chart of a random forest algorithm according to a preferred embodiment of the present invention.
Detailed Description
The invention will now be described in further detail with reference to the drawings and examples, which are simplified schematic illustrations of the basic structure of the invention, which are presented only by way of illustration, and thus show only the structures that are relevant to the invention.
It should be noted that, if a directional indication (such as up, down, bottom, top, etc.) is involved in the embodiment of the present invention, the directional indication is merely used to explain the relative positional relationship between the components, the movement situation, etc. in a certain specific posture, and if the specific posture is changed, the directional indication is correspondingly changed. The terms "first," "second," and the like, are used for descriptive purposes only and are not to be construed as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defining "a first" or "a second" may explicitly or implicitly include one or more such feature. Unless specifically stated or limited otherwise, the terms "disposed," "connected," and "connected" are to be construed broadly, and may be, for example, fixedly connected, detachably connected, or integrally connected, or may be directly connected, or may be indirectly connected through an intermediate medium, or may be in communication with the interior of two elements. The specific meaning of the above terms in the present invention will be understood in specific cases by those of ordinary skill in the art.
Example 1
As shown in fig. 1-13, the method for identifying the fatigue state of the lower limb muscles based on multi-modal feature fusion comprises the following steps:
Step S1, arranging a multi-mode sensor on an object to be detected, and collecting signals through the multi-mode sensor, wherein the signal collection comprises the steps of acquiring bioelectric signals, muscle form information and force position information through the multi-mode sensor;
and S2, obtaining a fatigue state through a fatigue state identification model according to the bioelectric signals, the muscle morphology information and the force position information.
The multi-mode sensor comprises a bioelectric signal sensor, a muscle form collector and a force position information collection device, wherein bioelectric signals on an object to be measured are collected through the bioelectric signal sensor, muscle form information on the object to be measured is collected through the muscle form collector, and force position information on the object to be measured is collected through the force position information collection device.
The myoelectricity characteristic extraction method comprises the steps of preprocessing bioelectricity signals acquired by a multi-mode sensor, including band-pass filtering, denoising and segmentation, and then obtaining time domain characteristics, frequency domain characteristics and nonlinear characteristics according to the preprocessed bioelectricity signals.
The method comprises the steps of capturing three-dimensional coordinates of a set mark point on an object to be detected by an infrared optical motion capturing system according to a bioelectric signal sensor in a multi-mode sensor, and dynamically calculating the radius of a muscle section circle, the stretching length of the surface and the triangular area change of the mark point according to the three-dimensional coordinates of the mark point.
Example two
As shown in fig. 1-13, the method for identifying the fatigue state of the lower limb muscles based on multi-modal feature fusion comprises the following steps:
Step S1, arranging a multi-mode sensor on an object to be detected, and collecting signals through the multi-mode sensor, wherein the signal collection comprises the steps of acquiring bioelectric signals, muscle form information and force position information through the multi-mode sensor;
and S2, obtaining a fatigue state through a fatigue state identification model according to the bioelectric signals, the muscle morphology information and the force position information.
The multi-mode sensor comprises a bioelectric signal sensor, a muscle form collector and a force position information collection device, wherein bioelectric signals on an object to be measured are collected through the bioelectric signal sensor, muscle form information on the object to be measured is collected through the muscle form collector, and force position information on the object to be measured is collected through the force position information collection device.
The myoelectricity characteristic extraction method comprises the steps of preprocessing bioelectricity signals acquired by a multi-mode sensor, including band-pass filtering, denoising and segmentation, and then obtaining time domain characteristics, frequency domain characteristics and nonlinear characteristics according to the preprocessed bioelectricity signals. More specifically, the preprocessing includes bandpass filtering, denoising and segmentation. The preprocessing of the data is to preprocess the data of three modes, namely bioelectric signals, muscle form information and force position information, which are acquired by a bioelectric signal sensor (namely a wireless myoelectricity acquisition module), a muscle form acquisition device (namely an infrared optical motion capture system) and a force position information acquisition device (namely a force measuring table system). The preprocessing comprises band-pass filtering, denoising and segmentation processing for bioelectricity signals, and segmentation processing for muscle morphology information and force position information. Wherein, the band-pass filter adopts a Chebyshev I-type filter (20-450 Hz), the denoising adopts wavelet decomposition denoising, the wavelet basis function adopts Daubechies4 (db 4), and the wavelet packet decomposition layer number is 4.
Further, the time domain features include a root mean square RMS, the root mean square RMS comprising:
Wherein Xi is the voltage value of the i-th discrete sample point in the electromyographic signal sequence.
Further, the frequency domain features include a median frequency MF, which includes fourier transforming the preprocessed signal, calculating a power spectral density P (ω) of the signal,Where ω is the signal frequency.
Further, the nonlinear feature includes a sample entropy SampEn, and the sample entropy SampEn includes:
for a sequence X (N) = { X (1), X (2), X (N), given an embedding dimension m, the original sequence forms a set of vectors Xm (i) of length N-m +1,
Xm (i) = { X (i), X (i+1),. The term, X (i+m-1) }, where 1≤i≤n-m+1; define d [ Xm(i),Xm (j) ] as the maximum distance between the corresponding vectors in the two different vector sets, N is the position index of the data point in the time series, representing the nth time point;
d [ Xm(i),Xm (j) ]=max (|x (i+k) -X (j+k) |), wherein 0≤k≤m-1, 1≤i, j≤N-m+1, i is not equal to j, given that r, r is a similarity tolerance; The ratio of the number defined as d [ Xm(i),Xm (j) ]. Ltoreq.r to the total number of vectors N-m+1 is: Wherein Bi (r) is the number of d [ Xm(i),Xm (j) ]. Ltoreq.r; Bi(r)=num{d[Xm(i),Xm (j) ]. Ltoreq.r },The average value of (2) is expressed asIs applicable to all the fields of i,;
When the embedding dimension is m+1, the ratio of the number of d [ Xm+1(i),Xm+1 (j) ]. Ltoreq.r to the total number of vectors N-m is recorded as,Wherein Ai (r) is the number of D [ Xm+1(i),Xm+1 (j) ]. Ltoreq.r, and the average value of Ai(r)=num{d[Xm+1(i),Xm+1(j)]≤r},Aim (r) is represented as Dm+1 (r), applicable to all i,The sample entropy of the time series is:
The method comprises the steps of capturing three-dimensional coordinates of a set mark point on an object to be detected by an infrared optical motion capturing system according to a bioelectric signal sensor in a multi-mode sensor, and dynamically calculating the radius of a muscle section circle, the stretching length of the surface and the triangular area change of the mark point according to the three-dimensional coordinates of the mark point.
Further, muscle morphological characteristics are extracted from muscle morphological information through a fatigue state recognition model, and the muscle morphological characteristics extraction step comprises the steps of capturing three-dimensional coordinates of a set mark point on an object to be detected by an infrared optical motion capturing system according to a bioelectric signal sensor in a multi-mode sensor, and dynamically calculating the radius of a muscle section circle, the stretching length of the surface and the triangular area change of the mark point according to the three-dimensional coordinates of the mark point.
Further, the calculating step of the triangular area change of the mark point comprises the steps of calculating the triangular area A through the side lengths of the triangle based on the sea-land formula,;Wherein a, b and c are three sides of a triangle.
Further, the calculation step of the radius of the muscle section circle comprises the following steps of according to a calculation formula of the radius R of the circumscribed circleAnd calculating the radius of the circumscribed circle, wherein the radius of the circumscribed circle is equal to the radius of the muscle section circle.
Further, the calculation step of the surface stretching length comprises the steps of calculating a central angle theta based on the radius of the muscle section circle by cosine theorem and calculating the arc length according to the radius and the central angle, wherein the radius of the muscle section circle is as follows: central angle: wherein R is the radius of the muscle section circle, C is the chord length and the arc length: wherein arc is arc length, and the surface stretching length is muscle surface strain obtained by multiplying the central angle radian value and the radius.
Specifically, the force-position joint feature is extracted from the force-position information through the fatigue state recognition model, and the force-position joint feature extraction step comprises the following steps:
Further, the step of obtaining the ground reaction force peak value Fz-max by finding the maximum ground reaction force during reverse jump, includes the step of determining the maximum value as the ground reaction force peak value Fz-max by traversing the ground reaction force maximum value during reverse jump,, wherein,As a function of the ground reaction force in the vertical direction Z over time.
Further, the reverse jump height is calculated through the vertical displacement of the hip joint mark points, and the method comprises the following steps:
The z-axis coordinate in the rest state is subtracted from the highest point value by traversing the highest point of the z-axis coordinate of the mark point in the reverse jump process to obtain the reverse jump height,, wherein,To mark the point z-axis coordinate highest point,A z-axis coordinate when in a resting state;
Further, the method for calculating the joint angle comprises the following steps:
And establishing an equivalent model of the object to be measured through three marking points arranged on the object to be measured, and calculating angles in x, y and z planes between relatively movable parts according to the equivalent model on the object to be measured to obtain the joint angle of the object to be measured.
Example III
On the basis of the second embodiment, the fatigue state recognition model comprises a random forest algorithm, wherein the random forest algorithm comprises the steps that a base unit adopts a decision tree, and optimal splitting characteristics and thresholds are selected by using a coefficient of a foundation. Specifically, the number of decision trees is 300, the number of minimum leaf nodes is 10, and the maximum depth is 20.
Example IV
On the basis of the third embodiment or the second embodiment, a ErgoLAB surface myoelectricity measurement system is adopted as the bioelectric signal sensor (myoelectric signal acquisition equipment) in the embodiment, the sampling frequency is 1000Hz, an 8-channel acquisition electrode is included, and Bluetooth 5.0 wireless signal transmission is supported. The muscle shape collector (motion capture system) adopts a Vicon optical motion capture system, the model number is ViconMx-GIGANET, the hardware comprises 16 cameras ViconMX, a PC host, viconDatastationADCPatchPanel information conversion boxes, MX special connecting wires and the like, and the sampling frequency is 100Hz. The force position information acquisition equipment (force measuring platform) adopts a three-dimensional force measuring platform in the prior art, is internally provided with six high-precision mechanical sensors, can acquire mechanical signals acting on the platform in real time, and can directly obtain the following mechanical parameters of Fx, fy, fz, mx, my, mz, a three-dimensional force vector, a pressure Center (COP) and a sampling frequency of 1000Hz. More specifically, in this embodiment, the object to be measured is a lower limb, so that the myoelectricity electrodes are arranged according to the anatomical position of main fatigue muscles of the lower limb (such as bilateral rectus femoris), the myoelectricity electrodes are stuck on the surface of the myoabdomen, the motion capture Mark points (Mark points) are arranged, the reflective Mark points (interval 40 mm) are transversely and equidistantly distributed on the surface of the muscle based on the distribution rule of the strain of the section of the rectus femoris, and meanwhile, the Mark points are stuck on the ankle, knee and hip joints to synchronously capture reverse jump kinematic data, and the mechanical data is acquired, wherein a pressure sensitive area is arranged on the surface of a force measuring table, and three-dimensional ground reaction force during reverse jump is synchronously acquired. Wherein Fx is a force component along the X-axis and is defined as a horizontal transverse force (such as a front-rear direction) of the sensor, fy is a force component along the Y-axis and is defined as a horizontal longitudinal force (such as a left-right direction) of the sensor, fz is a force component along the Z-axis and is defined as a vertical force (such as a top-bottom direction) of the sensor, mx is a moment component around the X-axis and represents a rotation action (such as torsion) of the object around the transverse axis of the sensor, my is a moment component around the Y-axis and represents a rotation action (such as inclination) of the object around the longitudinal axis, and Mz is a moment component around the Z-axis and represents a rotation action (such as rotation) of the object around the vertical axis.
Example five
On the basis of the second embodiment, the third embodiment or the fourth embodiment, the fatigue state recognition method of the fatigue state recognition model comprises the steps of classifying fatigue grades according to joint angle offset and force position characteristics and generating a label.
Model performance was assessed by cross-validation and fatigue status was output.
The fatigue state recognition model comprises a feature extraction module and a random forest algorithm comprising cross verification, wherein the feature extraction module comprises myoelectricity feature extraction, muscle morphological feature extraction and force position information extraction.
Random forests are an integrated learning method, and classification tasks are performed by integrating a plurality of decision trees. For a dataset containing K samples, each sample is assumed to contain S features (characterized by myoelectric features and muscle morphology features described above). K samples are randomly extracted from the data set (repetition is allowed) to form a training subset of a single decision tree, the training subset is repeatedly sampled for T times, T independent data subsets are generated, and the corresponding random forest algorithm comprises T decision trees.
Each decision tree randomly selects K candidate features (K < K) from the K features as each node splits. All possible split point indices are calculated using a binary number based on the selected k features, and the optimal split point is selected according to a minimization error principle.
And (3) independently carrying out the steps on all T decision trees in the random forest, outputting a fatigue state label to a sample by each decision tree, and determining the output of a final random forest algorithm by majority voting of all decision trees according to a final result.
More specifically, in the present embodiment, the jump height decrease rate is not less than 20% according to the joint angle offset, such as not less than 15% of knee joint angle attenuation, and the force level characteristics, and the fatigue grades include no fatigue, mild fatigue, and severe fatigue. More specifically, the change to the joint angle includes calculating the ratio of the joint angle to the base line in different fatigue states with the joint angle without fatigue as the base line, respectively, and considering moderate fatigue if the change exceeds 5%, and considering severe fatigue if the change exceeds 10%. The force level characteristics are the same, the ground acting force and jump height change exceeding 2% is regarded as moderate fatigue, and the change exceeding 5% is regarded as severe fatigue.
Specifically, the dataset was divided into 5 subsets (called "folds"), with 4 folds as training set and 1 fold as validation set in sequence, and finally the model generalization ability was evaluated by the average performance of 5 experiments.
Specifically, the recognition performance of the model is visualized by adopting an confusion matrix, each column of the confusion matrix represents a prediction category, the total number of each column represents the number of data predicted to be the category, each row represents the real attribution category of the data, the total number of data in each row represents the number of data instances of the category, the numerical value in each column represents the number of the real data predicted to be the category, the importance of the multi-modal features is ranked through OOB importance, the importance of the features is quantified through replacement or noise interference feature values, the change of the model performance is observed, about 36.8% of samples in the training set of each decision tree are not selected, the samples are used as verification sets for evaluating the feature importance, after random replacement is carried out on each feature, the difference of prediction errors of the model on the samples outside the bag before and after the replacement is calculated, and the larger difference indicates that the feature is more important for the model.
Specifically, the fatigue state is that the fatigue state is identified for the test set according to the trained model, and the model outputs fatigue labels corresponding to different fatigue states.
Example six
On the basis of any one of the second embodiment to the fifth embodiment, the lower limb muscle fatigue state recognition system based on multi-modal feature fusion comprises a fatigue state recognition model and a multi-modal sensor for providing bioelectric signals, muscle morphology information and force position information for the fatigue state recognition model.
The multi-mode sensor comprises a bioelectric signal sensor, a muscle form collector and a force position information collection device, wherein the bioelectric signal sensor, the muscle form collector and the force position information collection device are respectively connected with a fatigue state recognition model through a wireless myoelectricity collection module, an infrared optical motion capture system and a force measuring table system, and the method is used for realizing the lower limb muscle fatigue state recognition method based on multi-mode feature fusion. And acquiring original myoelectric signals of the target muscles in real time through a wireless myoelectric acquisition module. And capturing three-dimensional coordinates of Mark points by adopting an infrared optical motion capturing system, and dynamically calculating the radius of a muscle section circle, the surface stretching length and the triangular area change of the Mark points. The force measuring table data and the motion capture system realize time synchronization through hardware triggering.
Working principle:
The invention provides a lower limb muscle fatigue state identification method and system based on multi-mode feature fusion, which have the characteristics of high detection precision, quick dynamic response and strong anti-interference performance.
When the lower limb muscle fatigue state identification method and system based on multi-mode feature fusion are applied to lower limbs, provided by the invention:
according to the main fatigue muscle position of lower limb, myoelectricity electrodes are stuck on the surfaces of bilateral rectus femoris, according to the requirements of muscle section and muscle surface strain, motion capturing mark points are stuck on the surfaces of rectus femoris transversely and equidistantly, and according to the requirements of reverse jump, mark points are stuck on each part of lower limb.
The method comprises the steps of collecting electromyographic signals of target muscles according to electromyographic sensors, collecting three-dimensional coordinates of marking points pasted on the target muscles in real time according to a motion capture system, and collecting reverse jump ground reaction forces under different fatigue states according to a force measuring table.
According to the three-dimensional coordinates and the three-dimensional force table, extracting muscle morphological characteristics including a section circle radius and a muscle surface stretching length, geometrical characteristics including a mark point triangular area and a joint angle, and force position information including a reverse jump ground reaction force and a reverse jump maximum height, so as to form a multi-modal feature vector.
According to the multi-modal feature vector, a muscle fatigue recognition model is designed based on a random forest, wherein myoelectricity features, muscle morphological features and triangular areas of marking points are used as model inputs, joint angle and force position information are used as indexes of fatigue states, and different fatigue states are verified and marked.
The above specific embodiments are specific support for the solution idea provided by the present invention, and are not limited to the scope of the present invention, and any equivalent changes or equivalent modifications made on the basis of the technical solution according to the technical idea provided by the present invention still belong to the scope of the technical solution protection of the present invention.

Claims (10)

CN202510625199.5A2025-05-152025-05-15Lower limb muscle fatigue state identification method and system based on multi-modal feature fusionActiveCN120144985B (en)

Priority Applications (1)

Application NumberPriority DateFiling DateTitle
CN202510625199.5ACN120144985B (en)2025-05-152025-05-15Lower limb muscle fatigue state identification method and system based on multi-modal feature fusion

Applications Claiming Priority (1)

Application NumberPriority DateFiling DateTitle
CN202510625199.5ACN120144985B (en)2025-05-152025-05-15Lower limb muscle fatigue state identification method and system based on multi-modal feature fusion

Publications (2)

Publication NumberPublication Date
CN120144985Atrue CN120144985A (en)2025-06-13
CN120144985B CN120144985B (en)2025-08-01

Family

ID=95949398

Family Applications (1)

Application NumberTitlePriority DateFiling Date
CN202510625199.5AActiveCN120144985B (en)2025-05-152025-05-15Lower limb muscle fatigue state identification method and system based on multi-modal feature fusion

Country Status (1)

CountryLink
CN (1)CN120144985B (en)

Citations (5)

* Cited by examiner, † Cited by third party
Publication numberPriority datePublication dateAssigneeTitle
JP2007236663A (en)*2006-03-092007-09-20Shigeki Toyama Muscle fatigue evaluation method, muscle fatigue level evaluation apparatus, and exercise support system that reflects a user's physiological situation in real time
CN111973183A (en)*2019-05-212020-11-24中国科学院深圳先进技术研究院Joint measurement device and method for muscle fatigue and artificial limb
CN114732424A (en)*2022-04-292022-07-12苏州大学Method for extracting complex network attribute of muscle fatigue state based on surface electromyographic signal
CN118266951A (en)*2024-04-022024-07-02齐鲁工业大学(山东省科学院)Artificial intelligence sEMG signal feature extraction and muscle fatigue level discrimination method
CN118986339A (en)*2024-08-232024-11-22天津大学Motion assessment system and method based on multidimensional data analysis

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication numberPriority datePublication dateAssigneeTitle
JP2007236663A (en)*2006-03-092007-09-20Shigeki Toyama Muscle fatigue evaluation method, muscle fatigue level evaluation apparatus, and exercise support system that reflects a user's physiological situation in real time
CN111973183A (en)*2019-05-212020-11-24中国科学院深圳先进技术研究院Joint measurement device and method for muscle fatigue and artificial limb
CN114732424A (en)*2022-04-292022-07-12苏州大学Method for extracting complex network attribute of muscle fatigue state based on surface electromyographic signal
CN118266951A (en)*2024-04-022024-07-02齐鲁工业大学(山东省科学院)Artificial intelligence sEMG signal feature extraction and muscle fatigue level discrimination method
CN118986339A (en)*2024-08-232024-11-22天津大学Motion assessment system and method based on multidimensional data analysis

Also Published As

Publication numberPublication date
CN120144985B (en)2025-08-01

Similar Documents

PublicationPublication DateTitle
CN105877766B (en) A mental state detection system and method based on multi-physiological signal fusion
CN105769173B (en)A kind of cardioelectric monitor system with electrocardiosignal noise removal function
Radmand et al.A characterization of the effect of limb position on EMG features to guide the development of effective prosthetic control schemes
CN109259739A (en)A kind of myoelectricity estimation method of wrist joint motoring torque
CN109009586B (en) A method for continuous decoding of electromyography based on the natural driving angle of artificial wrist joint
Pancholi et al.Intelligent upper-limb prosthetic control (iULP) with novel feature extraction method for pattern recognition using EMG
KR101307515B1 (en)Apparatus for sensing bio-signal and method thereof
CN105534517A (en)Method for removing vehicle motion noise in three-lead electrocardiosignal
Hu et al.A novel fusion strategy for locomotion activity recognition based on multimodal signals
KR100994408B1 (en) Finger force estimation method and estimation device, muscle discrimination method and muscle determination device for finger force estimation
CN117079816A (en)System and method for assessing physical function development of children
Bayat et al.Human gait recognition using bag of words feature representation method
CN120144985B (en)Lower limb muscle fatigue state identification method and system based on multi-modal feature fusion
CN113033501A (en)Human body classification method and device based on joint quaternion
Ren et al.PDCHAR: human activity recognition via multi-sensor wearable networks using two-channel convolutional neural networks
CN120036792A (en)Electrocardiogram monitoring and data analysis system based on artificial intelligence
CN116115239A (en)Embarrassing working gesture recognition method for construction workers based on multi-mode data fusion
KR102603983B1 (en)Deep learning based system and method for recognizing behavior type using emg signal
Qakhkharov et al.Analysis of methods and algorithms for feature extraction of biosignals of muscle activity
CN111110268A (en)Human body muscle sound signal prediction method based on random vector function connection network technology
CN116807497A (en)Myoelectricity mode identification method based on IMU and sEMG feature image fusion
Akhil et al.Lower limb EMG signal analysis using scattering transform and support vector machine for various walking conditions
CN113133765A (en)Multi-channel fusion slight negative expression detection method and device for flexible electronics
CN110134242B (en) A method and system for arm stiffness identification based on surface EMG signals
VeerSpectral and mathematical evaluation of electromyography signals for clinical use

Legal Events

DateCodeTitleDescription
PB01Publication
PB01Publication
SE01Entry into force of request for substantive examination
SE01Entry into force of request for substantive examination
GR01Patent grant
GR01Patent grant

[8]ページ先頭

©2009-2025 Movatter.jp