CN113274039B - Prediction classification method and device based on surface electromyogram signals and motion signals - Google Patents

Prediction classification method and device based on surface electromyogram signals and motion signals Download PDF

Info

Publication number
CN113274039B
CN113274039B CN202110547866.4A CN202110547866A CN113274039B CN 113274039 B CN113274039 B CN 113274039B CN 202110547866 A CN202110547866 A CN 202110547866A CN 113274039 B CN113274039 B CN 113274039B
Authority
CN
China
Prior art keywords
motion
layer
signal
data
signals
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202110547866.4A
Other languages
Chinese (zh)
Other versions
CN113274039A (en
Inventor
陈昕
梁杰
陈述荣
苏婵娟
郑军凡
刘丽金
叶佳
姜海燕
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Fuzhou Second Hospital
Original Assignee
Fuzhou Second Hospital
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Fuzhou Second Hospital filed Critical Fuzhou Second Hospital
Priority to CN202110547866.4A priority Critical patent/CN113274039B/en
Publication of CN113274039A publication Critical patent/CN113274039A/en
Application granted granted Critical
Publication of CN113274039B publication Critical patent/CN113274039B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/103—Detecting, measuring or recording devices for testing the shape, pattern, colour, size or movement of the body or parts thereof, for diagnostic purposes
    • A61B5/1036—Measuring load distribution, e.g. podologic studies
    • A61B5/1038—Measuring plantar pressure during gait
    • 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
    • A61B5/112—Gait analysis
    • 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
    • A61B5/7225—Details of analog processing, e.g. isolation amplifier, gain or sensitivity adjustment, filtering, baseline or drift compensation
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7235—Details of waveform analysis
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7235—Details of waveform analysis
    • A61B5/7253—Details of waveform analysis characterised by using transforms
    • A61B5/726—Details of waveform analysis characterised by using transforms using Wavelet transforms
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7235—Details of waveform analysis
    • A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
    • 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

Landscapes

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

Abstract

The invention provides a prediction classification method and a prediction classification device based on a surface electromyogram signal and a motion signal, wherein the method comprises the following steps: collecting surface electromyographic signals and motion signals; carrying out first processing on the acquired motion signals, and calculating motion data corresponding to the gait coordination parameters; carrying out second processing on the collected surface electromyographic signals, and extracting corresponding characteristic data; and taking the motion data and the characteristic data as input data of a neural network of a recurrent cerebellum model, and carrying out training prediction to obtain a classification result. Through the technical scheme, the classification and identification of the surface myoelectric signals and the motion signals can be realized, and further the clinical diagnosis and curative effect evaluation of knee osteoarthritis are promoted.

Description

Prediction classification method and device based on surface electromyogram signals and motion signals
Technical Field
The invention relates to the technical field of classification, in particular to a prediction classification method and device based on surface electromyogram signals and motion signals.
Background
The surface electromyographic signal (sEMG) is formed by superposing action electric potential sequences generated by muscle excitation during movement on the surface of skin, and is a non-stable weak bioelectric signal. The intelligent rehabilitation device is generally generated 30-150 ms ahead of limb movement, has good anti-interference performance and is easy to obtain, can reflect the physiological state and the activity condition of a neuromuscular system, is always considered as the optimal physiological feedback signal of the intelligent rehabilitation device, and is widely applied to the human-computer interaction technology. The plantar pressure detection is to analyze dynamic characteristics and motility characteristics of the foot by measuring the characteristics of mechanical, geometric, time parameters and the like of the foot pressure and analyzing plantar pressure parameters in different states. The three-axis accelerometer, the angular velocity sensor, the plantar pressure sensor and the like can be used as motion signals to analyze step length, step width, step direction angle and step speed and analyze regularity and symmetry of gait as gait coordination parameters.
Osteoarthritis is a chronic arthritic condition characterized by degeneration of articular cartilage and secondary hyperosteogeny. With the development of society, the pace of life of people is also accelerated, the knee joint is the joint with the largest load among all joints of the human body and is also the joint with the most serious wear, and Knee Osteoarthritis (KOA) is the most common type of osteoarthritis, also called as knee proliferative arthritis and knee degenerative osteoarthritis. The analysis parameters of the fusion surface electromyogram signal and the motion signal reflect the disease process of the knee osteoarthritis, and the measurement and the related analysis processing of the fusion surface electromyogram signal and the motion signal have irreplaceable effects on the clinical diagnosis and the curative effect evaluation of the knee osteoarthritis.
Disclosure of Invention
Therefore, a technical scheme of prediction classification based on the surface electromyogram signal and the motion signal is required to be provided for realizing classification identification of the surface electromyogram signal and the motion signal.
To achieve the above object, a first aspect of the present application provides a predictive classification method based on a surface electromyogram signal and a motion signal, the method including:
collecting surface electromyographic signals and motion signals;
carrying out first processing on the acquired motion signals, and calculating motion data corresponding to the gait coordination parameters;
carrying out second processing on the collected surface electromyographic signals, and extracting corresponding characteristic data;
and taking the motion data and the characteristic data as input data of a neural network of a recurrent cerebellum model, and carrying out training prediction to obtain a classification result.
Further, the motion signal comprises any one or more of step size, step width, step angle, step speed.
Further, the motion data includes classification parameters, and the first processing includes: and calculating classification parameters corresponding to the gait harmony parameters based on the walking time phase.
Further, the classification parameters include a support phase plantar pressure symmetry index, and the support phase plantar pressure symmetry index is calculated according to the following method:
dividing the support phase into a plurality of periods, and calculating a mean matrix of N pressures in the plurality of periods according to a plurality of pressure data collected by the film pressure sensor, such as a left support phase pressure value matrix:
Figure BDA0003074288740000021
wherein: pLIs a matrix of left-hand support phase pressure values, PghThe pressure value of the h support phase of the g film pressure sensor is g ═ 1, 2, …, N1;
calculating a left pressure value matrix PLL ofP1L of norm and right pressure value matrixP2Norm, LP1And LP2The calculation formula of (a) is as follows:
Figure BDA0003074288740000022
Figure BDA0003074288740000031
calculating L of left and right support phase plantar pressure matrix in a plurality of periodsP1Norm and LP2And taking the norm as the plantar pressure symmetry index of the support phase.
Further, the second processing includes noise cancellation processing, and the extracting the corresponding feature data includes:
and performing dimensionality reduction on the data subjected to the denoising processing to obtain corresponding characteristic data.
Further, the characteristic data comprises an index of symmetry of the myoelectric signal of the supporting facies surface, and the index of symmetry of the myoelectric signal of the supporting facies surface is calculated according to the following method:
acquiring a characteristic parameter matrix of a plurality of surface electromyographic signals of a support phase after dimension reduction treatment, wherein the characteristic parameters comprise average power frequency or integral electromyography; for example, the left electromyographic signal matrix is as follows:
Figure BDA0003074288740000032
wherein: sLIs a left-side supporting phase electromyographic signal matrix, SghThe mean power frequency or integral myoelectricity of the z-th support phase of the t-th surface myoelectric signal, t is 1, 2, …, N2.
Calculating L of the left-hand eigen parameter matrixS1Norm and L of right-hand eigen parameter matrixS2Norm to obtain the symmetry index of the support phase surface electromyogram signal, and the calculation formula is as follows:
Figure BDA0003074288740000033
further, the recurrent cerebellum model neural network comprises an input layer, an association memory layer, a receiving domain layer, a weight memory layer and an output layer which are sequentially connected;
the input data of the input layer is the motion data and the feature data, and the output result of the output layer is the classification result;
the associative memory layer is a Gaussian wavelet function, the weight memory layer is used for providing weight, and the receiving domain layer is used for providing convolution kernel size.
Further, the gaussian wavelet function calculation formula is as follows:
Figure BDA0003074288740000041
wherein x isikIs a signal input to the associative memory layer, bikAnd aikRespectively a translation parameter and a dilation parameter, r, corresponding to said gaussian wavelet functionikOutputting the result for the associative memory layer;
wherein x isikIs represented as:
xik(t)＝xi(t)+wikrik(t-1) (5)
where t is the time series of data, wikIs the weight of this recursive element, which represents the influence of the output result at the previous instant on this instant, rikAnd (t-1) the output result at the previous time.
Further, the output layer comprises an activation function, the activation function adopts a sigmoid function, and the calculation formula of the sigmoid function is as follows:
Figure BDA0003074288740000042
where x is the signal input to the output layer;
the expression from the input layer to the output layer is:
Figure BDA0003074288740000043
wherein, IiInputting a feature vector, wherein m is a feature vector dimension; w is aikIs a weight value between the input layer and the associative memory layer; said wkjIs the weight value between the weight memory layer and the output layer, and n represents the division of the input vector characteristicsResolution, o, is the number of output layer classifications.
A second aspect of the present invention provides a predictive classification apparatus based on a surface electromyogram signal and a motion signal, the apparatus being configured to perform the predictive classification method based on a surface electromyogram signal and a motion signal according to the first aspect of the present invention.
Different from the prior art, the invention provides a prediction classification method and a prediction classification device based on a surface electromyogram signal and a motion signal, wherein the method comprises the following steps: collecting surface electromyographic signals and motion signals; carrying out first processing on the acquired motion signals, and calculating motion data corresponding to the gait coordination parameters; carrying out second processing on the collected surface electromyographic signals, and extracting corresponding characteristic data; and taking the motion data and the characteristic data as input data of a neural network of a recurrent cerebellum model, and carrying out training prediction to obtain a classification result. Through the technical scheme, the classification and identification of the surface myoelectric signals and the motion signals can be realized, and further the clinical diagnosis and the curative effect evaluation of knee osteoarthritis are promoted.
Drawings
Fig. 1 is a flowchart of a prediction classification method based on a surface electromyogram signal and a motion signal according to an embodiment of the present invention;
FIG. 2 is a flowchart of a prediction classification apparatus based on surface electromyography and motion signals according to an embodiment of the present invention;
fig. 3 is a schematic diagram of a neural network of a recursive cerebellar model according to an embodiment of the present invention.
Reference numerals:
101. an input layer; 102. an associative memory layer; 103. receiving a domain layer; 104. a weight memory layer; 105. an output layer; 106. a recursion unit;
1. a motion signal detection module; 10. a thin film pressure sensor; 11. a three-axis accelerometer;
2. a surface electromyogram signal detection unit; 20. a surface electromyography sensor; 21. a surface electromyography sensor; 22. a three-axis accelerometer; 23. an amplification filter circuit;
3. an embedded microprocessor;
30. a communication module; 31. serial port communication; 32. carrying out Bluetooth communication;
40. a storage unit; 41. FLASH; 42. SDRAM;
50. a terminal; 51. a recurrent cerebellar model neural network.
Detailed Description
To explain technical contents, structural features, and objects and effects of the technical solutions in detail, the following detailed description is given with reference to the accompanying drawings in conjunction with the embodiments.
As shown in fig. 1, which is a flowchart of a prediction classification method based on a surface electromyogram signal and a motion signal according to an embodiment of the present invention, the method includes:
firstly, step S101 is carried out to collect surface electromyographic signals and motion signals;
then, step S102 is carried out to carry out first processing on the collected movement signals, and movement data corresponding to the gait harmony parameters are calculated;
then, step S103 is carried out to carry out second processing on the collected surface electromyographic signals, and corresponding characteristic data are extracted;
and then, step S104 is carried out, the motion data and the feature data are used as input data of a neural network of the recurrent cerebellum model, training and prediction are carried out, and a classification result is obtained.
According to the scheme, the surface electromyographic signals and the motion signals are collected, corresponding motion data and characteristic data are generated according to the surface electromyographic signals and the motion signals, the motion data and the characteristic data are used as input data of the neural network of the recursion cerebellum model, training and prediction are carried out, a classification result is obtained, the classification result can correspondingly represent disease processes of knee osteoarthritis reflected by the currently input data, and the accuracy and the efficiency of clinical diagnosis of the knee osteoarthritis can be effectively improved.
In some embodiments, the motion signal comprises any one or more of a step size, a step width, a step angle, a pace. The motion signal can be collected through a plantar pressure sensor, a three-axis accelerometer, an angular velocity detection sensor and the like. The motion data includes classification parameters, and the first processing includes: and calculating classification parameters corresponding to the gait harmony parameters based on the walking time phase.
Preferably, the classification parameter includes a support phase plantar pressure symmetry index, and the support phase plantar pressure symmetry index is calculated according to the following method:
dividing the support phase into a plurality of periods (such as six), and calculating a mean matrix of N pressures in the plurality of periods according to a plurality of pressure data collected by the membrane pressure sensor, such as a left support phase pressure value matrix:
Figure BDA0003074288740000071
wherein: pLIs a matrix of left-hand support phase pressure values, PghThe pressure value of the h support phase of the g film pressure sensor is g ═ 1, 2, …, N1;
calculating a left pressure value matrix PLL ofP1L of norm and right pressure value matrixP2Norm, LP1And LP2The calculation formula of (a) is as follows:
Figure BDA0003074288740000072
Figure BDA0003074288740000073
calculating L of left and right support phase plantar pressure matrix in a plurality of periodsP1Norm and LP2And taking the norm as the plantar pressure symmetry index of the support phase.
In some embodiments, the second processing comprises noise cancellation processing, and the extracting the respective feature data comprises: and performing dimensionality reduction on the data subjected to the denoising processing to obtain corresponding characteristic data. In the practical application process, when the Median Frequency (MF), the average power frequency (MPF), the average myoelectricity (AEMG) and the integrated myoelectricity (IEMG) of the surface electromyogram signals of the rectus femoris, the external gastrocnemius muscle, the internal gastrocnemius muscle, the tibialis anterior muscle, the rectus femoris muscle, the biceps femoris muscle and the like are used, due to the coupling effect among the joint muscles, redundancy exists. The prediction accuracy cannot be improved on the basis of increasing the prediction time by data redundancy, so that the PCA is adopted to perform dimension reduction processing on the data in the embodiment to obtain the characteristic data of surface electromyogram signal classification so as to reduce redundant data.
In some embodiments, the characteristic data comprises an indicator of support facies surface electromyographic signal symmetry calculated according to:
acquiring a characteristic parameter matrix of a plurality of surface electromyographic signals of a support phase after dimension reduction treatment, wherein the characteristic parameters comprise average power frequency or integral electromyography; for example, the left electromyographic signal matrix is as follows:
Figure BDA0003074288740000081
wherein: sLIs a left-side supporting phase electromyographic signal matrix, SghThe average power frequency or integral myoelectricity of the z-th support phase of the t-th surface myoelectricity signal is t 1, 2, …, N2.
Calculating L of the left-hand eigen parameter matrixS1Norm and L of right-hand eigen parameter matrixS2Norm to obtain the symmetry index of the support phase surface electromyogram signal, and the calculation formula is as follows:
Figure BDA0003074288740000082
as shown in fig. 3, in some embodiments, the recurrent cerebellar model neural network includes an input layer 101, an association memory layer 102, a receiving domain layer 103, a weight memory layer 104, and an output layer 105, which are connected in sequence; the input data of the input layer 101 is the motion data and the feature data, and the output result of the output layer 105 is the classification result; the associative memory layer 102 is a gaussian wavelet function, the weight memory layer 104 is used for providing weight values, and the receiving domain layer 103 is used for providing convolution kernel values. The weight memory layer records weight values (i.e. convolution kernels), and the receiving domain layer (i.e. receiving domain) refers to the size of the convolution kernels. In other embodiments, the neural network of the recursive cerebellar model further includes a recursion unit 106, and the recursion unit 106 is configured to perform a recursion operation, that is, to continuously train according to data received by the input layer, so as to improve the accuracy of prediction.
In some embodiments, the gaussian wavelet function is calculated as follows:
Figure BDA0003074288740000083
wherein x isikIs a signal input to the associative memory layer, bikAnd aikRespectively a translation parameter and a dilation parameter, r, corresponding to said gaussian wavelet functionikOutputting the result for the associative memory layer;
wherein x isikIs represented as:
xik(t)＝xi(t)+wikrik(t-1) (5)
where t is the time series of data, wikIs the weight of this recursive element, which represents the influence of the output result at the previous instant on this instant, rikAnd (t-1) the output result at the previous time.
In some embodiments, the output layer includes an activation function that employs a sigmoid function whose formula is calculated as follows:
Figure BDA0003074288740000091
where x is the signal input to the output layer;
the expression from the input layer to the output layer is:
Figure BDA0003074288740000092
wherein, IiInputting a feature vector, wherein m is a feature vector dimension; w is aikIs a weight value between the input layer and the associative memory layer; said wkjThe weight value between the weight memory layer and the output layer, n represents the resolution of the input vector characteristics, and o is the number of output layer classifications.
The recurrent cerebellum neural network is a neural network which is fast, strong in generalization capability and close locally based on neurophysiology, in order to accelerate the overall convergence speed of the recurrent cerebellum model neural network and improve the network generalization, the recurrent cerebellum model neural network is improved by taking a wavelet function as a membership function and a method for increasing fuzzy logic, and the improved recurrent cerebellum model neural network is used for classification prediction. The improved recurrent cerebellar model neural network can effectively improve the network generalization.
As shown in fig. 2, a second aspect of the present invention provides a predictive classification apparatus based on a surface electromyogram signal and a motion signal, the apparatus being configured to perform the predictive classification method based on a surface electromyogram signal and a motion signal according to the first aspect of the present invention.
The device comprises a motion signal detection module 1, a surface electromyogram signal detection unit 2, an embedded microprocessor 3, a communication module 30, a storage unit 40 and a terminal 50, wherein the motion signal detection module 1 comprises a film pressure sensor 10 and a triaxial accelerometer 11, and the surface electromyogram signal detection unit 2 comprises a plurality of surface electromyogram sensors, a triaxial accelerometer 22 and an amplification filter circuit 23. The surface electromyography sensor comprises a first surface electromyography sensor 20 or a second surface electromyography sensor 21, but in other embodiments, the number of the surface electromyography sensors may be other values. The communication module 30 includes serial communication 31 or bluetooth communication 32, the storage unit 40 includes FLASH 41 or SDRAM 42, and the recurrent cerebellar model neural network 51 may be disposed on the terminal 50.
The motion signal detection module 1 and the surface electromyogram signal detection unit 2 are respectively connected with the embedded microprocessor 3, the embedded microprocessor 3 is further respectively connected with the communication module 30 and the storage unit 40, and the communication module 30 is further connected with the terminal 50. The terminal is an electronic device with a data processing function and can be a mobile phone, a tablet, a personal computer and the like. In other embodiments, the embedded microprocessor 3 may also be replaced by a CPU, MCU, GPU, etc.
The invention provides a prediction and classification method based on surface electromyogram signals and motion signals. And then selecting a surface electromyogram signal of the related muscle of the joint, adopting the Median Frequency (MF), the average power frequency (MPF), the Average Electromyogram (AEMG) and the Integral Electromyogram (IEMG) of the surface electromyogram signal, and performing dimensionality reduction processing on each item of data to obtain characteristic data of surface electromyogram signal classification. And then, fusing the characteristic data of the surface electromyogram signal and the motion signal classification, taking the characteristic data as the input data of the improved recursive cerebellar model network, and performing training prediction to realize classification and identification.
Finally, it should be noted that, although the above embodiments have been described herein, the scope of the present invention is not limited thereby. Therefore, based on the innovative concepts of the present invention, the technical solutions of the present invention can be directly or indirectly applied to other related technical fields by changing and modifying the embodiments described herein or by using the equivalent structures or equivalent processes of the content of the present specification and the attached drawings, and are included in the scope of the present invention.

Claims (5)

1. A predictive classification device based on a surface electromyography signal and a motion signal, characterized in that it is adapted to perform the following steps:
collecting surface electromyographic signals and motion signals;
carrying out first processing on the acquired motion signals, and calculating motion data corresponding to the gait coordination parameters;
carrying out second processing on the collected surface electromyographic signals, and extracting corresponding characteristic data;
taking the motion data and the characteristic data as input data of a neural network of a recurrent cerebellum model, and carrying out training prediction to obtain a classification result;
the motion data includes classification parameters, and the first processing includes: calculating classification parameters corresponding to the gait harmony parameters based on the walking time phase; the classification parameters comprise support phase plantar pressure symmetry indexes, and the support phase plantar pressure symmetry indexes are obtained by calculation according to the following method:
dividing the supporting phase into a plurality of periods, and calculating a mean value matrix of N pressures in the plurality of periods according to a plurality of pressure data collected by the film pressure sensor, wherein the left supporting phase pressure value matrix is as follows:
Figure FDA0003594960380000011
wherein: pLIs a matrix of left-hand support phase pressure values, PghThe pressure value of the h support phase of the g film pressure sensor is g ═ 1, 2, …, N1;
calculating a left pressure value matrix PLL ofP1L of norm and right pressure value matrixP2Norm, LP1And LP2The calculation formula of (a) is as follows:
Figure FDA0003594960380000012
Figure FDA0003594960380000013
wherein, PRA right side support phase pressure value matrix is obtained;
calculating L of left and right support phase plantar pressure matrix in a plurality of periodsP1Norm and LP2The norm is used as the plantar pressure symmetry index of the support phase;
the second processing includes noise cancellation processing, and the extracting the corresponding feature data includes: performing dimensionality reduction processing on the data subjected to the denoising processing to obtain corresponding characteristic data; the characteristic data comprises an index of the symmetry of the support facies surface electromyographic signals, and the index of the symmetry of the support facies surface electromyographic signals is calculated according to the following mode:
acquiring a characteristic parameter matrix of a plurality of surface electromyographic signals of a support phase after dimension reduction treatment, wherein the characteristic parameters comprise average power frequency or integral electromyography; the left electromyographic signal matrix is as follows:
Figure FDA0003594960380000021
wherein: sLIs a left-side supporting phase electromyographic signal matrix, SghThe mean power frequency or integral myoelectricity of the z-th support phase of the t-th surface myoelectric signal, t is 1, 2, …, N2;
calculating L of the left-hand eigen parameter matrixS1Norm and L of right-hand eigen parameter matrixS2Norm to obtain the symmetry index of the support phase surface electromyogram signal, and the calculation formula is as follows:
Figure FDA0003594960380000022
wherein S isRA right side support phase electromyographic signal matrix.
2. The apparatus for predictive classification based on a surface electromyography signal and a motion signal of claim 1, wherein the motion signal comprises any one or more of a step size, a step width, a step angle, a step speed.
3. The prediction classification device based on the surface electromyogram signal and the motion signal according to claim 1, wherein the neural network of the recursive cerebellum model comprises an input layer, an association memory layer, a receiving domain layer, a weight memory layer and an output layer which are connected in sequence;
the input data of the input layer is the motion data and the feature data, and the output result of the output layer is the classification result;
the associative memory layer is a Gaussian wavelet function, the weight memory layer is used for providing weight values, and the receiving domain layer is used for providing convolution kernel sizes.
4. The apparatus for predictive classification based on surface electromyography and motion signals according to claim 3, wherein the Gaussian wavelet function is calculated as follows:
Figure FDA0003594960380000031
wherein x isikIs a signal input to the associative memory layer, bikAnd aikRespectively, a translation parameter and a dilation parameter, r, corresponding to said gaussian wavelet functionikOutputting the result for the associative memory layer;
wherein x isikIs represented as:
xik(t)＝xi(t)+wikrik(t-1) (5)
where t is the time series of data, wikIs the weight value of the recursion unit, which represents the influence of the output result at the previous moment on this moment, rikAnd (t-1) the output result at the previous time.
5. The apparatus for predictive classification based on surface electromyography and motion signals according to claim 4, wherein the output layer comprises an activation function, the activation function uses a sigmoid function, and the sigmoid function is calculated as follows:
Figure FDA0003594960380000032
where x is the signal input to the output layer;
the expression from the input layer to the output layer is:
Figure FDA0003594960380000033
wherein the function r () is a Gaussian wavelet function rik，IiInputting a feature vector; m is the feature vector dimension; w is aikIs a weight value between the input layer and the associative memory layer; said wkjThe weight value between the weight memory layer and the output layer, n represents the resolution ratio of the input vector characteristics, and o is the number of output layer classifications;
the improved recurrent cerebellar neural network model output layer activation function adopts a sigmoid function:
Figure FDA0003594960380000041
where x is the signal input to the output layer;
the expression from input layer to output layer is:
Figure FDA0003594960380000042
wherein IiInputting a feature vector, wherein m is a feature vector dimension; w is aikAnd wkjThe weights between the input layer and the associative memory layer and between the receptive field and the output layer are respectively; n characterizes the resolution of the input vector features; o is the number of output layer classifications.
CN202110547866.4A 2021-05-19 2021-05-19 Prediction classification method and device based on surface electromyogram signals and motion signals Active CN113274039B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110547866.4A CN113274039B (en) 2021-05-19 2021-05-19 Prediction classification method and device based on surface electromyogram signals and motion signals

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110547866.4A CN113274039B (en) 2021-05-19 2021-05-19 Prediction classification method and device based on surface electromyogram signals and motion signals

Publications (2)

Publication Number Publication Date
CN113274039A CN113274039A (en) 2021-08-20
CN113274039B true CN113274039B (en) 2022-06-07

Family

ID=77279995

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110547866.4A Active CN113274039B (en) 2021-05-19 2021-05-19 Prediction classification method and device based on surface electromyogram signals and motion signals

Country Status (1)

Country Link
CN (1) CN113274039B (en)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113662535B (en) * 2021-09-14 2022-07-01 合肥综合性国家科学中心人工智能研究院(安徽省人工智能实验室) Gait detection method, device, equipment and storage medium
CN114392126B (en) * 2022-01-24 2023-09-22 佳木斯大学 Disabled children's hand cooperation training system

Family Cites Families (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106821680A (en) * 2017-02-27 2017-06-13 浙江工业大学 A kind of upper limb healing ectoskeleton control method based on lower limb gait
CN106937872B (en) * 2017-04-20 2019-11-12 杭州电子科技大学 A kind of gait bilateral symmetric property evaluation method based on regression curve
CN110032987B (en) * 2019-04-23 2022-07-08 福州大学 Surface electromyographic signal classification method based on cerebellar neural network model
CN110232412B (en) * 2019-05-30 2020-12-11 清华大学 Human gait prediction method based on multi-mode deep learning
CN110710984B (en) * 2019-10-18 2021-11-02 福州大学 Ankle moment prediction method of recursion cerebellum model based on surface electromyogram signal
CN112120697A (en) * 2020-09-25 2020-12-25 福州大学 Muscle fatigue advanced prediction and classification method based on surface electromyographic signals

Also Published As

Publication number Publication date
CN113274039A (en) 2021-08-20

Similar Documents

Publication Publication Date Title
CN113274039B (en) Prediction classification method and device based on surface electromyogram signals and motion signals
Xiao et al. Motion estimation from surface electromyogram using adaboost regression and average feature values
Gurve et al. Subject-specific EEG channel selection using non-negative matrix factorization for lower-limb motor imagery recognition
Cui et al. A multimodal framework based on integration of cortical and muscular activities for decoding human intentions about lower limb motions
Lu et al. Continuous and simultaneous estimation of lower limb multi-joint angles from sEMG signals based on stacked convolutional and LSTM models
Ma et al. Design on intelligent perception system for lower limb rehabilitation exoskeleton robot
Zhou et al. Accurate recognition of lower limb ambulation mode based on surface electromyography and motion data using machine learning
Koch et al. A recurrent neural network for hand gesture recognition based on accelerometer data
Sbargoud et al. WPT-ANN and Belief Theory Based EEG/EMG Data Fusion for Movement Identification.
Vijayvargiya et al. Hybrid deep learning approaches for sEMG signal-based lower limb activity recognition
Rahman et al. Automated detection of rehabilitation exercise by stroke patients using 3-layer CNN-LSTM model
Rani et al. Surface electromyography and artificial intelligence for human activity recognition-A systematic review on methods, emerging trends applications, challenges, and future implementation
Dam et al. In-bed posture classification using pressure sensor data and spiking neural network
Wei et al. sEMG signal-based lower limb movements recognition using tunable Q-factor wavelet transform and Kraskov entropy
Siu et al. Ankle torque estimation during locomotion from surface electromyography and accelerometry
Chen et al. A novel imbalanced dataset mitigation method and ECG classification model based on combined 1D_CBAM-autoencoder and lightweight CNN model
Chee et al. Gait acceleration-based diabetes detection using hybrid deep learning
Zhang et al. Research on pattern recognition of lower limb motion based on convolutional neural network
Wang et al. Research on multimodal fusion recognition method of upper limb motion patterns
Li et al. Continuous angle prediction of lower limb knee joint based on semg
Sorkhabadi et al. Human locomotion activity and speed recognition using electromyography based features
Thu et al. Freezing of gait detection using discrete wavelet transform and hybrid deep learning architecture
Thu et al. An Investigation on Deep Learning-Based Activity Recognition Using IMUs and Stretch Sensors
Astudillo et al. Lower limbs motion intention detection by using pattern recognition
Vijayvargiya et al. Research Article Hybrid Deep Learning Approaches for sEMG Signal-Based Lower Limb Activity Recognition

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant