CN110675933B - Finger mirror image rehabilitation training system - Google Patents
Finger mirror image rehabilitation training system Download PDFInfo
- Publication number
- CN110675933B CN110675933B CN201910840714.6A CN201910840714A CN110675933B CN 110675933 B CN110675933 B CN 110675933B CN 201910840714 A CN201910840714 A CN 201910840714A CN 110675933 B CN110675933 B CN 110675933B
- Authority
- CN
- China
- Prior art keywords
- module
- finger
- rehabilitation
- output
- domain
- 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
Links
Images
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H20/00—ICT specially adapted for therapies or health-improving plans, e.g. for handling prescriptions, for steering therapy or for monitoring patient compliance
- G16H20/30—ICT specially adapted for therapies or health-improving plans, e.g. for handling prescriptions, for steering therapy or for monitoring patient compliance relating to physical therapies or activities, e.g. physiotherapy, acupressure or exercising
Landscapes
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biophysics (AREA)
- Physical Education & Sports Medicine (AREA)
- Engineering & Computer Science (AREA)
- Epidemiology (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Primary Health Care (AREA)
- Public Health (AREA)
- Manipulator (AREA)
- Prostheses (AREA)
Abstract
The invention discloses a finger mirror image rehabilitation training system which comprises an electromyographic signal acquisition module, a pressure acquisition module, a main control module, a rehabilitation robot module, a human-computer interaction virtual environment module and a rehabilitation evaluation module. The advantages are that: the system of the invention constructs a finger mirror image rehabilitation training system through the electromyographic signal acquisition module, the pressure acquisition module, the main control module, the rehabilitation robot module, the human-computer interaction virtual environment module and the rehabilitation evaluation module, has a virtual rehabilitation interaction scene, can help a patient with finger muscle energy injury to carry out high-efficiency rehabilitation training, and can realize rehabilitation effect evaluation through the pressure acquisition module.
Description
Technical Field
The invention relates to a finger mirror image rehabilitation training system, and belongs to the technical field of automatic control.
Background
The hand motor dysfunction caused by the central nerve injury caused by the cerebral apoplexy operation inevitably brings great inconvenience to the daily life of the patient. Currently, repeated and intensive training is required to stimulate the reorganization of the cortex and strengthen the habitual application of the affected limb to the disabled condition, thereby helping the patient to recover the hand movement function.
Clinical rehabilitation methods for hand patients rely heavily on the one-to-one physical treatment of the patient by the treating physician. The traditional hand rehabilitation training method relies on professional physiotherapists to train the handles of the hands of patients. For the patient, although the assistant training of the doctor is helpful to some extent, the training is not ideal and cannot achieve the ideal rehabilitation effect completely, and the training mode has the following problems: firstly, a therapist can only perform exercise training on one patient at a time, the training efficiency is not high, and due to subjective attitude of the therapist in the operation process, the patient may not be guaranteed to obtain sufficient training intensity, or the training intensity exceeds the tolerance limit of the patient, and the treatment effect often depends on the experience and level of the therapist; secondly, lack of accurate control indexes and recording of training parameters (such as movement speed, trajectory, intensity, etc.) cannot provide reasonable theoretical basis for determining and improving treatment protocols; furthermore, the rehabilitation evaluation index is not objective enough, which is not beneficial to the follow-up deeper research on the rehabilitation mechanism; some intuitive feedback information cannot be provided for the patient in real time, and the patient passively participates in treatment in the whole training process, so that the rehabilitation effect is subjectively influenced.
Disclosure of Invention
The technical problem to be solved by the invention is to overcome the defects of the prior art and provide a finger mirror image rehabilitation training system.
In order to solve the technical problem, the invention provides a finger mirror image rehabilitation training system which comprises an electromyographic signal acquisition module, a pressure acquisition module, a main control module, a rehabilitation robot module, a human-computer interaction virtual environment module and a rehabilitation evaluation module;
the myoelectric signal draws the module and is used for drawing the myoelectric signal of healthy side and sick side and transmits for main control module, main control module is used for according to the myoelectric signal of healthy side and the action of control rehabilitation robot module, pressure acquisition module is used for gathering the finger pressure signal of healthy side and sick side to convert the signal of telecommunication into and transmit for rehabilitation robot module, the finger action of the synchronous healthy side of finger of rehabilitation robot module control sick side and output same effort, main control module controls human-computer interaction environment module simulation simultaneously and shows the motion gesture that the finger corresponds, the signal of telecommunication of pressure acquisition module output still transmits for rehabilitation evaluation module for judge recovered effect.
Furthermore, the electromyographic signal extraction module comprises an annular array type electromyographic sensor and Bluetooth equipment, wherein the annular array type electromyographic sensor is used for collecting the electromyographic signals of the front arm end of the healthy side and the affected side and transmitting the electromyographic signals to the main control module through the Bluetooth equipment. The control muscles of human knuckles are mainly distributed in the front arm part, including extensor muscles of fingers on the back of the hand and flexor muscles of fingers on the front. Therefore, the electromyographic signals extracted by the annular array type electromyographic sensors are optimal, and are transmitted wirelessly by Bluetooth equipment, so that the electromyographic signals are lighter and faster.
The system further comprises a preprocessing module, which is used for extracting the signal characteristics of the electromyographic signals collected by the electromyographic signal extracting module by adopting a time-varying power spectrum method, and performing learning classification by adopting an LSTM model to obtain the electromyographic signals without interference, and transmitting the electromyographic signals to the main control module. Such as: internal noise and external electromagnetic radiation interference from components such as power supplies, communications, torque, sensors, etc.
Further, the processing process of the preprocessing module is as follows:
the electromyographic signals collected by the electromyographic signal extraction module are divided into two groups.
The method is characterized in that the electromyographic signals of the first group are directly processed by a convolutional neural network, the convolutional neural network can convert a group of sampling signals of 8 channels into a group of characteristic quantities, and the method comprises the following steps: constructing a neural network having a structure of convolutional layers and pooling layers of I layers, wherein the convolutional layer of the I layer has NiA convolution kernel, the ith convolution layer generates N when the sampled signal passes through the convolution neural network layer by layeriAnd each eigen mapping matrix is sent to the (i + 1) th convolutional layer after being subjected to pooling layer down-sampling, b convolutional cores are arranged in the convolutional layer in the last layer of the convolutional neural network, the convolutional layer outputs b 2 x 2 eigen mapping matrixes, and then the b 2 x 2 eigen mapping matrixes are down-sampled into b 1 x 1 eigen mapping matrixes by the last pooling layer, namely, the eigen vectors with the length of b.
And extracting signal characteristics of the electromyographic signals of the second group by adopting a time-varying power spectrum method. And sending the second group of extracted signal features and the first group of extracted 1 x b-dimensional vectors into an MLP layer together, and completing splicing of the feature vectors in the MLP layer to obtain 1 x (b +6) -dimensional feature vectors. And inputting the data into a three-layer fully-connected neural network, wherein c neurons are output by the output layer of the network, so that a 1 x c matrix is output, and the obtained matrix is stored as a history characteristic into a first-in first-out queue with the size of 1 x c i. And saving the output from the current moment to i moments before as historical data, inputting the historical data into the LSTM neural network, wherein the output of the LSTM network layer is the recognition result of the whole neural network. The method can effectively remove the following components: internal noise and external electromagnetic radiation interference from components such as power supplies, communications, torque, sensors, etc.
Furthermore, the pressure acquisition module comprises a cylinder-like elastic body and a pressure sensor, and the top of the cylinder-like elastic body is attached with the pressure sensor module. The cylinder-like elastic body is deformed by finger pressure and then reacts to the pressure sensor to generate an electric signal, and the strength of the signal reflects the force application of the finger.
Further, recovered robot module includes driver, control panel and indicates recovered mechanism based on flexible joint ectoskeleton formula, indicate recovered mechanism to include base, drive module, back of the hand supporter, finger based on flexible joint ectoskeleton formula and wear the part, wherein the base is used for placing the forearm to as the container that holds power and driver, drive module is including the motor, gear and the relevant tendon rope that connect gradually for driving finger and wearing the part and realize snatching the motion, finger is worn the part and is five fingers and wears the part, the back of the hand supporter is used for fixed motor and connects finger and dress the part, main control module, pressure acquisition module and driver are connected respectively to the control panel, the driver is connected the motor. In the application, the rehabilitation robot realizes the connection of finger joints through flexible materials, so that the fingers of a patient can be effectively protected from secondary injury; the under-actuated control method based on the rope line is adopted to effectively reduce the control cost.
Furthermore, the rehabilitation robot module processes the electric signal input by the pressure acquisition module by adopting a variable domain fuzzy self-adaptive PD control method, outputs a moment control signal of a corresponding finger motor, and controls the finger on the affected side to output the same acting force as the finger on the healthy side. The finger on the affected side and the finger on the healthy side can perform the same action and output the same acting force.
Further, the processing process by adopting the variable domain fuzzy adaptive PD control method is as follows:
classifying the initial fuzzy domain of the pressure signal difference E between the healthy side and the affected side acquired by the pressure acquisition module, namely [ -E3, -E2, -E1, E0, E1, E2 and E3]Rate of change of pressure signal difference
The initial ambiguity domain is [ -CE3, -CE2, -CE1, CE0, CE1, CE2, CE3]The output variable is the motor of the corresponding finger, the outputThe output torque domain of the variable is denoted as U [ -U ═ U [ - ]min,Umax]Establishing a fuzzy producer
E＝6×(e-Emin)/(Emax-Emin)
Wherein the pressure signal difference E E is Emin,Emax]And rate of change
EminAnd EmaxRespectively representing the minimum and maximum values, CE, of the discourse domain of the signal difference eminAnd CEmaxRespectively representing the rate of change of the signal difference
The minimum value and the maximum value of the discourse domain, E and CE are respectively corresponding outputs of the fuzzy producer; wherein-E3, -E2, -E1, E0, E1, E2, E3 and-CE 3, -CE2, -CE1, CE0, CE1, CE2 and CE3 respectively represent basic symbols of fuzzy set partition of corresponding fuzzy discourse
Fuzzy rules are designed according to the adaptive PD control law as follows:
U＝KpΕ+KdCΕ
Kp＝Kp0+ΔKp×qp
Kd＝Kd0+ΔKd×qd
wherein, Kp，KdIs the final control parameter of the PD controller; kp0，Kd0Setting parameters for the PD controller; Δ KP，ΔKdIs the output quantity of the fuzzy controller; q. q.sp，qdIs a correction factor;
fuzzy elimination of system output is carried out by a membership maximum value method, a membership function peak value of an output fuzzy subset is directly selected as a determination value of output, and a fuzzy subset U is outputiThe logical "and" of (c) is:
finally, the accurate output of the control variable u is reversely solved by taking the median;
gradually decrease e and
the domain of the reduced pressure signal difference e is e (t [ - α [) ]1(t)Emin,α1(t)Emax]Rate of change
Is CE (t) [ - α [ ]2(t)CEmin,α2(t)CEmax]The output domain is U (t) [ - β (t) U [ ]min,β(t)Umax]Wherein α is1(t)、α2(t) and β (t) are respectively the scaling factors of the corresponding discourse domain, wherein;
α1(t)＝(|e|/E)τ+ε1，α2(t)＝(|ec|/EC)τ+ε2，
wherein 1 ≧ tau > 0, ε1And ε2Is a small positive number, taken as epsilon1＝(Emax-Emin)/1000，ε2＝(CEmax-CEmin) A value of/1000, ec representing the rate of change of the pressure difference; kIIs a design scale parameter; e (τ) represents the pressure difference at time τ; β (0) represents the scaling factor for the initial output discourse domain; t is time.
Pressure signal difference e and rate of change
The input domains E (t) and CE (t) are used as the input of domain fuzzy self-adaptive PD control, are used for feeding back the mirror image synchronization effect of the two hands at the key/affected side, and the output domain U (t) is used as the output of a driving motor of a corresponding finger.
Further, the rehabilitation evaluation module evaluates the rehabilitation effect according to the pressure difference delta P of the pressure acquisition module and the time domain characteristic of the characteristic signal difference delta f of the electromyographic sensor, wherein the shorter the time for the delta P and the delta f to reach stability is, the better the finger synchronism is, the better the rehabilitation effect is, and otherwise, the worse the motor ability of the affected side is. The rehabilitation evaluation module can simply, quickly and efficiently obtain an evaluation result.
The invention achieves the following beneficial effects:
the system of the invention constructs a finger mirror image rehabilitation training system through the electromyographic signal acquisition module, the pressure acquisition module, the main control module, the rehabilitation robot module, the human-computer interaction virtual environment module and the rehabilitation evaluation module, has a virtual rehabilitation interaction scene, can help a patient with finger muscle energy injury to carry out high-efficiency rehabilitation training, and can realize rehabilitation effect evaluation through the pressure acquisition module.
Drawings
FIG. 1 is a schematic diagram of the system of the present invention;
FIG. 2 is a schematic diagram of a rehabilitation robot according to the present invention;
FIG. 3 is a flow of action recognition based on the LSTM network model.
Detailed Description
The invention is further described below with reference to the accompanying drawings. The following examples are only for illustrating the technical solutions of the present invention more clearly, and the protection scope of the present invention is not limited thereby.
As shown in fig. 1, a finger mirror image rehabilitation training system comprises a MYO arm ring 1, a grip strength testing bottle 2, a pressure sensor 3, a driver and control panel 4, an exoskeleton-type finger rehabilitation mechanism 5 based on a flexible joint, a master control computer 6 and a human-computer interaction virtual environment 7;
MYO armlet 1 is used for wearing the forearm end at patient's both arms, and MYO armlet 1 surveys patient's muscular activity, connects with the master control computer through low-power bluetooth equipment, captures patient's motion intention. After receiving the arm ring control signal of the healthy side hand 8, the master control computer 6 identifies the finger joint action, sends a control command to the driver and control panel 4 by adopting the CAN bus, and synchronously controls the finger of the affected side hand 9 through the exoskeleton-type finger rehabilitation mechanism based on the flexible joint. And (3) sensing the motion state of the finger at the affected side according to the feedback information of the motor encoder of the exoskeleton-type finger rehabilitation mechanism 5 based on the flexible joint. In the constructed human interactive virtual environment 7, the corresponding motion gestures are displayed. A grip strength feedback system is constructed by a method of sticking the pressure sensor 3 on the top of the grip strength test bottle 2, and the magnitude of the grip strength of the fingers on the healthy and affected sides is monitored in real time and is used for a rehabilitation evaluation system.
As shown in fig. 2, the exoskeleton-type finger rehabilitation mechanism based on flexible joints comprises a base 10, a driving module, a hand back support 11 and a finger wearing part, wherein the base 10 is used for placing a forearm and is used as a container for accommodating a power supply and a driver; the back of the hand support 11 is used for fixing the motor 12 and connecting the finger wearing part; the driving module comprises a motor 12, a gear and a related tendon rope to drive the finger wearing part to realize grabbing motion; the finger wearing part is a five-finger wearing part and is respectively used for wearing on five fingers.
According to the finger rehabilitation action evaluation system, the design comprises: the five fingers are clenched, the five fingers are opened, the thumb is bent to the index finger, the thumb is bent to the middle finger, the thumb is bent to the ring finger/little finger and the natural state are six basic motion states. Since the surface electromyogram signal belongs to a weak signal, it is vulnerable to such as: internal noise from components such as power supplies, communications, torque, sensors, etc., and external electromagnetic radiation interference. Meanwhile, the posture of the arm, muscle fatigue, emotional state, etc. of the patient are all environmental variables that are difficult to grasp for motion recognition.
Therefore, the system adopts a time-varying power spectrum (TD-PSD) method to extract signal features aiming at the surface electromyogram signals of the finger actions of the user, and adopts an LSTM model to carry out learning classification, as shown in figure 3.
Let the sEMG signal be x [ j ], j ═ 1,2, …, N. The sampling frequency is fs Hz, so the square root of the first moment of zero, second and fourth order moments is:
the following 6 signal characteristics are defined:
f1＝log(m0)；f2＝log(m0-m2)；f3＝log(m0-m4)；
these signals have a smaller false recognition rate and a faster processing speed.
In order to accurately identify the surface electromyographic signals with time-varying characteristics, an RNN network model of a long-short term memory network (LSTM) is adopted for sample training. The LSTM module is used to replace the sliding window, and the biggest characteristic of the LSTM module is that the LSTM module can process an infinite time sequence by simulating a human memory mechanism and adaptively determine to memorize or forget certain information. After the signal of each period of time is extracted with the characteristics by the neural network, the LSTM can determine whether to 'forget' the previous memory (enter the next theme/action) or 'remember' the current parameters according to the characteristics, and the characteristics are integrated and sent to the next layer, at the moment, the LSTM controls when to enter the next theme (action), and the problem that the window length of the sliding window method is difficult to determine can be effectively solved.
The input of the neural network is divided into two groups, the first group directly processes the electromyographic signals by adopting the convolutional neural network, and a group of 8-channel sampling signals are output into a group of 1 × b characteristic vectors after passing through the multilayer convolutional layer. The second group consists of 6 basic features obtained by the above TD-PSD method (f)1～f6) And splicing the first group of extracted 1-x-b dimensional feature vectors and then sending the spliced feature vectors and the extracted feature vectors into an MLP layer. And outputting a group of 1 × c historical characteristics of the current moment through a three-layer full-communication neural network, and tracking the i historical moment characteristics to obtain an action recognition result through LSTM network training.
Since the electromyographic signal is generally only used as a motion classification signal, the degree of force applied to the motion is difficult to estimate. Therefore, we have constructed a force feedback system, as shown in fig. 1, and install a pressure sensor above two identical cylinder-like elastic bodies, and when the elastic bodies are deformed by finger pressure, the pressure sensor is responded to, and an electric signal is generated. The strength of the signal reflects the force applied by the finger.
Using the pressure signal difference E E of two pressure sensorsmin,Emax]And rate of change of error
And as an input of the control system, selecting the corresponding control motor according to the action result identified by the MYO. And a fuzzy controller is adopted for control in consideration of the nonlinear characteristic of the pressure signal.
The initial ambiguity domain of the selection force error E is classified into 6 classes, i.e., [ -E3, -E2, -E1, E0, E1, E2, E3]Rate of change of force error
The initial ambiguity domain is [ -CE3, -CE2, -CE1, CE0, CE1, CE2, CE3]The output variables are control motors M1-M4 corresponding to the fingers, and the number of U is [ -U [ ]min,Umax]Is the output torque domain. Establishing fuzzy producer
E＝6×(e-Emin)/(Emax-Emin)
For fuzzy rules, reference is made to the design of the adaptive PD control law, whereupon
U＝KpΕ+KdCΕ
Wherein, KpAnd KdControl parameters for PD control are respectively expressed and adjusted according to the following iterative formula.
Kp＝Kp0+ΔKp×qp
Kd＝Kd0+ΔKd×qd
In the formula, Kp，KdIs the final control parameter of the PD controller; k isp0，Kd0Setting parameters for the PD controller; Δ KP，ΔKdOutput for the fuzzy controller; q. q.sp，qdIs a correction coefficient.
The system selects a membership maximum method to eliminate the system fuzzy. The maximum membership method is a method of selecting an element with the maximum membership from a fuzzy set obtained through reasoning as a definition quantity. And directly selecting the determined value which is the most output by the peak value of the membership function of the output fuzzy subset. Here our output domain is U, the logical "AND" of the output fuzzy subset is:
and finally, the accurate output of the control variable u is reversely solved by taking the median value.
With the progress of the control process, the error and the change rate of the error are continuously reduced, and if the initial discourse domain and the division thereof are continuously used for fuzzy reasoning, the control precision is naturally not high. The reason for this is that the fuzzy set partitioning of the initial domain of discourse is too coarse with respect to the reduced error and the rate of change of the error, and therefore the size of the domain of discourse needs to be changed. According to our control objective, as the number of iterations increases, the system is stable and the error becomes smaller and smaller. Thus gradually becoming smaller e and
the universe of discourse of (2) can effectively improve the control precision.
The transformed error input domain is recorded as E (t) [ -alpha ]1(t)Emin,α1(t)Emax]， CE(t)＝[-α2(t)CEmin,α2(t)CEmax]The output domain of the system is U (t) [ -beta (t) U [)min,β(t)Umax]Wherein α is1、α2And β are the scaling factors of the respective domains of interest, respectively. With the aid of the layer theory, the following functions are respectively selected as scaling factors for the input domains: alpha is alpha1＝(|e|/E)τ+ε，α2＝(|ec|/EC)τ+ ε, where 1 ≧ τ > 0, ε is a small positive number, and for the scaling factor of the output domain, the following function is chosen as the scaling factor
The implementation method of the system comprises the following steps:
step 1: building an experimental environment;
the subject is comfortably seated in a chair with the affected finger coupled to the exoskeleton robot. The hand is locked in the neutral position and the forearm is supported by the control box. The position of the exoskeleton and the height of the table are adjusted such that the angle between the forearm and the upper arm of the subject is between 90 degrees and 120 degrees. The testee wears MYO sensors at 1/3-2/3 positions on the healthy side forearm and the affected side forearm respectively, and the MYO sensors are connected with an operating computer through Bluetooth. And the control system of the finger rehabilitation robot, the data feedback signals of the MYO sensor and the pressure sensor are tested to ensure that no contact problem exists.
Step 2: training myoelectric signal sampling network parameters of the surfaces of the healthy and affected sides;
first, it is given a period of time familiar with the six basic actions designed. Subjects performed each exercise once in a random order given by the experimenter, repeating 15 times. For each exercise, the subject was asked to assume a comfortable position and amount of applied force, held for about 2 seconds while EMG data was recorded, and then given a brief rest to avoid fatigue. Inputting the training data samples into a neural network, calculating the cross entropy of the real action and the predicted action, and updating the network parameters by using a back propagation algorithm. Experiments show that the basic actions can reach more than 93 percent of accuracy.
And step 3: synchronous mirror training;
the testee grips the same pressure test bottle with both hands, opens the virtual environment interface of the game, and the testee carries out the transformation of basic action according to the game instruction. In the process, the rehabilitation robot identifies the action intention according to the electromyographic signals of the two hands and determines the motor to be controlled. According to the pressure difference of the two pressure sensors, the fuzzy self-adaptive PD control with the variable discourse domain is adopted to control the motor to perform auxiliary motion, so that the purposes of synchronous motion of two hands and output of the same acting force are achieved. The subject can stimulate mirror image neurons in the brain to excite, promote the brain nerve to be remodeled and induce the motor function to be recovered through visual and skin surface tactile feedback.
And 4, step 4: evaluating the rehabilitation effect;
the rehabilitation effect is evaluated according to the time domain characteristics of the pressure difference delta P of the pressure sensor and the characteristic signal difference delta f of the electromyographic sensor. Generally, the shorter the time for Δ P and Δ f to reach stability, the better the finger synchronization, the better the rehabilitation effect, and conversely, the worse the motor ability of the affected side.
As will be appreciated by one skilled in the art, embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks
The above description is only a preferred embodiment of the present invention, and it should be noted that, for those skilled in the art, several modifications and variations can be made without departing from the technical principle of the present invention, and these modifications and variations should also be regarded as the protection scope of the present invention.
Claims (7)
1. A finger mirror image rehabilitation training system is characterized by comprising an electromyographic signal extraction module, a pressure acquisition module, a main control module, a rehabilitation robot module, a human-computer interaction virtual environment module and a rehabilitation evaluation module;
the rehabilitation robot comprises a myoelectric signal extraction module, a pressure acquisition module, a rehabilitation robot module, a human-computer interaction virtual environment module, a rehabilitation evaluation module and a pressure acquisition module, wherein the myoelectric signal extraction module is used for extracting myoelectric signals of a healthy side and an affected side and transmitting the myoelectric signals to the main control module, the main control module is used for controlling the action of the rehabilitation robot module according to the myoelectric signals of the healthy side, the pressure acquisition module is used for acquiring finger pressure signals of the healthy side and the affected side and converting the finger pressure signals into electric signals to transmit the electric signals to the rehabilitation robot module, the rehabilitation robot module controls the fingers of the affected side to synchronously move on the healthy side and output the same acting force, the main control module simultaneously controls the human-computer interaction virtual environment module to simulate and display the motion postures corresponding to the fingers, and the electric signals output by the pressure acquisition module are also transmitted to the rehabilitation evaluation module for judging the rehabilitation effect;
the system also comprises a preprocessing module, a main control module and a power supply module, wherein the preprocessing module is used for extracting signal characteristics of the electromyographic signals acquired by the electromyographic signal extracting module by adopting a time-varying power spectrum method, learning and classifying the electromyographic signals by adopting an LSTM (least squares metric) model to obtain the electromyographic signals without interference, and transmitting the electromyographic signals to the main control module;
the processing process of the preprocessing module comprises the following steps:
the electromyographic signals collected by the electromyographic signal extraction module are divided into two groups:
the method is characterized in that the electromyographic signals of the first group are directly processed by a convolutional neural network, the convolutional neural network converts a group of sampling signals of 8 channels into a group of characteristic quantities, and the method comprises the following steps: constructing a neural network having a structure of convolutional layers and pooling layers of I layers, wherein the convolutional layer of the I layer has NiA convolution kernel, the ith convolution layer generates N when the sampled signal passes through the convolution neural network layer by layeriEach feature mapping matrix is sent to the (i + 1) th convolutional layer after being subjected to pooling layer down-sampling, b convolutional cores are arranged in the convolutional layer in the last layer of the convolutional neural network, the convolutional layer outputs b 2 x 2 feature mapping matrices, and then the b 2 x 2 feature mapping matrices are down-sampled into b 1 x 1 feature mapping matrices by the last pooling layer, namely feature vectors with the length of b;
extracting signal features of a second group of electromyographic signals by adopting a time-varying power spectrum method, sending the second group of extracted signal features and the first group of extracted feature vectors with the length of b into an MLP layer, completing splicing of the feature vectors in the MLP layer to obtain feature vectors with the dimension of 1 x (b +6), inputting the feature vectors into a three-layer fully-connected neural network, wherein the network output layer has c neurons in total, outputting a 1 x c matrix, storing the obtained matrix as historical features into a first-in first-out queue with the size of 1 x c i, storing the output from the current moment to the previous i moments as historical data, inputting the historical data into an LSTM neural network, and outputting the LSTM network layer to obtain the identification result of the whole neural network.
2. The finger mirror image rehabilitation training system according to claim 1, wherein the electromyographic signal extraction module comprises a ring array type electromyographic sensor and a Bluetooth device, and the ring array type electromyographic sensor is used for collecting the electromyographic signals of the front arm ends of the healthy side and the affected side and transmitting the electromyographic signals to the main control module through the Bluetooth device.
3. The finger mirror image rehabilitation training system according to claim 1, wherein the pressure acquisition module comprises a cylinder-like elastic body and a pressure sensor, and the pressure sensor module is attached to the top of the cylinder-like elastic body.
4. The finger mirror image rehabilitation training system according to claim 1, wherein the rehabilitation robot module comprises a driver, a control board and a flexible joint-based exoskeleton-type finger rehabilitation mechanism, the flexible joint-based exoskeleton-type finger rehabilitation mechanism comprises a base, a driving module, a back of the hand support body and a finger wearing part, wherein the base is used for placing a forearm and serves as a container for accommodating a power supply and the driver, the driving module comprises a motor, a gear and a related tendon rope which are sequentially connected and used for driving the finger wearing part to realize grabbing movement, the finger wearing part is a five-finger wearing part, the back of the hand support body is used for fixing the motor and connecting the finger wearing part, the control board is respectively connected with the main control module, the pressure acquisition module and the driver, and the driver is connected with the motor.
5. The finger mirror image rehabilitation training system according to claim 1, wherein the rehabilitation robot module processes the electrical signal input by the pressure acquisition module by a variable domain fuzzy self-adaptive PD control method, outputs a motor torque control signal related to a corresponding action, and controls the finger on the affected side to output the same acting force as the finger on the healthy side through the rehabilitation robot device.
6. The finger image rehabilitation training system according to claim 5, wherein the processing by the variable domain fuzzy adaptive PD control method comprises:
classifying the initial fuzzy domain of the pressure signal difference E between the healthy side and the affected side acquired by the pressure acquisition module, namely [ -E3, -E2, -E1, E0 and E1,E2,E3]Rate of change of pressure signal difference
The initial ambiguity domain is [ -CE3, -CE2, -CE1, CE0, CE1, CE2, CE3]The output variable is a motor corresponding to a finger, and the output torque domain of the output variable is expressed as U [ -U [ ]min,Umax]Establishing a fuzzy producer
E＝6×(e-Emin)/(Emax-Emin)
Wherein the pressure signal difference E E is Emin,Emax]And rate of change
EminAnd EmaxRespectively representing the minimum and maximum values, CE, of the discourse domain of the signal difference eminAnd CEmaxRespectively representing the rate of change of the signal difference
The minimum value and the maximum value of the discourse domain, E and CE are respectively corresponding outputs of the fuzzy producer; wherein-E3, -E2, -E1, E0, E1, E2, E3 and-CE 3, -CE2, -CE1, CE0, CE1, CE2 and CE3 respectively represent basic symbols of fuzzy set partition of corresponding fuzzy discourse
The fuzzy rule is designed according to the self-adaptive PD control law as follows:
U＝KpΕ+KdCΕ
Kp＝Kp0+ΔKp×qp
Kd＝Kd0+ΔKd×qd
wherein, Kp，KdIs the final control parameter of the PD controller; kp0，Kd0Setting parameters for the PD controller; Δ KP，ΔKdIs the output quantity of the fuzzy controller; q. q.sp，qdIs a correction factor;
fuzzy elimination of system output is carried out by a membership maximum value method, a membership function peak value of an output fuzzy subset is directly selected as a determination value of output, and a fuzzy subset U is outputiThe logical "and" of (c) is:
finally, the precise output of the control variable u is reversely solved by taking the median;
gradually decrease e and
the domain of the reduced pressure signal difference e is e (t [ - α [) ]1(t)Emin,α1(t)Emax]Rate of change
Is CE (t) [ - α [ ]2(t)CEmin,α2(t)CEmax]The output domain is U (t) [ - β (t) U [ ]min,β(t)Umax]Wherein α is1(t)、α2(t) and β (t) are respectively the scaling factors of the corresponding discourse domain, wherein;
α1(t)＝(|e|/E)τ+ε1，α2(t)＝(|ec|/EC)τ+ε2，
wherein 1 ≧ tau > 0, ε1And ε2Are respectively taken as epsilon1＝(Emax-Emin)/1000，ε2＝(CEmax-CEmin) A value of/1000, ec representing the rate of change of the pressure difference; kIIs a design scale parameter; e (τ) represents the pressure difference at time τ; β (0) represents the scaling factor for the initial output discourse domain; t is time;
pressure signal difference e and rate of change
The input domains E (t) and CE (t) are used as the input of domain fuzzy self-adaptive PD control, are used for feeding back the mirror image synchronization effect of the two hands at the key/affected side, and the output domain U (t) is used as the output of a driving motor of a corresponding finger.
7. The finger mirror image rehabilitation training system according to claim 1, wherein the rehabilitation evaluation module performs rehabilitation effect evaluation according to the pressure difference Δ P of the pressure acquisition module and the time domain characteristics of the characteristic signal difference Δ f of the electromyographic sensor, and the shorter the time for Δ P and Δ f to reach stability is, the better the finger synchronism is, the better the rehabilitation effect is, otherwise, the worse the motor ability of the affected side is.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910840714.6A CN110675933B (en) | 2019-09-06 | 2019-09-06 | Finger mirror image rehabilitation training system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910840714.6A CN110675933B (en) | 2019-09-06 | 2019-09-06 | Finger mirror image rehabilitation training system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110675933A CN110675933A (en) | 2020-01-10 |
| CN110675933B true CN110675933B (en) | 2022-05-17 |
Family
ID=69076073
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910840714.6A Active CN110675933B (en) | 2019-09-06 | 2019-09-06 | Finger mirror image rehabilitation training system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110675933B (en) |
Families Citing this family (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111583263B (en) * | 2020-04-30 | 2022-09-23 | 北京工业大学 | Point cloud segmentation method based on joint dynamic graph convolution |
| CN112951360B (en) * | 2021-03-29 | 2023-05-02 | 深圳大学 | Bilateral finger cooperative motion control method and system |
| CN113223666B (en) * | 2021-05-18 | 2023-04-28 | 深圳大学 | Advanced prediction control method for limb movement rehabilitation based on bilateral coordination |
| CN114237042B (en) * | 2021-11-25 | 2022-06-03 | 季华实验室 | Pulse feeling pre-pressure control method and system |
| CN114171162B (en) * | 2021-12-03 | 2022-10-11 | 广州穗海新峰医疗设备制造股份有限公司 | Mirror neuron rehabilitation training method and system based on big data analysis |
| CN114822761A (en) * | 2022-04-26 | 2022-07-29 | 东南大学 | Wrist rehabilitation training system based on muscle cooperation and variable stiffness impedance control |
| CN116312947B (en) * | 2023-03-13 | 2023-11-24 | 北京航空航天大学 | Immersive ankle and foot rehabilitation training method based on upper limb movement signals and electronic equipment |
| CN117373614A (en) * | 2023-12-07 | 2024-01-09 | 深圳市微克科技有限公司 | Finger health training method, system and storage medium based on wearing product |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104666047A (en) * | 2013-11-28 | 2015-06-03 | 中国科学院沈阳自动化研究所 | Double-side mirror image rehabilitation system based on biological information sensing |
| CN107928980A (en) * | 2017-11-22 | 2018-04-20 | 南京航空航天大学 | A kind of autonomous rehabilitation training system of the hand of hemiplegic patient and training method |
| CN108785997A (en) * | 2018-05-30 | 2018-11-13 | 燕山大学 | A kind of lower limb rehabilitation robot Shared control method based on change admittance |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11246786B2 (en) * | 2016-12-22 | 2022-02-15 | Rehab-Robotcs Company Ltd. | Power assistive device for hand rehabilitation and a method of using the same |
| CN108743223B (en) * | 2018-06-05 | 2020-03-03 | 山东大学 | Intelligent exoskeleton rehabilitation manipulator and method based on healthy side biological electric control |
-
2019
- 2019-09-06 CN CN201910840714.6A patent/CN110675933B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104666047A (en) * | 2013-11-28 | 2015-06-03 | 中国科学院沈阳自动化研究所 | Double-side mirror image rehabilitation system based on biological information sensing |
| CN107928980A (en) * | 2017-11-22 | 2018-04-20 | 南京航空航天大学 | A kind of autonomous rehabilitation training system of the hand of hemiplegic patient and training method |
| CN108785997A (en) * | 2018-05-30 | 2018-11-13 | 燕山大学 | A kind of lower limb rehabilitation robot Shared control method based on change admittance |
Non-Patent Citations (1)
| Title |
|---|
| Design of Finger Exoskeleton Rehabilitation Robot Using the Flexible Joint and the MYO Armband;Jianxi Zhang 等;《Lecture Notes in Computer Science》;20190806;第11745卷;第213-225页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110675933A (en) | 2020-01-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110675933B (en) | Finger mirror image rehabilitation training system | |
| Bi et al. | A review on EMG-based motor intention prediction of continuous human upper limb motion for human-robot collaboration | |
| CN107378944B (en) | Multidimensional surface electromyographic signal artificial hand control method based on principal component analysis method | |
| Ma et al. | Grasping force prediction based on sEMG signals | |
| Cai et al. | SVM-based classification of sEMG signals for upper-limb self-rehabilitation training | |
| Fukuda et al. | A human-assisting manipulator teleoperated by EMG signals and arm motions | |
| Pancholi et al. | Electromyography-based hand gesture recognition system for upper limb amputees | |
| Ai et al. | Machine learning in robot assisted upper limb rehabilitation: A focused review | |
| Vonsevych et al. | Fingers movements control system based on artificial neural network model | |
| Strazzulla et al. | Online bimanual manipulation using surface electromyography and incremental learning | |
| CN109199783B (en) | Control method for controlling stiffness of ankle joint rehabilitation equipment by using sEMG | |
| Yang et al. | An intention-based online bilateral training system for upper limb motor rehabilitation | |
| Das et al. | A review on the advancements in the field of upper limb prosthesis | |
| Xiao et al. | Real time motion intention recognition method with limited number of surface electromyography sensors for A 7-DOF hand/wrist rehabilitation exoskeleton | |
| Kulkarni et al. | Overview: mechanism and control of a prosthetic arm | |
| Tang et al. | Continuous estimation of human upper limb joint angles by using PSO-LSTM model | |
| CN114897012A (en) | Intelligent prosthetic arm control method based on vital machine interface | |
| De la Cruz-Sánchez et al. | EMG-controlled hand exoskeleton for assisted bilateral rehabilitation | |
| CN113171214A (en) | Multi-path feedback myoelectricity control prosthetic hand based on self-adaptive enhanced classifier and method | |
| Fang et al. | Modelling EMG driven wrist movements using a bio-inspired neural network | |
| Pons et al. | Objectives and technological approach to the development of the multifunctional MANUS upper limb prosthesis | |
| KR100994408B1 (en) | Method and device for deducting pinch force, method and device for discriminating muscle to deduct pinch force | |
| Cho et al. | Estimating simultaneous and proportional finger force intention based on sEMG using a constrained autoencoder | |
| Fu et al. | Identification of finger movements from forearm surface EMG using an augmented probabilistic neural network | |
| Akdoğan et al. | A muscular activation controlled rehabilitation robot system |
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 | ||
| CB02 | Change of applicant information |
Address after: 210012 No.9 Wenyuan Road, Qixia District, Nanjing City, Jiangsu Province Applicant after: NANJING University OF POSTS AND TELECOMMUNICATIONS Address before: No.12 Huatai Road, Yuchun District, Nanjing, Jiangsu Province Applicant before: NANJING University OF POSTS AND TELECOMMUNICATIONS |
|
| CB02 | Change of applicant information | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |