CN116369867B - Plantar ground reaction force prediction method and plantar ground reaction force prediction system based on WOA-1DCNN-LSTM - Google Patents

Plantar ground reaction force prediction method and plantar ground reaction force prediction system based on WOA-1DCNN-LSTM Download PDF

Info

Publication number
CN116369867B
CN116369867B CN202310657622.0A CN202310657622A CN116369867B CN 116369867 B CN116369867 B CN 116369867B CN 202310657622 A CN202310657622 A CN 202310657622A CN 116369867 B CN116369867 B CN 116369867B
Authority
CN
China
Prior art keywords
woa
whale
lstm
1dcnn
data
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
CN202310657622.0A
Other languages
Chinese (zh)
Other versions
CN116369867A (en
Inventor
戴厚德
陈昱光
连阳林
邓盛中
武泠燏
黄巧园
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.)
Quanzhou Institute of Equipment Manufacturing
Original Assignee
Quanzhou Institute of Equipment Manufacturing
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 Quanzhou Institute of Equipment Manufacturing filed Critical Quanzhou Institute of Equipment Manufacturing
Priority to CN202310657622.0A priority Critical patent/CN116369867B/en
Publication of CN116369867A publication Critical patent/CN116369867A/en
Application granted granted Critical
Publication of CN116369867B publication Critical patent/CN116369867B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • 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
    • 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
    • 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
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
    • 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
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • 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
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/0464—Convolutional networks [CNN, ConvNet]
    • 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
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/048—Activation functions
    • 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
    • G06N3/086—Learning methods using evolutionary algorithms, e.g. genetic algorithms or genetic programming
    • 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
    • G06N3/0985—Hyperparameter optimisation; Meta-learning; Learning-to-learn

Landscapes

  • Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Theoretical Computer Science (AREA)
  • Biomedical Technology (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Biophysics (AREA)
  • Artificial Intelligence (AREA)
  • General Physics & Mathematics (AREA)
  • Software Systems (AREA)
  • Mathematical Physics (AREA)
  • General Engineering & Computer Science (AREA)
  • Computing Systems (AREA)
  • Evolutionary Computation (AREA)
  • Data Mining & Analysis (AREA)
  • Computational Linguistics (AREA)
  • Medical Informatics (AREA)
  • Veterinary Medicine (AREA)
  • Pathology (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Public Health (AREA)
  • Animal Behavior & Ethology (AREA)
  • Surgery (AREA)
  • Physiology (AREA)
  • Psychiatry (AREA)
  • Signal Processing (AREA)
  • Oral & Maxillofacial Surgery (AREA)
  • Dentistry (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Footwear And Its Accessory, Manufacturing Method And Apparatuses (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention relates to the technical field of neural network prediction, in particular to a plantar ground reaction force prediction method and system based on WOA-1 DCNN-LSTM. A plantar ground reaction force prediction method and system based on WOA-1DCNN-LSTM comprises S1, data acquisition; s2, preprocessing data; s3, data set distribution; s4, building a WOA-1DCNN-LSTM neural network model; s5, outputting data. According to the invention, by arranging the intelligent insole with the sensor, the walking posture of the user is monitored, the ground reaction force generated next is predicted, the WOA-1DCNN-LSTM neural network model is arranged in the prediction system, and the method for predicting the ground reaction force by introducing the neural network model is used for realizing long-time and long-distance measurement of the intelligent insole, and the user can obtain timely feedback, so that the accuracy of the obtained feedback result is high.

Description

Plantar ground reaction force prediction method and plantar ground reaction force prediction system based on WOA-1DCNN-LSTM
Technical Field
The invention relates to the technical field of neural network prediction, in particular to a plantar ground reaction force prediction method and system based on WOA-1 DCNN-LSTM.
Background
The plantar total ground reaction force (Ground Reaction Force, GRF) in the human body movement state comprises Vertical force (Fv), anterior-posterior shear force (analytical-posterior shear force, F) AP ) And Medial-lateral shear force (media-lateral shear force, F ML ) A schematic diagram of the plantar total ground reaction force space is shown in fig. 2. Only the relationship between vertical pressure and foot conditions is often of concern in previous studies. In recent years, however, it has been increasingly recognized that foot disorders are associated with not only plantar vertical pressure, but also plantar shear forces, such as motor blisters, diabetic foot ulcers, hallux valgus, and the like. Shear forces refer to forces in the medial-lateral direction perpendicular to the coronal plane and forces in the anterior-posterior direction perpendicular to the sagittal plane. When a shear force acts on the deep layer, it will cause a relative displacement of the tissue, cutting off the supply of large-area small blood vessels, resulting in a decrease in tissue oxygen tension. Therefore, sometimes the shearing force is more serious than the vertical pressure to the human body. In order to avoid the partial lesion of the sole caused by the long-term incorrect walking posture, the pressure and the shearing force of the sole can be monitored to monitor the change condition of the reaction of the sole ground in three directions during the walking of a human body, and the physical problem during the walking is found out through the data result, so that the unhealthy walking mode is corrected.
The traditional sole three-dimensional reaction force monitoring method is based on a pressure force measuring flat plate, the method can only be used for measuring in a laboratory, long distance and measurement in multiple environments cannot be realized, and the method for predicting the complete ground reaction force through the pressure force measuring flat plate is mostly a linear fitting method, and the method has a good fitting effect on vertical pressure, but has a poor fitting effect on shearing force of front side, rear side, inner side and outer side, and cannot adapt to the prediction requirement in the actual walking process.
Disclosure of Invention
Additional features and advantages of the invention will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. The objectives and other advantages of the invention may be realized and attained by the structure particularly pointed out in the written description and the appended drawings.
The invention aims to overcome the defects, and provides a plantar ground reaction force prediction method and a plantar ground reaction force prediction system based on WOA-1 DCNN-LSTM.
The invention provides a plantar ground reaction force prediction method based on WOA-1DCNN-LSTM, which comprises the following steps:
s1, data acquisition: arrange capacitive pressure sensor, this capacitive pressure sensor sets up on intelligent shoe-pad and is array arrangement, gathers the pressure data that the sole produced to ground when walking, including vertical direction pressure F V Shear force F in front-rear direction AP Shear force F in the inner and outer directions ML ；
S2, data preprocessing: selecting the first non-zero data to the last non-zero data, normalizing each data, and establishing a data set;
s3, dividing the data set into a training set, a testing set and a verification set according to a proportion;
s4, building a WOA-1DCNN-LSTM neural network model, wherein the model comprises a whale optimization algorithm WOA, a one-dimensional convolutional neural network 1DCNN and a long and short time memory network LSTM;
the one-dimensional convolutional neural network 1DCNN is used for extracting spatial features of a data set with plantar gait features to obtain time sequence data required by the long-short-term memory network LSTM; the long-short-term memory network LSTM performs characteristic processing and predictive fitting on the time sequence data to obtain predicted plantar ground reaction force; the whale optimization algorithm WOA is used for optimizing model parameters in the long-short-term memory network LSTM, and the prediction accuracy is improved;
s5, outputting prediction data, and visually evaluating the prediction effect of the model.
In some embodiments, the whale optimization algorithm WOA is essentially a process of searching for an optimal solution in a solution set, where one solution is represented by one whale individual, a plurality of solutions can be represented by a plurality of whale individuals, and searching for the optimal solution using the whale optimization algorithm WOA can be regarded as a process that the plurality of whale individuals continuously update their individual locations until a satisfactory solution is searched.
In some embodiments, the whale optimization algorithm WOA includes three update methods, namely, surround predation, random search, and spiral bubble search, the mathematical model of which is as follows:
surrounding predation:
A＝2a(r 1 -1) (2)
C＝2ar 2 (3)
a＝2-2t/t max (4)
Wherein the method comprises the steps ofIs the ith whale in the t-th iteration,>is the best whale after the t-th cycle, A and C are matrix coefficients, a is the convergence factor, t max Is the maximum iteration number, r 1 And r 2 Is [0,1 ]]Random vectors within the range;
random search:
wherein the method comprises the steps ofIs random whales in the population at the t iteration;
spiral bubble search:
wherein the method comprises the steps ofRepresenting the distance between the ith whale and the best whale in the t-th iteration, the spiral shape constant b is typically set to 1, l is [ -1,1]Depending on the distance between the individual and the best whale;
for the whale optimization algorithm WOA, when the predation strategy probability p is less than 0.5, selecting an updating method by the whale group according to |A|; when the A is less than or equal to 1, the surrounding predation is adopted to realize local development; when the shrink wrap mechanism is not satisfied, the whale group is no longer updated according to the optimal whale position, but performs a random search to achieve a global search; when p is equal to or greater than 0.5, the whale moves in a spiral motion to the optimal whale position and bubbles are generated to surround the game to complete hunting, i.e., spiral bubble updating is performed.
In some embodiments, the specific operation steps of S4 are:
s41: setting the initial parameters of the one-dimensional convolutional neural network 1 DCNN;
s42, determining the model layer number of the long-short-time memory network LSTM, and taking the node number and the learning rate of two hidden layers in the long-short-time memory network LSTM as optimizing super parameters;
s43, WOA algorithm initialization, namely setting the iteration times as 100, setting the initial population quantity as N, selecting whale population positions by using a reverse learning strategy, and taking root mean square error of a test set as a loss function;
s44, calculating the fitness value of each whale, and storing the current optimal individual and position;
s45, updating A, C, l, p, r and r2 when t is less than tmax;
for p <0.5, if |A|+.1, updating the whale position according to (formula 1); otherwise, if |a| >1, updating the whale position according to (formula 5);
updating the whale position according to (formula 6) for p.gtoreq.0.5;
calculating the optimal individual in the current group, storing the individual position, and updating the optimal individual position through a cross optimization algorithm;
s46, outputting an optimal individual, wherein the optimal individual value is the optimal super parameter of the model of the long-short-time memory network LSTM;
s47: and (3) giving the value of the optimal super parameter to a 1DCNN-LSTM model, so as to establish a mature WOA-1DCNN-LSTM model, and obtaining a prediction result of the reaction force after the capacitive pressure sensor inputs data.
In some embodiments, the one-dimensional convolutional neural network 1DCNN is composed of an input layer, a convolutional layer, a pooling layer, a full-connection layer, and an output layer.
In some examples, the long-short-term memory network LSTM is comprised of a forget gate, an input gate, and an output gate, and the number of neuronal nodes and the learning rate of the long-short-term memory network LSTM are selected by the whale optimization algorithm WOA adaptive update.
In some embodiments, the first 70% of the data of the dataset is selected as the training set, 20% of the data is selected as the test set, and 10% of the data is selected as the validation set.
A plantar ground reaction force prediction system based on WOA-1DCNN-LSTM comprises an intelligent insole and upper computer software;
the intelligent insole is provided with a plurality of capacitive pressure sensors which are arranged in an array manner, and the capacitive pressure sensors distributed in the array manner can obtain the change condition of the plantar array pressure;
the upper computer software receives pressure data from the intelligent insole in a Bluetooth transmission mode, a mature WOA-1DCNN-LSTM neural network model is arranged in the upper computer software, and when the upper computer software receives pressure change data from the intelligent insole, data processing can be carried out on the pressure change data to obtain a prediction result of the complete ground reaction force of the sole.
In some embodiments, the number of the capacitive pressure sensors provided on the intelligent insole is 13, and the number of the capacitive pressure sensors is S1, S2, S3, S13, and each capacitive pressure sensor can obtain the vertical pressure FV, the front-rear side shearing force FAP, and the medial-lateral shearing force FML.
In some embodiments, the prediction system further comprises a host computer system, and the host computer software can be applied at a mobile terminal and a PC terminal of the mobile phone for displaying pressure data of the intelligent insole and ground reaction force predicted by the neural network model.
By adopting the technical scheme, the invention has the beneficial effects that:
according to the intelligent insole provided by the invention, the walking posture of a user is monitored and the ground reaction force generated next is predicted by arranging the intelligent insole with the sensor, so that the user can adjust the walking posture according to the prediction information to prevent the plantar lesion caused by incorrect posture, the prediction system is provided with the WOA-1DCNN-LSTM neural network model, the one-dimensional convolutional neural network 1DCNN is used for extracting the characteristics of plantar pressure data, the long-short-term memory network LSTM integrates and predicts the data output by the one-dimensional convolutional neural network 1DCNN, the whale optimization algorithm WOA is used for optimizing the model parameters in the long-short-term memory network LSTM, the model prediction accuracy is improved, and the long-term and long-distance measurement of the intelligent insole is realized by introducing the neural network model to predict the ground reaction force, the user can obtain timely feedback, and the obtained feedback result is high in accuracy.
It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure.
It is apparent that these and other objects of the present invention will become more apparent from the following detailed description of the preferred embodiments, which is to be read in connection with the accompanying drawings and figures.
The foregoing and other objects, features and advantages of the invention will be apparent from the following more particular description of the preferred embodiments, as illustrated in the accompanying drawings.
Drawings
The accompanying drawings are included to provide a further understanding of the invention, and are incorporated in and constitute a part of this specification, illustrate the invention and together with the embodiments of the invention, serve to explain the invention, without limitation to the invention.
In the drawings, like parts are designated with like reference numerals and are illustrated schematically and are not necessarily drawn to scale.
In order to more clearly illustrate the embodiments of the invention or the technical solutions in the prior art, the following description will briefly explain the drawings used in the embodiments or the description of the prior art, and it is obvious that the drawings in the following description are only one or several embodiments of the invention, and that other drawings can be obtained according to such drawings without inventive effort to a person skilled in the art.
FIG. 1 is a schematic overall flow chart of a prediction method according to some embodiments of the present invention;
FIG. 2 is a schematic representation of a three-dimensional reaction force spatial distribution in some embodiments of the invention;
FIG. 3 is a schematic diagram of a 1DCNN structure in accordance with some embodiments of the invention;
FIG. 4 is a schematic diagram of LSTM cell structures according to some embodiments of the invention;
FIG. 5 is a schematic illustration of a WOA predation process in some embodiments of the invention;
FIG. 6 is a schematic diagram of the overall structure of a WOA-1DCNN-LSTM neural network in accordance with some embodiments of the invention;
FIG. 7 is a schematic diagram of an array distribution of thirteen capacitive pressure sensors in some embodiments of the invention;
FIG. 8 is a schematic overall flow diagram of a prediction system according to some embodiments of the invention;
FIG. 9 is a diagram illustrating predicted reaction force results according to some embodiments of the invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention will be further described in detail with reference to the following detailed description. It should be understood that the detailed description is presented merely to illustrate the invention, and is not intended to limit the invention.
In addition, in the description of the present invention, it should be understood that the terms "center", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", "axial", "radial", "circumferential", etc. indicate orientations or positional relationships based on the drawings, are merely for convenience in describing the present invention and simplifying the description, and do not indicate or imply that the devices or elements referred to must have a specific orientation, be configured and operated in a specific orientation, and thus should not be construed as limiting the present invention.
In the present invention, unless explicitly specified and limited otherwise, the terms "mounted," "connected," "secured," and the like are to be construed broadly, and may be, for example, fixedly connected, detachably connected, or integrally formed; can be directly connected or indirectly connected through an intermediate medium, and can be communicated with the inside of two elements or the interaction relationship of the two elements. However, it is noted that direct connection indicates that the two bodies connected together do not form a connection relationship through a transition structure, but are connected together to form a whole through a connection structure. The specific meaning of the above terms in the present invention can be understood by those of ordinary skill in the art according to the specific circumstances.
In the present invention, unless expressly stated or limited otherwise, a first feature "up" or "down" a second feature may be the first and second features in direct contact, or the first and second features in indirect contact via an intervening medium. In the description of the present specification, a description referring to terms "one embodiment," "some embodiments," "examples," "specific examples," or "some examples," etc., means that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the present invention. In this specification, schematic representations of the above terms are not necessarily directed to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.
Referring to fig. 1-4, fig. 1 is a schematic overall flow chart of plantar ground reaction force prediction in some embodiments of the invention; FIG. 2 is a schematic representation of a three-dimensional reaction force spatial distribution in some embodiments of the invention; FIG. 3 is a schematic diagram of a 1DCNN structure in accordance with some embodiments of the invention; fig. 4 is a schematic diagram of an LSTM cell according to some embodiments of the present invention.
According to some embodiments of the present invention, there is provided a method for predicting plantar surface reaction force based on WOA-1DCNN-LSTM, comprising:
s1, data acquisition: arrange capacitive pressure sensor, this capacitive pressure sensor sets up on intelligent shoe-pad and is the array and arranges, gathers the pressure data that the plantar produced to ground when walking, including vertical direction pressure FV, front and back side direction shearing force FAP, inside and outside direction shearing force F ML ；
Resistive pressure sensors and capacitive pressure sensors are often applied to designing intelligent wearable insoles, but intelligent insoles designed by using piezoresistive pressure sensors generally have the defects of low precision, high power consumption, easy breakage and the like, so that the accuracy of data input is improved by adopting the capacitive pressure sensors.
S2, data preprocessing: selecting the first non-zero data to the last non-zero data, normalizing each data, and establishing a data set;
since only the data when the foot wearing the intelligent insole is in contact with the ground is valid data, the input data needs to be preprocessed after the course of action is finished, and the valid data range is from the beginning of the first non-zero data to the end of the last non-zero data.
S3, dividing the data set into a training set, a testing set and a verification set according to a proportion; selecting the first 70% of data of the data set as a training set, 20% of data as a test set and 10% of data as a verification set;
s4, building a WOA-1DCNN-LSTM neural network model, wherein the model comprises a whale optimization algorithm WOA, a one-dimensional convolutional neural network 1DCNN and a long and short time memory network LSTM;
the one-dimensional convolutional neural network 1DCNN is used for extracting spatial features of a data set with plantar gait features to obtain time sequence data required by the long-short-term memory network LSTM; the long-short-term memory network LSTM performs characteristic processing and predictive fitting on the time sequence data to obtain predicted plantar ground reaction force; the whale optimization algorithm WOA is used for optimizing model parameters in the long-short-term memory network LSTM, and the prediction accuracy is improved;
according to some embodiments of the invention, the one-dimensional convolutional neural network 1DCNN is optionally composed of an input layer, a convolutional layer, a pooling layer, a fully-connected layer, and an output layer.
The method has the advantages that the method adopts a local connection and weight sharing mode to reduce the risk of network overfitting, and meanwhile, compared with the traditional method, the convolutional neural network can deeply mine the internal relation of data and improve the quality of the extracted feature vectors.
In the prior art, 2DCNN is often adopted for processing the image problem, but the plantar gait feature information in the invention is one-dimensional data based on a time sequence, and 1DCNN is selected to have better expressive performance, so that not only can the time sequence features of the data be reserved as much as possible, but also the computational complexity can be reduced, the commonly used 1DCNN structure is shown in fig. 3, the input feature data is subjected to rolling and pooling, then is subjected to unfolding dimension reduction processing, and finally is subjected to an activation function to obtain the time sequence data required by LSTM.
According to some embodiments of the invention, optionally, the long-short-term memory network LSTM is composed of a forgetting gate, an input gate and an output gate, and the number of neuronal nodes and the learning rate of the long-short-term memory network LSTM are selected by the whale optimization algorithm WOA adaptive update.
In the training process of the traditional cyclic neural network, the problems of gradient disappearance and gradient explosion are easy to occur along with the increase of the number of network layers and the increase of iteration times, the long-and-short-term memory neural network LSTM designs various gate structures on the basis of the cyclic neural network, wherein the structures comprise forgetting gates, input gates and output gates, the number of the passed characteristics is controlled, the structure diagram of an LSTM unit is shown as a figure 4, the forgetting gates can discard redundant information in data, the input gates can retain key data in the data, and the output gates can be used for selecting important data output; therefore, the long-short-term memory network can ensure that effective information is reserved and updated for a long time through the operation of three gates on data, and the problems of gradient explosion and gradient disappearance are avoided.
According to some embodiments of the present invention, the selection of the initial parameters of the LSTM prediction model may have a great influence on the prediction performance, so many researchers have studied the optimization of the LSTM parameters, and the research results indicate that the number of hidden layer nodes and the learning rate are the most critical super parameters; values that are too small or too large will have a large impact on the final predicted effect, so the range of parameters can be selected empirically first, and then the optimal super-parameters can be selected according to an optimization algorithm.
The whale optimization algorithm WOA is essentially a process of searching for an optimal solution in a solution set, wherein one solution is represented by one whale individual, a plurality of solutions can be represented by a plurality of whale individuals, and the process of searching for the optimal solution by using the whale optimization algorithm WOA can be regarded as that the plurality of whale individuals continuously update the individual positions until a satisfactory solution is searched.
The whale optimization algorithm WOA comprises three updating methods, namely surround predation, random search and spiral bubble search, wherein mathematical models of the three updating methods are as follows:
surrounding predation:
A＝2a(r 1 -1) (2)
C＝2ar 2 (3)
a＝2-2t/t max (4)
Wherein the method comprises the steps ofIs the ith whale in the t-th iteration,>is the best whale after the t-th cycle, A and C are matrix coefficients, a is the convergence factor, t max Is the maximum iteration number, r 1 And r 2 Is [0,1 ]]Random vectors within the range;
random search:
wherein the method comprises the steps ofIs random whales in the population at the t iteration;
spiral bubble search:
wherein the method comprises the steps ofRepresenting the distance between the ith whale and the best whale in the t-th iteration, the spiral shape constant b is typically set to 1, l is [ -1,1]Depending on the distance between the individual and the best whale;
referring to fig. 5, for the whale optimization algorithm WOA, when the predation strategy probability p <0.5, the whale group selects the update method according to |a|; when the A is less than or equal to 1, the surrounding predation is adopted to realize local development; when the shrink wrap mechanism is not satisfied, the whale group is no longer updated according to the optimal whale position, but performs a random search to achieve a global search; when p is equal to or greater than 0.5, the whale moves in a spiral motion to the optimal whale position and bubbles are generated to surround the game to complete hunting, i.e., spiral bubble updating is performed.
S5, outputting prediction data, and visually evaluating the prediction effect of the model.
In summary, a complete structure of the WOA-1DCNN-LSTM model is obtained, and referring to FIG. 6, the model is used for integrating the advantages of the CNN neural network and the LSTM model aiming at the characteristics of data acquisition of the intelligent insole, has good characteristic extraction and prediction fitting effects on time series data, and simultaneously is used for solving the problem that incorrect parameter selection of the LSTM model greatly influences a prediction result, and optimizing the node number and learning rate of two hidden layers of the LSTM model by using a whale optimization algorithm so as to enable the node number and learning rate to be more matched with the data of an insole sensor.
The specific operation steps of the S4 are as follows:
s41: setting the initial parameters of the one-dimensional convolutional neural network 1 DCNN;
s42, determining the model layer number of the long-short-time memory network LSTM, and taking the node number and the learning rate of two hidden layers in the long-short-time memory network LSTM as optimizing super parameters;
s43, WOA algorithm initialization, namely setting the iteration times as 100, setting the initial population quantity as N, selecting whale population positions by using a reverse learning strategy, and taking root mean square error of a test set as a loss function;
s44, calculating the fitness value of each whale, and storing the current optimal individual and position;
s45, updating A, C, l, p, r and r2 when t is less than tmax;
for p <0.5, if |A|+.1, updating the whale position according to (formula 1); otherwise, if |a| >1, updating the whale position according to (formula 5);
updating the whale position according to (formula 6) for p.gtoreq.0.5;
calculating the optimal individual in the current group, storing the individual position, and updating the optimal individual position through a cross optimization algorithm;
s46, outputting an optimal individual, wherein the optimal individual value is the optimal super parameter of the model of the long-short-time memory network LSTM;
s47: and (3) giving the value of the optimal super parameter to a 1DCNN-LSTM model, so as to establish a mature WOA-1DCNN-LSTM model, and obtaining a prediction result of the reaction force after the capacitive pressure sensor inputs data.
Referring to fig. 7-8, fig. 7 is a schematic diagram of an array distribution of thirteen capacitive pressure sensors in some embodiments of the invention; FIG. 8 is a schematic overall flow diagram of a prediction system according to some embodiments of the invention.
The invention provides a plantar ground reaction force prediction system based on WOA-1DCNN-LSTM, which comprises an intelligent insole and upper computer software;
the intelligent insole is provided with a plurality of capacitive pressure sensors which are arranged in an array manner, and the capacitive pressure sensors distributed in the array manner can obtain the change condition of the plantar array pressure;
the upper computer software receives pressure data from the intelligent insole in a Bluetooth transmission mode, a mature WOA-1DCNN-LSTM neural network model is arranged in the upper computer software, and when the upper computer software receives pressure change data from the intelligent insole, data processing can be carried out on the pressure change data to obtain a prediction result of the complete ground reaction force of the sole.
The upper computer software can be applied to a mobile terminal and a PC terminal of the mobile phone and is used for displaying pressure data of the intelligent insole and ground reaction force predicted by the neural network model.
The number of the capacitive pressure sensors arranged on the intelligent insole is 13, the specific distribution conditions are shown in figure 7, S1, S2 and S3 are respectively S13, and each capacitive pressure sensor can obtain the pressure F in the vertical direction V The front-rear side shearing force F AP The shear force F in the inner and outer directions ML 。
Referring to fig. 9, fig. 9 is a schematic diagram of a reaction force prediction result in some embodiments of the invention.
The whole plantar ground reaction force prediction system based on WOA-1DCNN-LSTM is tested, thirteen capacitive pressure sensor array type pressure change values are selected as output, and ground reaction forces in three directions are output. And the Pelson correlation coefficient (R) is selected as an evaluation standard of a prediction result, the test result is compared with the RBF neural network test result after the radial basis function neural network RBFNN, the particle swarm and the genetic algorithm are optimized, and the prediction result of the plantar ground reaction force prediction system based on the WOA-1DCNN-LSTM on the ground reaction force is more accurate from the test result.
It is to be understood that the disclosed embodiments are not limited to the specific process steps or materials disclosed herein, but are intended to extend to equivalents of such features as would be understood by one of ordinary skill in the relevant arts. It is also to be understood that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting.
Reference in the specification to "an embodiment" means that a particular feature, or characteristic, described in connection with the embodiment is included in at least one embodiment of the invention. Thus, appearances of the phrase or "an embodiment" in various places throughout this specification are not necessarily all referring to the same embodiment.
Furthermore, the described features or characteristics may be combined in any other suitable manner in one or more embodiments. In the above description, certain specific details are provided, such as thicknesses, numbers, etc., to provide a thorough understanding of embodiments of the invention. One skilled in the relevant art will recognize, however, that the invention can be practiced without one or more of the specific details, or with other methods, components, materials, etc.

Claims (10)

1. A WOA-1DCNN-LSTM based plantar surface reaction force prediction method, comprising:
s1, data acquisition: arrange a plurality of capacitive pressure sensor, this capacitive pressure sensor sets up on intelligent shoe-pad and is the array and arrange, gathers the pressure data that the sole produced to ground when walking, including vertical direction pressure F V Shear force F in front-rear direction AP Shear force F in the inner and outer directions ML ；
S2, data preprocessing: selecting the first non-zero data to the last non-zero data, normalizing each data, and establishing a data set;
s3, dividing the data set into a training set, a testing set and a verification set according to a proportion;
s4, building a WOA-1DCNN-LSTM neural network model, wherein the model comprises a whale optimization algorithm WOA, a one-dimensional convolutional neural network 1DCNN and a long and short time memory network LSTM;
the one-dimensional convolutional neural network 1DCNN is used for extracting spatial features of a data set with plantar gait features to obtain time sequence data required by the long-short-term memory network LSTM; the long-short-term memory network LSTM performs characteristic processing and predictive fitting on the time sequence data to obtain predicted plantar ground reaction force; the whale optimization algorithm WOA is used for optimizing model parameters in the long-short-term memory network LSTM, and the prediction accuracy is improved;
s5, outputting prediction data, and visually evaluating the prediction effect of the model.
2. The WOA-1DCNN-LSTM based plantar surface reaction force prediction method according to claim 1, wherein the whale optimization algorithm WOA is essentially a process of searching for an optimal solution in a solution set, wherein one solution is represented by one whale individual, a plurality of solutions can be represented by a plurality of whale individuals, and the process of searching for the optimal solution using the whale optimization algorithm WOA can be regarded as that a plurality of whale individuals continuously update their individual positions until a satisfactory solution is searched.
3. The WOA-1DCNN-LSTM based plantar surface reaction force prediction method according to claim 2, characterized in that the whale optimization algorithm WOA includes three updating methods, namely, surround predation, random search and spiral bubble search, the mathematical model of the three updating methods is as follows:
surrounding predation:
A＝2a(r 1 -1) (2)
C＝2ar 2 (3)
a＝2-2t/t max (4)
Wherein isIth whale in the t-th iteration, < >>Is the best whale after the t-th cycle, A and C are matrix coefficients, a is the convergence factor, t max Is the maximum iteration number, r 1 And r 2 Is [0,1 ]]Random vectors within the range;
random search:
wherein the method comprises the steps ofIs random whales in the population at the t iteration;
spiral bubble search:
wherein the method comprises the steps ofRepresenting the distance between the ith whale and the best whale in the t-th iteration, the spiral shape constant b is set to 1, l is [ -1,1]Depending on the distance between the individual and the best whale;
for the whale optimization algorithm WOA, when the predation strategy probability p is less than 0.5, selecting an updating method by the whale group according to |A|; when the A is less than or equal to 1, the surrounding predation is adopted to realize local development; when the shrink wrap mechanism is not satisfied, the whale group is no longer updated according to the optimal whale position, but performs a random search to achieve a global search; when p is equal to or greater than 0.5, the whale moves in a spiral motion to the optimal whale position and bubbles are generated to surround the game to complete hunting, i.e., spiral bubble updating is performed.
4. The WOA-1DCNN-LSTM based plantar surface reaction force prediction method according to claim 3, wherein the specific operation steps of S4 are as follows:
s41: setting the initial parameters of the one-dimensional convolutional neural network 1 DCNN;
s42, determining the model layer number of the long-short-time memory network LSTM, and taking the node number and the learning rate of two hidden layers in the long-short-time memory network LSTM as optimizing super parameters;
s43, WOA algorithm initialization, namely setting the iteration times as 100, setting the initial population quantity as N, selecting whale population positions by using a reverse learning strategy, and taking root mean square error of a test set as a loss function;
s44, calculating the fitness value of each whale, and storing the current optimal individual and position;
s45, updating A, C, l, p, r and r2 when t is less than tmax;
for p <0.5, if |A|+.1, updating the whale position according to formula 1; otherwise, if |a| >1, updating the whale position according to equation 5;
for p.gtoreq.0.5, updating the whale position according to formula 6;
calculating the optimal individual in the current group, storing the individual position, and updating the optimal individual position through a cross optimization algorithm;
s46, outputting an optimal individual, wherein the optimal individual value is the optimal super parameter of the model of the long-short-time memory network LSTM;
s47: and (3) giving the value of the optimal super parameter to a 1DCNN-LSTM model, so as to establish a mature WOA-1DCNN-LSTM model, and obtaining a prediction result of the reaction force after the capacitive pressure sensor inputs data.
5. The WOA-1DCNN-LSTM based plantar surface reaction force prediction method according to claim 1, wherein the one-dimensional convolutional neural network 1DCNN is composed of an input layer, a convolutional layer, a pooling layer, a full-connection layer, and an output layer.
6. The WOA-1DCNN-LSTM based plantar surface reaction force prediction method according to claim 1, wherein the long-short-term memory network LSTM is composed of a forgetting gate, an input gate and an output gate, and the number of neuronal nodes and the learning rate of the long-short-term memory network LSTM are selected by the whale optimization algorithm WOA adaptive update.
7. The WOA-1DCNN-LSTM based plantar surface reaction force prediction method as defined in claim 1, wherein the first 70% of the data set is selected as a training set, 20% of the data is selected as a test set, and 10% of the data is selected as a verification set.
8. A WOA-1DCNN-LSTM based plantar surface reaction force prediction system that performs the method of any one of claims 1-7, comprising
The intelligent insole is provided with a plurality of capacitive pressure sensors which are arranged in an array manner, and the capacitive pressure sensors which are arranged in the array manner can obtain the change condition of the array pressure of the sole;
the upper computer software receives pressure data from the intelligent insoles in a Bluetooth transmission mode, a mature WOA-1DCNN-LSTM neural network model is arranged in the upper computer software, and when the upper computer software receives pressure change data from the intelligent insoles, data processing can be carried out on the pressure change data to obtain a prediction result of the full ground reaction force of the soles.
9. The WOA-1DCNN-LSTM based plantar surface reaction force prediction system as defined in claim 8, wherein the number of capacitive pressure sensors provided on the intelligent insole is 13, S1, S2, S3S 13, each of the capacitive pressure sensors can obtain the vertical pressure F V The front-rear side shearing force F AP The shear force F in the inner and outer directions ML 。
10. The WOA-1DCNN-LSTM based plantar surface reaction force prediction system of claim 8, wherein the host computer software is capable of being applied to a mobile terminal and a PC terminal of a mobile phone, and is used for displaying pressure data of the intelligent insole and the ground reaction force predicted by the neural network model.
CN202310657622.0A 2023-06-06 2023-06-06 Plantar ground reaction force prediction method and plantar ground reaction force prediction system based on WOA-1DCNN-LSTM Active CN116369867B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202310657622.0A CN116369867B (en) 2023-06-06 2023-06-06 Plantar ground reaction force prediction method and plantar ground reaction force prediction system based on WOA-1DCNN-LSTM

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202310657622.0A CN116369867B (en) 2023-06-06 2023-06-06 Plantar ground reaction force prediction method and plantar ground reaction force prediction system based on WOA-1DCNN-LSTM

Publications (2)

Publication Number Publication Date
CN116369867A CN116369867A (en) 2023-07-04
CN116369867B true CN116369867B (en) 2023-11-21

Family

ID=86961947

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202310657622.0A Active CN116369867B (en) 2023-06-06 2023-06-06 Plantar ground reaction force prediction method and plantar ground reaction force prediction system based on WOA-1DCNN-LSTM

Country Status (1)

Country Link
CN (1) CN116369867B (en)

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR101502235B1 (en) * 2013-10-25 2015-03-12 성균관대학교산학협력단 Method for deriving ground reaction force predicting model and ground reaction force predicting apparatus using the same
CN114741952A (en) * 2022-03-11 2022-07-12 江苏大学 Short-term load prediction method based on long-term and short-term memory network
CN114912673A (en) * 2022-04-29 2022-08-16 河海大学 Water level prediction method based on whale optimization algorithm and long-term and short-term memory network
CN115759321A (en) * 2022-09-14 2023-03-07 国网山西省电力公司大同供电公司 Intelligent reactive power compensation method taking minimum line loss as optimization target

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112653142B (en) * 2020-12-18 2022-09-02 武汉大学 Wind power prediction method and system for optimizing depth transform network

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR101502235B1 (en) * 2013-10-25 2015-03-12 성균관대학교산학협력단 Method for deriving ground reaction force predicting model and ground reaction force predicting apparatus using the same
CN114741952A (en) * 2022-03-11 2022-07-12 江苏大学 Short-term load prediction method based on long-term and short-term memory network
CN114912673A (en) * 2022-04-29 2022-08-16 河海大学 Water level prediction method based on whale optimization algorithm and long-term and short-term memory network
CN115759321A (en) * 2022-09-14 2023-03-07 国网山西省电力公司大同供电公司 Intelligent reactive power compensation method taking minimum line loss as optimization target

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
A Deep Learning Model for 3D Ground Reaction Force Estimation Using Shoes with Three Uniaxial Load Cells;Junggil Kim等;《Sensors》;第1-13页 *

Also Published As

Publication number Publication date
CN116369867A (en) 2023-07-04

Similar Documents

Publication Publication Date Title
CN107767946B (en) Breast cancer diagnosis system based on PCA (principal component analysis) and PSO-KELM (particle swarm optimization-KeLM) models
CN111161879B (en) Disease prediction system based on big data
CN111667884A (en) Convolutional neural network model for predicting protein interactions using protein primary sequences based on attention mechanism
CN103064283B (en) A kind of self study wheelchair control method based on gravity center of human body&#39;s change
CN108731732B (en) Method and system for processing plantar pressure based on periodic modal information
US20060129034A1 (en) Medical decision support systems utilizing gene expression and clinical information and method for use
JP6998518B2 (en) Motor function estimation information generation device, motor function estimation system, motor function estimation information generation method, motor function estimation method and recording medium
CN112259229A (en) Method, system and equipment for realizing traditional Chinese medicine diagnosis and treatment based on neural network
CN110893100A (en) Device and method for monitoring posture change based on plantar pressure sensor
CN112509696A (en) Health data detection method based on convolution autoencoder Gaussian mixture model
CN114203295B (en) Cerebral apoplexy risk prediction intervention method and system
CN116369867B (en) Plantar ground reaction force prediction method and plantar ground reaction force prediction system based on WOA-1DCNN-LSTM
CN109567313B (en) Intelligent insole with biological characteristic recognition function
CN112712895B (en) Data analysis method of multi-modal big data aiming at type 2 diabetes complications
Nipas et al. Burned Calories Prediction using Supervised Machine Learning: Regression Algorithm
CN106056167A (en) Normalization possibilistic fuzzy entropy clustering method based on Gaussian kernel hybrid artificial bee colony algorithm
CN117275661A (en) Deep reinforcement learning-based lung cancer patient medication prediction method and device
CN115147768A (en) Fall risk assessment method and system
CN116164880A (en) CNN-GRU neural network-based pressure sensor fault detection method
CN114504298A (en) Physiological feature distinguishing method and system based on multi-source health perception data fusion
Purnami et al. A new expert system for diabetes disease diagnosis using modified spline smooth support vector machine
CN116798621A (en) System and method for predicting cardiovascular and cerebrovascular diseases
CN117796768A (en) Intelligent diabetes diagnosis system based on gait dynamic change and plantar pressure
CN113459158B (en) Exoskeleton assistance efficiency evaluation method and device and computer readable storage medium
CN117672523A (en) Sugar control level evaluation system and method for diabetics

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