CN115192050A - Lower limb exoskeleton gait prediction method based on surface electromyography and feedback neural network - Google Patents
Lower limb exoskeleton gait prediction method based on surface electromyography and feedback neural network Download PDFInfo
- Publication number
- CN115192050A CN115192050A CN202210867468.5A CN202210867468A CN115192050A CN 115192050 A CN115192050 A CN 115192050A CN 202210867468 A CN202210867468 A CN 202210867468A CN 115192050 A CN115192050 A CN 115192050A
- Authority
- CN
- China
- Prior art keywords
- neural network
- network model
- angle
- knee joint
- user
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 91
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 32
- 210000003141 lower extremity Anatomy 0.000 title claims abstract description 30
- 230000005021 gait Effects 0.000 title claims abstract description 29
- 238000002567 electromyography Methods 0.000 title claims description 6
- 210000000629 knee joint Anatomy 0.000 claims abstract description 46
- 238000001914 filtration Methods 0.000 claims abstract description 36
- 238000012549 training Methods 0.000 claims abstract description 26
- 230000008569 process Effects 0.000 claims description 72
- 238000003062 neural network model Methods 0.000 claims description 52
- 238000012545 processing Methods 0.000 claims description 34
- 238000005452 bending Methods 0.000 claims description 33
- 238000005070 sampling Methods 0.000 claims description 20
- 230000033001 locomotion Effects 0.000 claims description 13
- 210000003205 muscle Anatomy 0.000 claims description 12
- 210000003127 knee Anatomy 0.000 claims description 9
- 230000010354 integration Effects 0.000 claims description 4
- 230000033228 biological regulation Effects 0.000 claims description 3
- 238000004364 calculation method Methods 0.000 claims description 3
- 210000000109 fascia lata Anatomy 0.000 claims description 3
- 210000003314 quadriceps muscle Anatomy 0.000 claims description 3
- 210000000689 upper leg Anatomy 0.000 claims description 3
- 210000001699 lower leg Anatomy 0.000 claims 1
- 238000004422 calculation algorithm Methods 0.000 abstract description 8
- 230000006870 function Effects 0.000 description 17
- 210000002569 neuron Anatomy 0.000 description 10
- 238000005516 engineering process Methods 0.000 description 7
- 230000000694 effects Effects 0.000 description 4
- 230000003183 myoelectrical effect Effects 0.000 description 4
- 238000004088 simulation Methods 0.000 description 4
- 238000012546 transfer Methods 0.000 description 4
- 230000009471 action Effects 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000003909 pattern recognition Methods 0.000 description 2
- 230000004044 response Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 244000309466 calf Species 0.000 description 1
- 230000001364 causal effect Effects 0.000 description 1
- 230000003750 conditioning effect Effects 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000004069 differentiation Effects 0.000 description 1
- 210000003414 extremity Anatomy 0.000 description 1
- 230000009123 feedback regulation Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000012905 input function Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 210000005036 nerve Anatomy 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Images
Classifications
-
- 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]
- A61B5/397—Analysis of electromyograms
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/103—Detecting, measuring or recording devices for testing the shape, pattern, colour, size or movement of the body or parts thereof, for diagnostic purposes
- A61B5/11—Measuring movement of the entire body or parts thereof, e.g. head or hand tremor, mobility of a limb
- A61B5/1121—Determining geometric values, e.g. centre of rotation or angular range of movement
-
- 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
- A61B5/25—Bioelectric electrodes therefor
- A61B5/251—Means for maintaining electrode contact with the body
- A61B5/257—Means for maintaining electrode contact with the body using adhesive means, e.g. adhesive pads or tapes
-
- 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
- A61B5/25—Bioelectric electrodes therefor
- A61B5/279—Bioelectric electrodes therefor specially adapted for particular uses
- A61B5/296—Bioelectric electrodes therefor specially adapted for particular uses for electromyography [EMG]
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/68—Arrangements of detecting, measuring or recording means, e.g. sensors, in relation to patient
- A61B5/6801—Arrangements of detecting, measuring or recording means, e.g. sensors, in relation to patient specially adapted to be attached to or worn on the body surface
- A61B5/6813—Specially adapted to be attached to a specific body part
- A61B5/6828—Leg
-
- 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
- A61B5/7271—Specific aspects of physiological measurement analysis
- A61B5/7275—Determining trends in physiological measurement data; Predicting development of a medical condition based on physiological measurements, e.g. determining a risk factor
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
Abstract
The invention belongs to the technical field of sensing and control engineering, and particularly relates to a lower limb exoskeleton gait prediction method based on surface myoelectricity and a feedback neural network; the method comprises the steps of collecting and filtering surface electromyographic signals and training and using a BP neural network control algorithm; the method takes the angle quantity of the knee joint in the predicted gait information as ideal output of the exoskeleton, and realizes accurate prediction of the gait through real-time feedback adjustment of an angle sensor on the exoskeleton; the invention realizes the accurate and rapid prediction of the gait of the user and the accurate control of the exoskeleton by using the cooperation of surface myoelectricity and a neural network and feedback control.
Description
Technical Field
The invention belongs to the technical field of sensing and control engineering, and particularly relates to a lower limb exoskeleton gait prediction method based on surface myoelectricity and a feedback neural network.
Background
The development of the robot technology has profound influence on human life, and along with the progress of the biological monitoring technology, the exoskeleton of the robot is developed to a certain extent. In the control field, the control mode of the robot exoskeleton can be basically considered as that the relation between relevant parameters and the positions and angles of four limbs is explored during human body movement, and an accurate and rapid model is searched. In the present development of exoskeleton technology, the traditional exoskeleton technology has basically matured and gained wide acceptance through the improvement of one generation and another. The exoskeleton technology of the robot is completely new after integrating functions in various fields.
The gait prediction of the traditional exoskeleton in the using process is realized based on a force sensor and an inclination angle sensor, but the force sensor, the inclination angle sensor and the like used in the driving process of the traditional exoskeleton technology have the characteristic of information lag, and are not intelligent and flexible enough.
In addition, in practical use, because the filter in the sensor used in the conventional exoskeleton technology is too simple, noise cannot be filtered effectively, and the problem of inaccurate analysis of the action intention of the human body is caused.
Disclosure of Invention
In order to overcome the problems, the invention provides a lower limb exoskeleton gait prediction method based on surface myoelectricity and a feedback neural network, which realizes accurate and rapid prediction of user gait and accurate control of exoskeleton by matching the surface myoelectricity and the neural network with feedback control.
In order to achieve the purpose, the invention provides the following scheme:
a lower limb exoskeleton gait prediction method based on surface myoelectricity and a feedback neural network comprises the following contents:
respectively pasting electrode plates of an sEMG collector on tensor fascia lata, rectus femoris and biceps femoris of a thigh muscle area of a user and gastrocnemius and tibialis anterior muscle of a calf muscle area, wearing an angle sensor on a knee joint of the user, and starting the movement of the user;
secondly, collecting a bioelectricity signal generated in the motion process of a user in real time by an electrode slice, transmitting the signal to an sEMG collector, carrying out primary filtering processing on the signal by the sEMG collector, carrying out formal filtering processing on the signal subjected to the primary filtering processing by the sEMG collector by adopting a Butterworth filter to obtain electromyographic signal data subjected to the formal filtering processing, and simultaneously collecting angle data of the bending of the knee joint of the user by an angle sensor;
step three, predicting the knee joint bending angle of the lower limb exoskeleton by adopting a BP neural network model, which specifically comprises the following steps: inputting electromyographic signal data subjected to formal filtering processing into a BP neural network model as input data, and after the electromyographic signal data are processed by the BP neural network model, taking output data of the BP neural network model as a predicted knee joint bending angle ideal value of the lower limb exoskeleton;
step four, realizing assistance of the lower limb exoskeleton on the gait of the user through PID feedback control and regulation, wherein the ideal value of the knee joint bending angle predicted by the BP neural network model is added with an adjustment quantity u (i) to be used as a final knee joint bending prediction angle; the adjustment amount u (i) is calculated as follows:
wherein K is p Is a proportionality coefficient, T i For integration period, T d The differential period is adopted, T is the sampling period of the sensor, n =0, 1 \8230, \8230i, e (i) is the error between the ideal value of the knee joint bending angle and the actual angle value acquired by the ith angle sensor, and the calculation is carried out according to the following formula;
e(i)＝C t -C d (i)
wherein C t Data output for the BP neural network model, i.e. the predicted knee joint bending angle ideal value, C, of the lower extremity exoskeleton d (i) For the ith acquisition of the angle sensorThe actual angle value.
The training process of the BP neural network model in the third step comprises the following contents:
step 1, recording a knee of a user as a sampling process once when the user bends, acquiring bioelectricity signal data once by each electrode plate in the process, acquiring knee joint bending angle data once by an angle sensor, and repeating the sampling process for m times;
and 2, respectively taking the electromyographic signal data obtained in the primary sampling process after formal filtering processing and the angle data acquired by the angle sensor as an input data group and an output data group of the BP neural network model so as to realize a primary training process of the BP neural network model, and repeating the training process 1000 or 2000 times to obtain the trained BP neural network model, wherein the data used in each training process are the data acquired in different sampling processes.
The method comprises the following steps that a sampling process is continuously carried out in the process that a user wears a lower limb exoskeleton, a BP neural network model is continuously trained, when the user wears the lower limb exoskeleton, a knee is bent once and is marked as a using process, in the using process, each electrode plate acquires bioelectrical signal data once, an angle sensor acquires angle data of one knee joint bending, and the using process is repeated for k times;
the electromyographic signal data obtained in the k times of use process after formal filtering processing and the electromyographic signal data obtained in the m times of training process after formal filtering processing are jointly used as an input data set of a BP neural network model, and the angle data obtained in the k times of use process and the angle data obtained in the m times of training process and collected by an angle sensor are jointly used as an output data set of the BP neural network model, so that the BP neural network model is trained, the BP neural network model is enabled to further predict the knee joint bending angle, and the knee joint bending angle is used as an ideal value of the knee joint bending angle of the lower limb exoskeleton predicted by the BP neural network model.
The invention has the beneficial effects that:
the invention combines the collection and conditioning of electromyographic signals, the algorithm of a neural network and a feedback control system, has the deep learning capability, can adjust the output quantity in real time along with the change of the body of a user, achieves the rapid and accurate prediction of the human movement intention, improves the response speed of the system, and can adapt to different occasions when the human body moves. Meanwhile, the collected electromyographic signals are preprocessed, noise and interference are eliminated, and the exoskeleton is controlled more accurately by further combining a feedback system. The method provides reference for other exoskeleton driving control equipment, provides a new method for the active exoskeleton to identify the human motion, and has certain application value.
Drawings
Fig. 1 is a model structure diagram of a BP neural network according to the present invention.
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings and examples. It is to be understood that the specific embodiments described herein are merely illustrative of the invention and are not to be construed as limiting the invention.
Example 1
A lower limb exoskeleton gait prediction method based on surface electromyography and a feedback neural network comprises a surface electromyography signal acquisition process, a filtering process, a training process and a using process of a neural network algorithm:
the acquisition process comprises:
step one, preparing work, respectively pasting electrode plates (n pieces in total) of an sEMG collector on tensor fascia lata, rectus femoris, biceps femoris and gastrocnemius and tibialis anterior muscle of a thigh muscle area of a user, wearing an angle sensor on a knee joint of the user, connecting motor equipment and a motor driver of an exoskeleton with a single chip microcomputer, connecting the single chip microcomputer, the sEMG collector and the angle sensor with the computer, and starting the movement of the user;
secondly, the electrode plate collects a bioelectricity signal generated in the movement process of a user in real time, the signal is transmitted to an sEMG collector, the sEMG collector performs primary filtering processing on the signal and then stores the signal in a computer, a Butterworth filter is adopted in the computer to perform formal filtering processing on the signal subjected to the primary filtering processing by the sEMG collector, electromyographic signal data subjected to the formal filtering processing are obtained, and meanwhile, an angle sensor collects angle data in real time and stores the angle data in the computer;
the process of performing formal filtering processing on the signal after the primary filtering processing of the sEMG collector by the Butterworth filter comprises the following steps:
(1) A Butterworth filter in an IIR digital filter is selected, the IIR digital filter belongs to a recursive linear time-invariant causal system, and difference variances of the IIR digital filter are as follows:
wherein: a is i And b i M is the order of the recursive filter, N is the order of the non-recursive filter, x is an input function, y is an output function, i is an order, and N is an independent variable in a time domain;
and performing z transformation on the digital signal to obtain a transfer function of the IIR digital filter:
the equation for the butterworth filter is as follows:
where ω is the frequency of the input signal, ω c For the cut-off frequency, n is the filter order, Z is the discrete domain variable, and H (ω) is its transfer function in the frequency domain;
(2) The method comprises the following steps that a Butterworth filter is used and comprises a Butterworth band-pass filter and a Butterworth band-stop filter, wherein the Butterworth band-pass filter is used for carrying out band-pass filtering on surface myoelectric signals, then the Butterworth band-stop filter is used for carrying out filtering processing on filtered environment signals, the environment signals are removed, and useful signals are reserved so as to reduce power frequency interference in the environment;
the Butterworth filter has a built-in algorithm in Matlab, and can directly call functions, input the passband boundary simulation frequency wp, the stopband boundary simulation frequency ws (simulation angular frequency, unit is rad/s), passband maximum attenuation Rp and stopband minimum attenuation As (unit is dB), can solve the N-filter order and wc-3dB cut-off simulation frequency (unit is rad/s, N and wc are calculated by using a butterd function), the filter function which inputs N and wc into the Matlab can output coefficients ai and bi, and then the filter function in the Matlab is used to input ai and bi and the signal x (N) to be filtered, so As to obtain the filtered signal y (N).
Step three, as shown in fig. 1, predicting the knee joint bending angle of the lower limb exoskeleton by adopting a BP neural network model, inputting the electromyographic signal data subjected to formal filtering processing as input data into the BP neural network model, and after processing by the BP neural network model, taking the output data of the BP neural network model as a predicted ideal value of the joint bending angle of the lower limb exoskeleton;
step four, the using process comprises the following steps: taking the established functional relation as a basic prediction function of gait prediction, gradually modifying and perfecting the prediction function along with the habit of a user in the later use process, taking knee joint angle information obtained by the prediction function as ideal output of the exoskeleton equipment, and realizing assistance of the lower limb exoskeleton on the gait of the user through PID (proportion integration differentiation) feedback control regulation, wherein the ideal value of the knee joint bending angle predicted by the BP (Back propagation) neural network model and an adjustment amount u (i) are taken as a final knee joint bending prediction angle; the adjustment amount u (i) is calculated as follows:
wherein K p Is a proportionality coefficient, T i For integration period, T d The differential period is adopted, T is the sampling period of the sensor, n =0, 1 \8230, \8230i, e (i) is the error between the ideal value of the knee joint bending angle and the actual angle value acquired by the ith angle sensor, and the calculation is carried out according to the following formula;
e(i)＝C t -C d (i)
wherein C is t Data output for the BP neural network model, i.e. the predicted knee joint bending angle ideal value, C, of the lower extremity exoskeleton d (i) The actual angle value collected by the angle sensor at the ith time.
The BP neural network model training process comprises the following steps:
firstly, recording a knee of a user as a sampling process once when the user bends, acquiring bioelectricity signal data once by each electrode plate in the process, acquiring knee joint bending angle data once by an angle sensor, and repeating the sampling process for m times;
and secondly, the electromyographic signal data obtained in the primary sampling process after formal filtering processing and the angle data acquired by the angle sensor are respectively used as an input data group and an output data group of the BP neural network model so as to realize the training process of the BP neural network model for one time, and the trained BP neural network model is obtained by repeating the training process 1000 or 2000 times, wherein the data used in each training process is the data acquired in different sampling processes.
The electromyographic signal data obtained in the m-time sampling process after formal filtering processing is used as an input data set of a BP neural network model, and the angle data obtained in the m-time sampling process and collected by an angle sensor is used as an output data set of the BP neural network model;
the electromyographic signal data of n electrode plates after formal filtering processing is used as the input of a BP neural network model, an input layer for establishing a neural network algorithm is composed of n neurons, each neuron represents electric signal data which is acquired by one electrode plate and is subjected to filtering processing, the number of output layers is 1, namely the angle value of gait, the number of nodes of a hidden layer is determined by an empirical formula, and the empirical formula is as follows:
h is the number of hidden layer nodes, p is the number of electric signal data subjected to filtering processing, namely the number of input layer nodes, q is the number of output layer nodes, namely the corresponding human knee joint angle, and u is an adjusting constant between 1 and 10;
the subsequent sampling process is continuously carried out in the process that the user wears the exoskeleton, the BP neural network model is continuously trained, the knee is bent once when the user wears the exoskeleton and is marked as a using process, in the process, each electrode plate collects once bioelectricity signal data, the angle sensor collects once bent angle data of the knee joint, and the using process is repeated for k times;
the electromyographic signal data obtained in the k times of use process after formal filtering processing and the electromyographic signal data obtained in the m times of training process after formal filtering processing are jointly used as an input data set of a BP neural network model, and the angle data obtained in the k times of use process and the angle data obtained in the m times of training process and collected by an angle sensor are jointly used as an output data set of the BP neural network model, so that the BP neural network model is trained;
continuously adjusting the weight, gradually establishing a functional relation between the knee joint angle and the surface myoelectricity of the user, further predicting the gait of the user by the BP neural network model, namely predicting the knee joint bending angle, and taking the knee joint bending angle to be controlled by the exoskeleton predicted by the BP neural network model as ideal output of the exoskeleton equipment.
Example 2
A lower limb exoskeleton gait prediction method based on surface electromyography and a feedback neural network, an electromyography analysis process, comprises the following steps:
(1) In a specified gait cycle (one gait cycle is calculated by the walking of the user), acquiring an electric signal detected by each electrode plate once as a surface electromyographic signal of the user when the lower limb of the user moves in each gait cycle, and forming an original data set;
(2) Taking each group of surface myoelectric signals in the original data group as the input of a BP neural network;
(3) Analyzing the surface myoelectric signal, and the algorithm comprises the following steps:
(a) The sEMG acquisition sensor, the angle sensor and the exoskeleton are worn by a user, motor equipment and a motor driver of the exoskeleton are connected with the single chip microcomputer, the sEMG acquisition device and the angle sensor are all connected with the computer, and the user starts to move; the user repeats k gait cycles for m times, repeats the acquisition process for m times, and performs filtering processing on the electric signal data acquired by the electrode plate stored in the computer to obtain m groups of training data, and applies the data to the MATLAB to perform the input data group of the BP neural network model.
The number of nodes of an input layer, namely the number of electromyographic signal data after filtering processing, is set to be n, the number of nodes of an output layer is set to be 1, the number of nodes of a hidden layer is set to be h, firstly, parameters in a neural network are initialized, namely, the parameters are input to a hidden weight w ij Weight v implicit to output jt Threshold value theta of the hidden layer j Threshold value gamma of output layer j To [ -1,1 [)]Set the ith neuron of the input layer as
The input of each neuron of the hidden layer is s j Output is
Output layer neuron input is L j With an output of C t ；
It is known from experience in this method that the hidden layer transfer function uses a tan-sigmoid function, i.e.
The output layer transfer function uses purelin function, i.e. g (x) = x;
calculating the input of each neuron of the hidden layer as s j And output is
Computing output layer neuron inputs as L j And the output is C t ：
C t ＝g(L j ) t＝1,2,…q
Calculating the real error of the output layer:
calculating node errors of each neuron in the middle layer:
for threshold value gamma in neural network j And the weight v jt And (5) correcting:
wherein t =1,2, \8230q; j =1,2, \8230; p;0< alpha <1
For threshold value theta in neural network j And the weight value w ij And (3) adjusting:
wherein i =1,2, \8230n; j =1,2, \8230; p;0< beta <1
(b) After a plurality of times of training (the training times are 1000 times or 2000 times), the functional relation between the surface electromyographic signals and the knee joint angles when the user walks is gradually established. The predicted knee angle in the next gait cycle of the user can be obtained from the average value of the knee angles in the first N gait cycles, and the predicted knee angles are detected in real time on the basis to adapt to the exercise habits of the user. The set safety threshold value of the angle sensor cannot be larger than 180 degrees and cannot be smaller than 0 degree, and the phenomenon that the rotation angle is too large due to interference is avoided.
The angle sensor collects angle information of devices such as an exoskeleton and the like and knee joint angle information of a user in real time, feedback regulation is built according to the angle information, the functional relation between surface electromyographic signals and the knee joint angle when the user walks is continuously corrected, correction is carried out through a neural network algorithm, the angle sensor collects the knee joint angle of the user and uses the knee joint angle as output data of the neural network algorithm to build the functional relation, so that the walking state and the road surface condition of the user are adapted, and a good control effect is achieved.
Along with the continuous walking of the user, the function relation between the surface myoelectric signals and the knee joint angle is continuously corrected by the neural network and the feedback control, the rapid and accurate prediction of the movement intention of the user is achieved, the feedback system is combined to achieve more accurate control on the outer skeleton, and the walking habit of the user is gradually adapted.
The method is based on the principle that exoskeleton gait prediction is realized on the basis of surface electromyographic signals, a Tewoth filter and a neural network; the surface electromyogram signal is a bioelectrical signal which is generated by muscle movement when a human body autonomously moves, and the physiological signal is picked up from the surface of the muscle by adopting an electrode. The generation of the surface electromyographic signals is conducted by nerves, is ahead of the final muscle activity to effect the movement, and the activities of muscle parts in different movements are obviously different, and the corresponding surface electromyographic signals are also obviously different. Therefore, the human action intention can be quickly identified by utilizing the surface electromyogram signal.
An artificial neural network is formed by connecting a large number of neurons to each other, and in the neural network, a forward neural network and a feedback neural network are divided according to a network structure. The learning mode can be divided into a learning network with a guide and a learning network without a guide, and the main difference between the two learning modes is whether a target output corresponding to an input guides training in the learning process. At present, neural networks are widely applied to the fields of automatic control, machine learning, pattern recognition and the like. In pattern recognition, the neural network has a mapping function capable of simulating any complex nonlinearity due to the difference of the number of neurons in the hidden layer, and has great advantages for nonlinear data operation, especially when a training sample is large, the advantages of the neural network are more obvious.
And the frequency response curve of Butterworth in a passband is smooth, the amplitude-frequency characteristic is stable, the distortion is less, the effective information of the original signal is kept as far as possible, and the accuracy of analyzing the human body action intention is improved.
Although the preferred embodiments of the present invention have been described in detail with reference to the accompanying drawings, the scope of the present invention is not limited to the specific details of the above embodiments, and any person skilled in the art can substitute or change the technical solution of the present invention and its inventive concept within the technical scope of the present invention, and these simple modifications belong to the scope of the present invention.
Claims (3)
1. A lower limb exoskeleton gait prediction method based on surface myoelectricity and a feedback neural network is characterized by comprising the following contents:
step one, respectively pasting electrode plates of sEMG collectors on tensor fascia lata, rectus femoris and biceps femoris of thigh muscle areas of a user and gastrocnemius and tibialis anterior muscles of crus muscle areas, wearing an angle sensor on knee joints of the user, and starting the movement of the user;
secondly, collecting a bioelectricity signal generated in the motion process of a user in real time by an electrode slice, transmitting the signal to an sEMG collector, carrying out primary filtering processing on the signal by the sEMG collector, carrying out formal filtering processing on the signal subjected to the primary filtering processing by the sEMG collector by adopting a Butterworth filter to obtain electromyographic signal data subjected to the formal filtering processing, and simultaneously collecting angle data of the bending of the knee joint of the user by an angle sensor;
step three, predicting the knee joint bending angle of the lower limb exoskeleton by adopting a BP neural network model, which specifically comprises the following steps: inputting electromyographic signal data subjected to formal filtering processing into a BP neural network model as input data, and after the electromyographic signal data is processed by the BP neural network model, taking output data of the BP neural network model as a predicted knee joint bending angle ideal value of the lower limb exoskeleton;
step four, realizing the power assistance of the lower limb exoskeleton on the gait of the user through PID feedback control and regulation, wherein the ideal value of the knee joint bending angle predicted by the BP neural network model is added with an adjustment amount u (i) to be used as a final knee joint bending prediction angle; the adjustment amount u (i) is calculated as follows:
wherein K p Is a proportionality coefficient, T i For integration period, T d The differential period is adopted, T is the sampling period of the sensor, n =0, 1 \8230, \8230i, e (i) is the error between the ideal value of the knee joint bending angle and the actual angle value acquired by the ith angle sensor, and the calculation is carried out according to the following formula;
e(i)＝C t -C d (i)
wherein C is t Data output for the BP neural network model, i.e. the predicted knee joint bending angle ideal value, C, of the lower extremity exoskeleton d (i) The actual angle value collected by the angle sensor at the ith time.
2. The method for predicting gait of lower extremity exoskeleton of claim 1, wherein the training process of BP neural network model in step three comprises the following steps:
step 1, recording a knee of a user as a sampling process once when the user bends, acquiring bioelectricity signal data once by each electrode plate in the process, acquiring knee joint bending angle data once by an angle sensor, and repeating the sampling process for m times;
and 2, respectively taking the electromyographic signal data obtained in the primary sampling process after formal filtering processing and the angle data acquired by the angle sensor as an input data group and an output data group of the BP neural network model so as to realize a primary training process of the BP neural network model, and repeating the training process 1000 or 2000 times to obtain the trained BP neural network model, wherein the data used in each training process are the data acquired in different sampling processes.
3. The method for predicting gait of lower extremity exoskeleton of claim 2 based on surface electromyography and feedback neural network, wherein the sampling process is continuously performed during the process that a user wears the lower extremity exoskeleton, the training is continuously performed on the BP neural network model, the knee of the user is bent once when wearing the lower extremity exoskeleton is recorded as a one-time use process, in the process, each electrode plate collects one-time bioelectricity signal data, the angle sensor collects one-time knee joint bending angle data, and the use process is repeated for k times;
the electromyographic signal data obtained in the k times of use process after formal filtering processing and the electromyographic signal data obtained in the m times of training process after formal filtering processing are jointly used as an input data set of a BP neural network model, and the angle data obtained in the k times of use process and the angle data obtained in the m times of training process and collected by an angle sensor are jointly used as an output data set of the BP neural network model, so that the BP neural network model is trained, the BP neural network model is enabled to further predict the knee joint bending angle, and the knee joint bending angle is used as an ideal value of the knee joint bending angle of the lower limb exoskeleton predicted by the BP neural network model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210867468.5A CN115192050A (en) | 2022-07-22 | 2022-07-22 | Lower limb exoskeleton gait prediction method based on surface electromyography and feedback neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210867468.5A CN115192050A (en) | 2022-07-22 | 2022-07-22 | Lower limb exoskeleton gait prediction method based on surface electromyography and feedback neural network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115192050A true CN115192050A (en) | 2022-10-18 |
Family
ID=83583298
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210867468.5A Pending CN115192050A (en) | 2022-07-22 | 2022-07-22 | Lower limb exoskeleton gait prediction method based on surface electromyography and feedback neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115192050A (en) |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2014155998A (en) * | 2013-02-18 | 2014-08-28 | Advanced Telecommunication Research Institute International | Power assist robot |
| CN105943207A (en) * | 2016-06-24 | 2016-09-21 | 吉林大学 | Intelligent artificial limb movement system based on idiodynamics and control methods thereof |
| CN110141239A (en) * | 2019-05-30 | 2019-08-20 | 东北大学 | A kind of motion intention identification and installation method for lower limb exoskeleton |
| CN110339024A (en) * | 2019-07-19 | 2019-10-18 | 中国科学院深圳先进技术研究院 | Lower limb exoskeleton robot and its real-time gait switching method and storage device |
| CN110653817A (en) * | 2019-08-20 | 2020-01-07 | 南京航空航天大学 | Exoskeleton robot power-assisted control system and method based on neural network |
| CN111643321A (en) * | 2020-04-30 | 2020-09-11 | 北京精密机电控制设备研究所 | Exoskeleton joint angle prediction method and system based on sEMG signals |
| CN111803250A (en) * | 2020-07-03 | 2020-10-23 | 北京联合大学 | Knee joint angle prediction method and system based on electromyographic signals and angle signals |
| CN114757339A (en) * | 2022-04-18 | 2022-07-15 | 华中科技大学鄂州工业技术研究院 | Hip joint track generation method based on human limb cooperative law and related equipment |
-
2022
- 2022-07-22 CN CN202210867468.5A patent/CN115192050A/en active Pending
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2014155998A (en) * | 2013-02-18 | 2014-08-28 | Advanced Telecommunication Research Institute International | Power assist robot |
| CN105943207A (en) * | 2016-06-24 | 2016-09-21 | 吉林大学 | Intelligent artificial limb movement system based on idiodynamics and control methods thereof |
| CN110141239A (en) * | 2019-05-30 | 2019-08-20 | 东北大学 | A kind of motion intention identification and installation method for lower limb exoskeleton |
| CN110339024A (en) * | 2019-07-19 | 2019-10-18 | 中国科学院深圳先进技术研究院 | Lower limb exoskeleton robot and its real-time gait switching method and storage device |
| CN110653817A (en) * | 2019-08-20 | 2020-01-07 | 南京航空航天大学 | Exoskeleton robot power-assisted control system and method based on neural network |
| CN111643321A (en) * | 2020-04-30 | 2020-09-11 | 北京精密机电控制设备研究所 | Exoskeleton joint angle prediction method and system based on sEMG signals |
| CN111803250A (en) * | 2020-07-03 | 2020-10-23 | 北京联合大学 | Knee joint angle prediction method and system based on electromyographic signals and angle signals |
| CN114757339A (en) * | 2022-04-18 | 2022-07-15 | 华中科技大学鄂州工业技术研究院 | Hip joint track generation method based on human limb cooperative law and related equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109394476B (en) | Method and system for automatic intention recognition of brain muscle information and intelligent control of upper limbs | |
| CN110653817B (en) | Exoskeleton robot power-assisted control system and method based on neural network | |
| CN105213153B (en) | Lower limb rehabilitation robot control method based on brain flesh information impedance | |
| CN110755070B (en) | Multi-sensor fusion-based lower limb movement pose rapid prediction system and method | |
| CN109262618B (en) | Muscle cooperation-based upper limb multi-joint synchronous proportional myoelectric control method and system | |
| CN102727361B (en) | Sitting and lying type lower limb rehabilitation robot | |
| CN106067178A (en) | A kind of hand joint based on muscle synergistic activation model continuous method of estimation of motion | |
| CN112842824B (en) | Training method for rehabilitation and recovery of lower limbs | |
| CN106110587A (en) | Lower limb exoskeleton rehabilitation system based on man-computer cooperation and method | |
| CN106236503A (en) | The wearable exoskeleton system of the electrically driven (operated) upper limb of flesh and control method | |
| CN110339024A (en) | Lower limb exoskeleton robot and its real-time gait switching method and storage device | |
| CN113058157B (en) | Feedback type functional electrical stimulation system with multi-signal fusion | |
| Liu et al. | An adaptive upper-arm EMG-based robot control system | |
| CN109758336A (en) | A kind of control method carrying out active rehabilitation care using patient's Ipsilateral sEMG | |
| CN112206124B (en) | Neural loop-guided upper limb function rehabilitation training system and method | |
| Li et al. | A temporally smoothed MLP regression scheme for continuous knee/ankle angles estimation by using multi-channel sEMG | |
| CN115177864A (en) | Functional electrical stimulation closed-loop regulation and control method based on muscle activation degree and LSTM | |
| Zangene et al. | Continuous estimation of knee joint angle during squat from sEMG using artificial neural networks | |
| CN109645995A (en) | Joint motions estimation method based on myoelectricity model and Unscented kalman filtering | |
| Turgunov et al. | Comparative analysis of the results of EMG signal classification based on machine learning algorithms | |
| CN115192050A (en) | Lower limb exoskeleton gait prediction method based on surface electromyography and feedback neural network | |
| Khanjani et al. | Estimate human-force from sEMG signals for a lower-limb rehabilitation robot | |
| CN105769186B (en) | Electromyographic signal collection position choosing method based on complex network | |
| Anwar et al. | The estimation of knee joint angle based on generalized regression neural network (GRNN) | |
| Khoshdel et al. | Application of statistical techniques and artificial neural network to estimate force from sEMG signals |
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 |