CN109009586B - Myoelectric continuous decoding method for man-machine natural driving angle of artificial wrist joint - Google Patents

Myoelectric continuous decoding method for man-machine natural driving angle of artificial wrist joint Download PDF

Info

Publication number
CN109009586B
CN109009586B CN201810664049.5A CN201810664049A CN109009586B CN 109009586 B CN109009586 B CN 109009586B CN 201810664049 A CN201810664049 A CN 201810664049A CN 109009586 B CN109009586 B CN 109009586B
Authority
CN
China
Prior art keywords
angle
wrist joint
wrist
joint
axis
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.)
Active
Application number
CN201810664049.5A
Other languages
Chinese (zh)
Other versions
CN109009586A (en
Inventor
张小栋
孙晓峰
陆竹风
李瀚哲
李睿
郭健
杨昆才
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.)
Xian Jiaotong University
Original Assignee
Xian Jiaotong 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 Xian Jiaotong University filed Critical Xian Jiaotong University
Priority to CN201810664049.5A priority Critical patent/CN109009586B/en
Publication of CN109009586A publication Critical patent/CN109009586A/en
Application granted granted Critical
Publication of CN109009586B publication Critical patent/CN109009586B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61F—FILTERS IMPLANTABLE INTO BLOOD VESSELS; PROSTHESES; DEVICES PROVIDING PATENCY TO, OR PREVENTING COLLAPSING OF, TUBULAR STRUCTURES OF THE BODY, e.g. STENTS; ORTHOPAEDIC, NURSING OR CONTRACEPTIVE DEVICES; FOMENTATION; TREATMENT OR PROTECTION OF EYES OR EARS; BANDAGES, DRESSINGS OR ABSORBENT PADS; FIRST-AID KITS
    • A61F2/00—Filters implantable into blood vessels; Prostheses, i.e. artificial substitutes or replacements for parts of the body; Appliances for connecting them with the body; Devices providing patency to, or preventing collapsing of, tubular structures of the body, e.g. stents
    • A61F2/50—Prostheses not implantable in the body
    • A61F2/68—Operating or control means
    • A61F2/70—Operating or control means electrical
    • A61F2/72—Bioelectric control, e.g. myoelectric
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
    • A61B5/316—Modalities, i.e. specific diagnostic methods
    • A61B5/389—Electromyography [EMG]
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7235—Details of waveform analysis
    • A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems

Landscapes

  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Engineering & Computer Science (AREA)
  • Public Health (AREA)
  • Biomedical Technology (AREA)
  • Veterinary Medicine (AREA)
  • Physics & Mathematics (AREA)
  • General Health & Medical Sciences (AREA)
  • Animal Behavior & Ethology (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Medical Informatics (AREA)
  • Molecular Biology (AREA)
  • Biophysics (AREA)
  • Pathology (AREA)
  • Artificial Intelligence (AREA)
  • Surgery (AREA)
  • Physiology (AREA)
  • Cardiology (AREA)
  • Psychiatry (AREA)
  • Signal Processing (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Computation (AREA)
  • Mathematical Physics (AREA)
  • Fuzzy Systems (AREA)
  • Oral & Maxillofacial Surgery (AREA)
  • Transplantation (AREA)
  • Vascular Medicine (AREA)
  • Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
  • Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
  • Prostheses (AREA)

Abstract

The invention discloses a myoelectricity continuous decoding method for a man-machine natural driving angle of a prosthetic wrist joint, which enables a disabled person to imagine that a missing hand and a healthy hand of the disabled person do wrist joint movement with the same angle, and collects surface myoelectricity signals of the forearm of the disabled person and the movement angle of a healthy side wrist joint so as to establish a surface myoelectricity signal continuous decoding model for the man-machine natural driving angle of the prosthetic wrist joint. The invention has simple operation and high accuracy. The three-dimensional motion capture system can simultaneously realize high resolution and high capture frequency, and the joint angle calculated by the three-dimensional motion capture system has high accuracy; compared with the traditional wearable angle sensor, the wearable angle sensor has the advantages that compression interference on surface muscle electrical signals is avoided; the three-dimensional motion capture device is provided with an interface which is synchronously collected with an electromyograph, and can realize simultaneous collection of three-dimensional motion capture angles and surface electromyographic signals. In addition, the sampling frequency of the electromyograph is 2048Hz, and the change condition of the surface electromyogram signal can be collected in real time.

Description

Myoelectric continuous decoding method for man-machine natural driving angle of artificial wrist joint
Technical Field
The invention belongs to the technical field of intelligent artificial hand and raw electro-mechanical integration, and relates to a myoelectric continuous decoding method for a man-machine natural driving angle of a wrist joint of an artificial hand.
Background
According to the latest statistical data of Chinese couplets, the number of people with physical disabilities is about 2472 million in China, and the number of patients with hand disabilities reaches ten million. The loss of the hand function not only affects the life and work of the disabled, but also brings heavy impact to the mind of the disabled. The traditional artificial hand only has a decoration function and cannot meet the daily life needs. The intelligent artificial hand overcomes the defects of the traditional artificial hand, wherein the myoelectricity control artificial hand becomes a research hotspot due to convenient wearing, accurate control and strong function.
The human body surface electromyographic signal is a bioelectricity signal, can objectively reflect the motion state of a human body and can be generated in advance of actual motion, has predictability, and can realize the perception of human body motion intention. Most of the existing myoelectric artificial hands focus on the research on the utilization of surface myoelectric signals to identify human hand action classification so as to realize the grasping action prediction of the artificial hand, however, in the motion process of the artificial hand, the driving angle of a wrist joint greatly determines the flexibility of the artificial hand operation, so that in order to realize better personification of the artificial hand, the decoding of the human-computer natural driving angle of the artificial wrist joint by utilizing the surface myoelectric signals remained on the arm is extremely important. Based on the above, it is a key point of current research to continuously decode the wrist joint movement angle by using the arm surface electromyogram signal so as to provide a suitable artificial wrist joint man-machine natural driving angle.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provide a myoelectricity continuous decoding method for a man-machine natural driving angle of a artificial wrist joint, which enables a disabled person to imagine that the missing hand and the healthy hand do wrist joint movement with the same angle, and collects surface myoelectricity signals of the disabled person residual limb forearm and the healthy side wrist joint movement angle, thereby establishing a man-machine natural driving angle model for the artificial wrist joint through continuous decoding of the surface myoelectricity signals. The artificial wrist joint man-machine natural driving angle is continuously decoded by using the electromyographic signals on the surface of the forearm of the disabled limb, so that the requirements of the hand disabled patients on daily life are met.
In order to achieve the purpose, the invention adopts the following technical scheme to realize the purpose:
a method for continuously decoding man-machine natural driving angles of artificial wrist joints by using human arm surface myoelectric signals comprises the following steps:
the first step is as follows: the motion capture system is used for recording the three-dimensional coordinates of the movement of the wrist joint at the healthy side, and the wrist bending/stretching angle is calculated by a human wrist joint kinematics modeling method.
In the method, a reasonable wrist joint motion capture scheme is selected, a local coordinate system of the wrist joint is established, a wrist joint motion model is established, and the angle change of the wrist joint in the bending and stretching process is calculated by using a kinematics method. The final formula of the wrist joint flexion and extension angle is as follows:
θ＝arccos(T·i2·i1)
where T is the transformation matrix from the local coordinate system of the wrist joint to the local coordinate system of the elbow joint, i1、i2Respectively are unit vectors in sagittal axis directions of the elbow joint and the wrist joint.
The second step is that: and synchronously acquiring surface electromyographic signals of the six muscles on the residual side of the forearm by using an electromyography acquisition instrument. The six residual muscles are extensor carpi radialis longus, flexor carpi radialis longus, extensor carpi ulnaris, flexor carpi ulnaris, extensor digitorum communis and flexor digitorum superficialis respectively.
The third step: and preprocessing and extracting the characteristics of the collected six-channel surface electromyographic signals. Because the sampling frequency of the surface electromyograph is 2048Hz and the sampling frequency of the three-dimensional motion capture system is 100Hz, the characteristic value of the surface electromyography signal is resampled, and the surface electromyography signal and the wrist joint kinematics data have the same sampling frequency.
The fourth step: and establishing a BP neural network by adopting a machine learning method to realize the continuous decoding of the bending/stretching angle of the wrist joint by the myoelectric signals on the surface of the arm. Firstly, setting network parameters of a BP neural network, secondly training the BP neural network, and finally testing the BP neural network.
In the method, a three-layer BP neural network is constructed, the myoelectricity activity intensity characteristic value of a surface myoelectricity signal is extracted to be used as network input, a joint angle calculated by a wrist joint kinematics model is used as network output, 10 neurons are arranged in the middle layer, and each neuron adopts a Sigmoid action function.
The fifth step: collecting the surface electromyogram signal of the forearm of the disabled side, and preprocessing and extracting the characteristic of the collected surface electromyogram signal. Inputting the characteristic value of the myoelectric activity intensity into a wrist joint angle continuous decoding model, outputting continuously changed wrist joint motion angles, calculating a linear correlation coefficient between a network prediction joint angle and a joint angle calculated by kinematics, and judging the accuracy of the human-computer natural driving angle of the artificial wrist joint continuously decoded by the myoelectric signal on the surface of the human arm.
Compared with the prior art, the invention has the following beneficial effects:
the invention has simple operation and high accuracy. The three-dimensional motion capture system can simultaneously realize high resolution and high capture frequency, and the joint angle calculated by the three-dimensional motion capture system has high accuracy; compared with the traditional wearable angle sensor, the wearable angle sensor has the advantages that compression interference on surface muscle electrical signals is avoided; the three-dimensional motion capture device is provided with an interface which is synchronously collected with an electromyograph, and can realize simultaneous collection of three-dimensional motion capture angles and surface electromyographic signals. In addition, the sampling frequency of the electromyograph is 2048Hz, and the change condition of the surface electromyogram signal can be collected in real time.
The invention establishes a prediction model for continuously decoding the driving angle of the artificial wrist joint by the surface electromyogram signal. The stretching and contraction of skeletal muscles drive the wrist joint to move, and the corresponding surface electromyographic signal amplitude changes differently in the muscle contraction process, so that the man-machine natural driving angle of the artificial wrist joint can be continuously decoded by utilizing the myoelectricity active strength of the surface electromyographic signal. The disabled can imagine that the wrist joints of the missing hand and the healthy hand synchronously perform the same movement, and after the movement imagination training, the movement angle of the healthy side wrist joint is used as the man-machine natural driving angle of the wrist joint of the disabled hand. The model input is the surface electromyographic signals of the forearm of the disabled, individual differences caused by modeling from a healthy person to a disabled person are avoided, so that the surface electromyographic signals are continuously decoded to realize the man-machine natural driving angle of the joint of the artificial wrist, and the method has potential application values in the fields of biological medicine, human-machine engineering and the like.
Drawings
FIG. 1 is a block diagram of a method for continuously decoding a man-machine natural driving angle of a joint of a fake wrist by using an electromyographic signal of the surface of an arm;
FIG. 2 is a schematic diagram of the right upper limb marker point position and coordinate system setup; wherein P is1At the elbow joint, P2At the radial position of the forearm, P3At the outer side of the wrist, P4At the medial side of the wrist, P5The middle finger metacarpophalangeal joint of the right hand;
FIG. 3 is a schematic diagram showing the calculation result of the flexion and extension angles of the wrist joint;
FIG. 4 is a characteristic value of myoelectric activity intensity of myoelectric signals on six-channel surfaces of an arm;
FIG. 5 is a basic flowchart of the BP neural network algorithm;
FIG. 6 is a graph showing the result of human-machine natural driving angles of the wrist joints by continuously decoding myoelectric signals on the surface of the arm.
Detailed Description
The invention is described in further detail below with reference to the accompanying drawings:
referring to fig. 1, the invention uses a three-dimensional motion capture system to record kinematic data of the bending and stretching process of the wrist joint at the healthy side, and calculates the bending and stretching angle of the wrist joint; the electromyograph synchronously acquires surface electromyographic signals of the forearm muscle of the disabled side, and the electromyographic active intensity characteristics are obtained through pretreatment and characteristic extraction; training the BP neural network by taking the myoelectric activity intensity characteristics as the input of the BP neural network and the wrist joint bending and stretching angle as the output of the BP neural network, setting an error range, stopping iteration after error conditions are met, and obtaining a stable human-computer natural driving angle model of the arm decoding artificial wrist joint; and finally, inputting the tested activity intensity of the electrical signals of the musculus signals on the surface of the disabled side, and outputting the predicted man-machine natural driving angle of the artificial wrist joint. The specific embodiment is as follows:
the first step is as follows: calculating the bending and stretching angle of the wrist joint according to the kinematic data, wherein the specific method comprises the following steps:
(1) selecting mark positions of wrist joint motion capture mark points and establishing a joint local coordinate system:
fig. 2 shows the sticking scheme of the mark point of the upper limb on the right side, in order to ensure that the mark point is not blocked, the examinee adopts a sitting posture, the big arm and the forearm on the right side are horizontally lifted and kept still, and the back of the hand is bent and extended upwards to form the wrist joint. According to the kinematics principle, each rigid body has 6 degrees of freedom in 3-dimensional space, and the position coordinates of 3 non-collinear points on the rigid body are required to be known for determining the pose of the moving rigid body in the three-dimensional space. Therefore, a mark point is set at the elbow joint, and is marked as P1The radial side of the forearm is provided with a mark point marked as P2A mark point is respectively arranged at the outer side and the inner side of the wrist and is marked as P3And P4Then, a mark point is arranged at the middle finger metacarpophalangeal joint of the right hand and is marked as P5。
The local coordinate system of the elbow joint is shown in figure 2:
the local coordinate origin of the elbow joint is P1，P1And P2The line is the x-axis and the direction is towards P2(ii) a From P1、P2、P3The normal line of the plane formed by the three points is the y axis, and the direction points to the inner side of the body; as can be seen from the right-hand rule, the normal vector of the plane formed by the x-axis and the y-axis is the z-axis, and the direction is upward. The unit vector calculation formulas of the x axis, the y axis and the z axis are respectively as follows:
Figure BDA0001707335980000051
Figure BDA0001707335980000052
k1＝i1×j1
the local coordinate system of the wrist joint is shown in figure 2:
the origin of local coordinates of the wrist joint is P3And P4Midpoint of line, origin and P5The line is the x-axis and the direction is towards P5(ii) a From P3、P4、P5The normal vector of the plane formed by the three points is the z axis, and the direction is downward; as can be seen from the right-hand rule, the normal vector of the plane formed by the x-axis and the z-axis is the y-axis, and the direction points to the inside of the body. The unit vector calculation formula of the x, y and z axes is as follows:
Figure BDA0001707335980000053
Figure BDA0001707335980000061
k2＝i2×j2
(2) calculating the wrist joint motion angle:
and solving the bending/stretching angle of the wrist joint in the human body sagittal plane on the basis of the elbow joint and wrist joint local coordinate system, and expressing the bending/stretching angle of the wrist by theta.
θ＝arccos(T·i2·i1)
Where T is the transformation matrix from the local coordinate system of the wrist joint to the local coordinate system of the elbow joint, i1、i2Respectively are unit vectors in sagittal axis directions of the elbow joint and the wrist joint.
Figure 3 shows the wrist flexion/extension angle change calculated by this method. When the palm hand flat position is defined to be 0 degree, the bending angle is positive, the stretching angle is negative, and the bending and stretching motion range of the wrist joint is-60-75 degrees, so that the wrist bending/stretching angle calculated by the method is reasonable and in the bending/stretching motion range of the hand.
The second step is that: and synchronously acquiring surface electromyographic signals of six related muscles of the forearm stump side by using an electromyograph. The specific method comprises the following steps:
analyzing the forearm muscles of the human hand, the extensor carpi radialis longus, extensor carpi ulnaris longus and extensor digitorum communis are closely related to the stretching action of the wrist, and the flexor carpi radialis, flexor carpi ulnaris and flexor digitorum superficialis are closely related to the bending action of the wrist, so that the six muscles are selected as surface electromyographic signal sources. Firstly, disinfecting the forearm of a subject by using alcohol, and reducing the interference of skin surface grease on surface myoelectric signals; secondly, respectively sticking the differential electrodes on the surfaces of six muscles along the muscle fiber direction, and arranging the zero point of the electromyographic signal at the elbow joint; and finally, realizing synchronous acquisition of the electromyograph and the three-dimensional motion capture system by using a trigger.
The third step: preprocessing the collected six-channel arm surface electromyographic signals, extracting the characteristic of the electromyographic activity intensity, finally resampling the active intensity of the arm surface electromyographic signals, and unifying the sampling frequency of the surface electromyographic signals and the sampling frequency of the wrist joint angle value. The specific method comprises the following steps:
(1) preprocessing an arm surface electromyographic signal:
the collected arm surface electromyographic signals are usually mixed with noises and interferences, including inherent equipment noises, ambient noise interferences, 50Hz power frequency interferences, artifact noises and the like. The effective frequency range of the surface electromyogram signal is 20-500Hz, in order to reduce the interference of noise signals, a 4-order Butterworth filter is adopted to carry out 20-500Hz band-pass filtering on the surface electromyogram signal, and then notch filtering is utilized to remove 50Hz power frequency interference.
(2) Extracting the electromyographic signal characteristics of the surface of the arm:
in order to continuously decode the man-machine natural driving angle of the artificial wrist joint by the myoelectric signal on the surface of the arm, the relation between the myoelectric signal strength on the surface of the arm and the man-machine natural driving joint angle of the artificial hand needs to be found, and the myoelectric active strength characteristic of the myoelectric signal on the surface can be extracted. And (3) performing full-wave finishing on the filtered surface electromyogram signal by adopting an envelope curve method, performing low-pass filtering, and selecting a cutoff frequency of 4-10 Hz to obtain the peak value change of the surface electromyogram signal, as shown in fig. 4.
The fourth step: a machine learning method is adopted to establish a BP neural network, so that the continuous decoding of the myoelectric signals of the surface of the arm on the man-machine natural driving angle of the joint of the artificial wrist is realized. The specific method comprises the following steps:
(1) construction of a BP neural network:
in general, a three-layer BP neural network can solve most pattern recognition problems, so the three-layer BP neural network is constructed here. The input layer is a surface myoelectricity intensity characteristic value, wherein b is 6, namely the input layer has 6 neuron nodes; the output layer is at a wrist bending and stretching angle, namely the output layer is provided with 1 neuron node; the number of intermediate layer neuron nodes is determined by the following empirical disclosure:
Figure BDA0001707335980000071
wherein g is the number of intermediate layer nodes, b is the number of input layer nodes, c is the number of output layer nodes, and d is 1-10 as an adjusting constant. Where d is 8, i.e. g is 10.
(2) Training of the BP neural network:
in fig. 5, let X be [ X ] as the network input vector1x2x3x4x5x6]TThe network output vector is Y ═ Y]TThe neuron output of the intermediate layer network is:
Figure BDA0001707335980000081
Figure BDA0001707335980000082
the output of the output layer is:
Figure BDA0001707335980000083
Figure BDA0001707335980000084
Figure BDA0001707335980000085
wherein the neuron action function is:
Figure BDA0001707335980000086
defining an error function:
Figure BDA0001707335980000087
wherein the content of the first and second substances,
Figure BDA0001707335980000088
Figure BDA0001707335980000089
n is the number of samples, m is the number of samples in each sample, dpiFor the wrist flexion-extension angle, y, calculated from kinematic datapiThe wrist joint flexion and extension angle estimated by the BP neural network, and q is the number of network layers.
And searching the local minimum value of E by using a gradient descent method, wherein each connection weight needs to be corrected along the reverse direction of the derivative of the connection weight by the E. If the error function is in the ideal range, stopping iteration, otherwise, continuing to correct the connection weight until the error is small enough.
(3) Testing of the BP neural network:
and inputting 3 groups of arm surface electromyographic signal characteristic values to the trained BP neural network, and outputting 3 groups of predicted values of wrist bending and stretching angles. And calculating a correlation coefficient between the wrist bending and stretching angle input by the BP neural network and an angle obtained by calculating real kinematic data captured by the three-dimensional motion capture system so as to reflect the linear correlation degree between the wrist bending and stretching angle and the real kinematic data. The correlation coefficient calculation formula is as follows:
Figure BDA0001707335980000091
wherein Cov (X, Y) is the covariance of X and Y, Var [ X ]]And Var [ Y]The variances of X and Y, respectively. Correlation coefficient | ρxy|≤1，|ρxyA closer | to 1 indicates a higher degree of correlation of X with Y, and | ρxyA closer | to 0 indicates a lower degree of correlation of X with Y.
The decoding result of extracting the myoelectric activity intensity characteristic of the myoelectric signal on the surface of the arm and predicting the man-machine natural driving angle of the artificial wrist joint by adopting a BP neural network method is shown in figure 6. As can be seen from the figure, the method can stably and effectively realize the continuous decoding of the wrist joint bending and stretching angle of the myoelectric signals on the surface of the arm, and can be used for the man-machine natural driving control of the myoelectric intelligent artificial hand.
The invention relates to a method for continuously decoding a man-machine natural driving angle of a fake wrist joint by using an arm surface electromyographic signal, which is used for identifying the bending and stretching driving angle of the fake wrist joint through amplitude change of the forearm surface electromyographic signal and increasing anthropomorphic natural operation of an intelligent artificial hand. The method extracts surface electromyographic signals of six muscles of extensor carpi radialis longus, flexor carpi radialis, extensor carpi ulnaris, flexor carpi ulnaris, extensor digitorum communis and flexor digitorum superficialis in the process of bending and stretching the wrist, combines an angle change value in the process of bending and stretching the wrist joint, and establishes a nonlinear mapping model by utilizing a BP (back propagation) neural network, so that the aims of continuously decoding the man-machine natural driving angle of the artificial wrist joint by the electromyographic signals of the surface of the arm and flexibly controlling the action of the artificial wrist joint are fulfilled. The model established by the invention is reliable, the angle prediction accuracy is high, the intelligent artificial hand is favorably used by the hand disabled patients to carry out natural operation better, the functional requirements of the artificial hand in life and work are met, and the method has good social benefit and economic benefit.
The above-mentioned contents are only for illustrating the technical idea of the present invention, and the protection scope of the present invention is not limited thereby, and any modification made on the basis of the technical idea of the present invention falls within the protection scope of the claims of the present invention.

Claims (4)

1. A myoelectric continuous decoding method for man-machine natural driving angles of artificial wrist joints is characterized by comprising the following steps:
step 1: the motion capture system is used for recording the three-dimensional coordinates of the movement of the wrist joint at the exercise side, and the wrist bending/stretching angle is calculated by a wrist joint kinematics modeling method, wherein the method comprises the following specific steps:
1-1) selecting mark positions of wrist joint motion capture mark points and establishing a joint local coordinate system:
setting a mark point at the elbow joint, and marking the mark point as P1The radial side of the forearm is provided with a mark point marked as P2A mark point is respectively arranged at the outer side and the inner side of the wrist and is marked as P3And P4Then, a mark point is arranged at the middle finger metacarpophalangeal joint of the right hand and is marked as P5；
The local coordinate origin of the elbow joint is P1，P1And P2The line is the x-axis and the direction is towards P2(ii) a From P1、P2、P3The normal line of the plane formed by the three points is the y axis, and the direction points to the inner side of the body; the vector is obtained by a right hand rule, a normal vector of a plane formed by the x axis and the y axis is a z axis, and the direction is upward; the unit vector calculation formulas of the x axis, the y axis and the z axis are respectively as follows:
Figure FDA0002448457830000011
Figure FDA0002448457830000012
k1＝i1×j1
the origin of local coordinates of the wrist joint is P3And P4Midpoint of line, origin and P5The line is the x-axis and the direction is towards P5(ii) a From P3、P4、P5The normal vector of the plane formed by the three points is the z axis, and the direction is downward; the right hand rule shows that the normal vector of a plane formed by the x axis and the z axis is the y axis, and the direction points to the inner side of the body; the unit vector calculation formula of the x, y and z axes is as follows:
Figure FDA0002448457830000013
Figure FDA0002448457830000014
k2＝i2×j2
1-2) calculating the wrist joint motion angle:
solving the bending/stretching angle of the wrist joint in the human sagittal plane on the basis of the elbow joint and wrist joint local coordinate system, and expressing the bending/stretching angle of the wrist by theta;
θ＝arccos(T·i2·i1)
where T is the transformation matrix from the local coordinate system of the wrist joint to the local coordinate system of the elbow joint, i1、i2Respectively are unit vectors in sagittal axis directions of the elbow joint and the wrist joint;
step 2: synchronously acquiring surface electromyographic signals of six muscles on the residual side of the forearm by using an electromyographic acquisition instrument; the six residual muscles are extensor carpi radialis longus, flexor carpi radialis longus, extensor carpi ulnaris, flexor carpi ulnaris, extensor digitorum communis and flexor digitorum superficialis respectively;
and step 3: preprocessing and extracting characteristics of the collected six-channel surface electromyographic signals; resampling the surface electromyographic signal characteristic value to realize that the surface electromyographic signal and the wrist joint kinematic data have the same sampling frequency;
and 4, step 4: establishing a BP neural network by adopting a machine learning method to realize the continuous decoding of the wrist joint bending/stretching angle of the arm surface electromyographic signals; firstly, setting network parameters of a BP neural network, secondly training the BP neural network, and finally testing the BP neural network;
and 5: collecting surface electromyographic signals of the forearm of the disabled side, and preprocessing and extracting characteristics of the collected surface electromyographic signals; inputting the characteristic value of the myoelectricity activity intensity into a wrist joint angle continuous decoding model, outputting continuously changed wrist joint motion angles, calculating a linear correlation coefficient between a network prediction joint angle and a joint angle calculated by kinematics, and judging the accuracy of the human-computer natural driving angle of the artificial wrist joint continuously decoded by the myoelectricity signal on the surface of the arm.
2. The myoelectric continuous decoding method of the man-machine natural driving angle of the artificial wrist joint according to claim 1, characterized in that the specific method of step 3 is as follows:
3-1) preprocessing the myoelectric signal of the surface of the arm:
performing 20-500Hz band-pass filtering on the surface myoelectric signal by adopting a 4-order Butterworth filter, and removing 50Hz power frequency interference by utilizing notch filtering;
3-2) extracting the electromyographic signal characteristics of the surface of the arm:
and performing full-wave finishing on the filtered surface electromyogram signal by adopting an envelope curve method, performing low-pass filtering, and selecting a cutoff frequency of 4-10 Hz to obtain the peak value change of the surface electromyogram signal.
3. The method for continuously decoding the myoelectricity of the artificial wrist joint man-machine natural driving angle according to claim 1, wherein in step 4, a three-layer BP neural network is constructed, the myoelectricity activity intensity characteristic value of the surface myoelectricity signal is extracted as the network input, the joint angle calculated by a wrist joint kinematics model is used as the network output, 10 neurons are arranged in the middle layer, and each neuron adopts a Sigmoid action function.
4. The myoelectric continuous decoding method of man-machine natural driving angles of the artificial wrist joint according to claim 1 or 3, characterized in that the specific method of step 4 is as follows:
4-1) construction of BP neural network:
constructing a three-layer BP neural network; the input layer is a surface myoelectricity intensity characteristic value, and b is 6, namely the input layer has 6 neuron nodes; the output layer is at a wrist bending and stretching angle, namely the output layer is provided with 1 neuron node; the number of intermediate layer neuron nodes is determined by the following empirical disclosure:
Figure FDA0002448457830000031
g is the number of intermediate layer nodes, b is the number of input layer nodes, c is the number of output layer nodes, and d is 1-10 and is an adjusting constant; taking d as 8, namely g as 10;
4-2) training of BP neural network:
let the network input vector be X ═ X1x2x3x4x5x6]TThe network output vector is Y ═ Y]TThe neuron output of the intermediate layer network is:
Figure FDA0002448457830000041
Figure FDA0002448457830000042
the output of the output layer is:
Figure FDA0002448457830000043
Figure FDA0002448457830000044
Figure FDA0002448457830000045
wherein the neuron action function is:
Figure FDA0002448457830000046
defining an error function:
Figure FDA0002448457830000047
wherein the content of the first and second substances,
Figure FDA0002448457830000048
Figure FDA0002448457830000049
n is the number of samples, m is the number of samples in each sample, dpiFor the wrist flexion-extension angle, y, calculated from kinematic datapiThe wrist joint bending and stretching angle estimated by the BP neural network, and q is the number of network layers;
searching a local minimum value of E by using a gradient descent method, wherein each connection weight needs to be corrected along the reverse direction of the derivative of the connection weight of E; if the error function is in the ideal range, stopping iteration, otherwise, continuously correcting the connection weight until the error is small enough;
4-3) testing of BP neural network:
inputting 3 groups of arm surface electromyographic signal characteristic values to a trained BP neural network, and outputting 3 groups of predicted values of wrist bending and stretching angles; calculating a correlation coefficient between a wrist bending and stretching angle input by the BP neural network and an angle obtained by real kinematics data calculation captured by a three-dimensional motion capture system to reflect the linear correlation degree between the wrist bending and stretching angle and the angle; the correlation coefficient calculation formula is as follows:
Figure FDA0002448457830000051
wherein Cov (X, Y) is the covariance of X and Y, Var [ X ]]And Var [ Y]The variances of X and Y, respectively; correlation coefficient | ρxy|≤1，|ρxyA closer | to 1 indicates a higher degree of correlation of X with Y, and | ρxyA closer | to 0 indicates a lower degree of correlation of X with Y.
CN201810664049.5A 2018-06-25 2018-06-25 Myoelectric continuous decoding method for man-machine natural driving angle of artificial wrist joint Active CN109009586B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810664049.5A CN109009586B (en) 2018-06-25 2018-06-25 Myoelectric continuous decoding method for man-machine natural driving angle of artificial wrist joint

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810664049.5A CN109009586B (en) 2018-06-25 2018-06-25 Myoelectric continuous decoding method for man-machine natural driving angle of artificial wrist joint

Publications (2)

Publication Number Publication Date
CN109009586A CN109009586A (en) 2018-12-18
CN109009586B true CN109009586B (en) 2020-07-28

Family

ID=64610702

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810664049.5A Active CN109009586B (en) 2018-06-25 2018-06-25 Myoelectric continuous decoding method for man-machine natural driving angle of artificial wrist joint

Country Status (1)

Country Link
CN (1) CN109009586B (en)

Families Citing this family (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109259739B (en) * 2018-11-16 2020-08-18 西安交通大学 Myoelectricity estimation method of wrist joint movement moment
CN110801226A (en) * 2019-11-01 2020-02-18 西安交通大学 Human knee joint moment testing system method based on surface electromyographic signals and application
CN110827987B (en) * 2019-11-06 2021-03-23 西安交通大学 Myoelectricity continuous prediction method and system for wrist joint torque in multi-grabbing mode
CN111374808A (en) * 2020-03-05 2020-07-07 北京海益同展信息科技有限公司 Artificial limb control method and device, storage medium and electronic equipment
CN111616848B (en) * 2020-06-02 2021-06-08 中国科学技术大学先进技术研究院 Five-degree-of-freedom upper arm prosthesis control system based on FSM
CN111743667A (en) * 2020-06-29 2020-10-09 北京海益同展信息科技有限公司 Artificial limb control method, device, system and storage medium
CN111920416B (en) * 2020-07-13 2024-05-03 张艳 Hand rehabilitation training effect measuring method, storage medium, terminal and system
CN112587242B (en) * 2020-12-11 2023-02-03 山东威高手术机器人有限公司 Master hand simulation method of surgical robot, master hand and application
CN114224577B (en) * 2022-02-24 2022-05-17 深圳市心流科技有限公司 Training method and device for intelligent artificial limb, electronic equipment, intelligent artificial limb and medium
CN117953413A (en) * 2024-03-27 2024-04-30 广东工业大学 Electromyographic signal validity judging method, electromyographic signal validity judging device, electromyographic signal validity judging equipment and storage medium

Family Cites Families (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
GB2163570B (en) * 1984-08-24 1988-05-25 Hanger & Co Ltd J E Artificial hand
KR101371359B1 (en) * 2012-03-20 2014-03-19 한국과학기술연구원 Peripheral Nerve Interface System and Method for Prosthetic Hand Control
CN104665828A (en) * 2013-11-27 2015-06-03 中国科学院深圳先进技术研究院 System and method based on electromyographic signal controlling remote controller
CN105615890B (en) * 2015-12-24 2018-08-10 西安交通大学 Human body lower limbs walking joint angles myoelectricity continuous decoding method
CN105963100B (en) * 2016-04-19 2018-07-17 西安交通大学 By the lower limb rehabilitation robot self-adaptation control method assisted needed for patient motion
CN106236336A (en) * 2016-08-15 2016-12-21 中国科学院重庆绿色智能技术研究院 A kind of myoelectric limb gesture and dynamics control method
CN106923942B (en) * 2017-02-15 2018-08-31 上海术理智能科技有限公司 Upper and lower extremities motion assistant system based on the control of human body electromyography signal

Also Published As

Publication number Publication date
CN109009586A (en) 2018-12-18

Similar Documents

Publication Publication Date Title
CN109009586B (en) Myoelectric continuous decoding method for man-machine natural driving angle of artificial wrist joint
CN107397649B (en) Upper limb exoskeleton movement intention identification method based on radial basis function neural network
CN109259739B (en) Myoelectricity estimation method of wrist joint movement moment
Lei An upper limb movement estimation from electromyography by using BP neural network
Jiang et al. A method of recognizing finger motion using wavelet transform of surface EMG signal
Chen et al. Multiple hand gesture recognition based on surface EMG signal
WO2018233435A1 (en) Multi-dimensional surface electromyographic signal based artificial hand control method based on principal component analysis method
Nasr et al. MuscleNET: mapping electromyography to kinematic and dynamic biomechanical variables by machine learning
CN102499797B (en) Artificial limb control method and system
Ma et al. Design on intelligent perception system for lower limb rehabilitation exoskeleton robot
Balbinot et al. Decoding arm movements by myoelectric signal and artificial neural networks
Liu et al. sEMG-based continuous estimation of knee joint angle using deep learning with convolutional neural network
Tang et al. Continuous estimation of human upper limb joint angles by using PSO-LSTM model
Wojtczak et al. Hand movement recognition based on biosignal analysis
Zeng et al. Feature fusion of sEMG and ultrasound signals in hand gesture recognition
CN111714339A (en) Brain-myoelectricity fusion small-world neural network prediction method for human lower limb movement
Pan et al. Musculoskeletal model for simultaneous and proportional control of 3-DOF hand and wrist movements from EMG signals
KR100994408B1 (en) Method and device for deducting pinch force, method and device for discriminating muscle to deduct pinch force
Pan et al. A reliable multi-user EMG interface based on a generic-musculoskeletal model against loading weight changes
Favieiro et al. Decoding arm movements by myoeletric signals and artificial neural networks
Wang et al. EMG signal classification for myoelectric teleoperating a dexterous robot hand
Ke et al. Deep convolutional spiking neural network based hand gesture recognition
Guo et al. Study on motion recognition for a hand rehabilitation robot based on sEMG signals
Ahmad et al. Classification of surface electromyographic signal using fuzzy logic for prosthesis control application
Yang et al. Comparison of Isometric Force Estimation Methods for Upper Limb Elbow Joints

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
GR01 Patent grant
GR01 Patent grant