A method and device for predicting hip joint torque

Abstract

The present application is applicable to the field of exoskeleton assistance technology, and provides a method and device for predicting hip joint torque, including: obtaining a torque prediction model, wherein the torque prediction model is obtained by training a pre-constructed initial torque prediction model based on a hip joint torque training data set, and the hip joint torque training data set includes multiple hip joint torque training data, and each hip joint torque training data includes: body parameter information, hip joint motion information and actual hip joint torque; through the torque prediction model, torque prediction is performed on the hip joint data to be predicted, and the hip joint torque corresponding to the hip joint data to be predicted is obtained. The present invention can solve the technical problems of low prediction accuracy and low efficiency of the existing hip joint torque estimation method.

Description

Technical Field

[0001] The present application belongs to the field of exoskeleton power-assist technology, and in particular, relates to a method and device for predicting hip joint torque. Background Art

[0002] Exoskeleton assistance refers to the use of external mechanical devices to provide force support to assist or enhance the natural movement of the human body. The lower limb exoskeleton assistance system can assist the human body's joint force, thereby reducing the human body's joint torque, metabolism, and axial pressure between joints, etc., and assist the human body in walking. It is widely used in groups such as weight-bearing walking, arthritis patients, and the elderly.

[0003] The design and control of the lower limb exoskeleton power-assist system requires accurate calculation and adjustment of the torque acting on each joint of the lower limbs of the human body to simulate or enhance the force required for natural human movement and achieve effective power assistance and movement assistance. The lower limb exoskeleton power-assist system includes hip exoskeleton power-assist, knee exoskeleton power-assist and ankle exoskeleton power-assist. The hip joint is the joint that does the most work during walking, so calculating and adjusting the torque of the hip joint when the human body walks is an important basis for the design and control of the lower limb exoskeleton power-assist system.

[0004] At present, the estimation methods of hip joint torque include: predicting hip joint torque based on the angle data of hip joint during movement, based on time domain convolutional neural network (TCN network) or ensemble neural network (Ensemble network). However, the above hip joint torque prediction methods have a long calculation cycle and low efficiency, are suitable for offline hip joint torque estimation, have timeliness problems, and the predicted hip joint torque has low accuracy. Summary of the invention

[0005] The embodiments of the present application provide a method and device for predicting hip joint torque, which can solve the technical problems of low prediction accuracy and low efficiency of existing hip joint torque estimation methods.

[0006] In a first aspect, an embodiment of the present application provides a method for predicting hip joint torque, comprising:

[0007] Acquire a torque prediction model, wherein the torque prediction model is obtained by training a pre-constructed initial torque prediction model based on a hip joint torque training data set, wherein the hip joint torque training data set includes a plurality of hip joint torque training data, and each of the hip joint torque training data includes: body parameter information, hip joint motion information, and actual hip joint torque;

[0008] The torque prediction model is used to perform torque prediction on the hip joint data to be predicted, and the hip joint torque corresponding to the hip joint data to be predicted is obtained.

[0009] In a possible implementation manner of the first aspect, before acquiring the torque prediction model, the method further includes:

[0010] Acquire a hip joint torque training data set, wherein the hip joint torque training data set includes a plurality of hip joint torque training data, and each of the hip joint torque training data includes: body parameter information, hip joint motion information and actual hip joint torque;

[0011] Extracting features of the hip joint torque according to the body parameter information and the actual hip joint torque in the hip joint torque training data set to obtain a generalized torque eigenvalue;

[0012] According to the hip joint motion information in the hip joint torque training data set, eigenvalue estimation is performed on the hip joint torque to obtain an estimated torque eigenvalue;

[0013] Reducing the hip joint torque according to the body parameter information and the generalized torque eigenvalue in the hip joint torque training data set to obtain an estimated hip joint torque;

[0014] According to the generalized torque eigenvalue, the estimated torque eigenvalue, the estimated hip joint torque and the actual hip joint torque, the pre-constructed initial torque prediction model is optimized to obtain a torque prediction model.

[0015] In a possible implementation of the first aspect, optimizing a pre-constructed initial torque prediction model according to the generalized torque eigenvalue, the estimated torque eigenvalue, the estimated hip joint torque, and the actual hip joint torque to obtain a torque prediction model includes:

[0016] Calculating a first loss value of the initial moment prediction model according to the generalized moment eigenvalue and the estimated moment eigenvalue;

[0017] Calculating a second loss value of the initial torque prediction model according to the estimated hip joint torque and the actual hip joint torque;

[0018] Performing a weighted summation on the first loss value and the second loss value of the torque prediction model to obtain a total loss value of the initial torque prediction model;

[0019] According to the total loss value of the initial torque prediction model, the initial torque prediction model is optimized to obtain a torque prediction model.

[0020] In a possible implementation of the first aspect, the hip joint torque training data further includes time information and user information, and performing eigenvalue estimation on the hip joint torque according to the hip joint motion information in the hip joint torque training data set to obtain an estimated torque eigenvalue includes:

[0021] Selecting target user's hip joint torque training data from the hip joint torque training data set according to the user information in the hip joint torque training data;

[0022] Selecting target historical hip joint torque training data from the target user's hip joint torque training data according to the time information in the hip joint torque training data;

[0023] According to the hip joint motion information in the hip joint torque training data and the hip joint motion information in the target historical hip joint torque training data, the eigenvalue of the hip joint torque is estimated to obtain the estimated torque eigenvalue corresponding to the hip joint torque training data.

[0024] In a possible implementation of the first aspect, the hip joint motion information includes: hip joint angle information, hip joint angular velocity information, and hip joint angular acceleration information;

[0025] The method of estimating the characteristic value of the hip joint torque according to the hip joint motion information in the hip joint torque training data and the hip joint motion information in the historical target hip joint torque training data to obtain the estimated torque characteristic value corresponding to the hip joint torque training data includes:

[0026] According to the hip joint angle information, hip joint angular velocity information and hip joint angular acceleration information in the hip joint torque training data, as well as the hip joint angle information, hip joint angular velocity information and hip joint angular acceleration information in the historical target hip joint torque training data, the eigenvalue of the hip joint torque is estimated to obtain the estimated torque eigenvalue corresponding to the hip joint torque training data.

[0027] In a possible implementation of the first aspect, the body parameter information includes: human height information and human weight information;

[0028] The step of extracting features of the hip joint torque according to the body parameter information and the actual hip joint torque in the hip joint torque training data set to obtain a generalized torque feature value includes:

[0029] According to the human body height information, human body weight information and the actual hip joint torque in the hip joint torque training data, feature extraction is performed on the hip joint torque to obtain a generalized torque feature value corresponding to the hip joint torque training data.

[0030] In a possible implementation of the first aspect, the torque prediction model includes a feature estimator and a torque decoder, and performing torque prediction on the hip joint data to be predicted through the torque prediction model to obtain the hip joint torque corresponding to the hip joint data to be predicted, including:

[0031] Inputting the hip joint motion information in the to-be-predicted hip joint data into the feature estimator, performing feature value estimation on the hip joint torque, and obtaining the estimated torque feature value corresponding to the to-be-predicted hip joint data;

[0032] The body parameter information in the hip joint data to be predicted and the estimated torque characteristic value corresponding to the hip joint data to be predicted are input into the torque decoder to restore the hip joint torque to obtain the hip joint torque corresponding to the hip joint data to be predicted.

[0033] In a second aspect, an embodiment of the present application provides a method for training a torque prediction model, comprising:

[0034] Acquire a hip joint torque training data set, wherein the hip joint torque training data set includes a plurality of hip joint torque training data, and each of the hip joint torque training data includes: body parameter information, hip joint motion information and actual hip joint torque;

[0035] Extracting features of the hip joint torque according to the body parameter information and the actual hip joint torque in the hip joint torque training data set to obtain a generalized torque eigenvalue;

[0036] According to the hip joint motion information in the hip joint torque training data set, eigenvalue estimation is performed on the hip joint torque to obtain an estimated torque eigenvalue;

[0037] Reducing the hip joint torque according to the body parameter information and the generalized torque eigenvalue in the hip joint torque training data set to obtain an estimated hip joint torque;

[0038] According to the generalized torque eigenvalue, the estimated torque eigenvalue, the estimated hip joint torque and the actual hip joint torque, the pre-constructed initial torque prediction model is optimized to obtain a torque prediction model.

[0039] In a third aspect, an embodiment of the present application provides a device for predicting hip joint torque, comprising:

[0040] A torque prediction model acquisition module is used to acquire a torque prediction model, wherein the torque prediction model is obtained by training a pre-constructed initial torque prediction model based on a hip joint torque training data set, wherein the hip joint torque training data set includes a plurality of hip joint torque training data, and each of the hip joint torque training data includes: body parameter information, hip joint motion information and actual hip joint torque;

[0041] The torque prediction module is used to perform torque prediction on the hip joint data to be predicted through the torque prediction model to obtain the hip joint torque corresponding to the hip joint data to be predicted.

[0042] In a fourth aspect, an embodiment of the present application provides a device for predicting hip joint torque, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements any of the methods described above when executing the computer program.

[0043] In a fifth aspect, an embodiment of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, it implements any of the methods described above.

[0044] In a sixth aspect, an embodiment of the present application provides a computer program product, which, when executed on a terminal device, enables the terminal device to execute any of the methods described above.

[0045] Compared with the prior art, the beneficial effects of the embodiments of the present application are as follows: by acquiring a torque prediction model, torque prediction is performed on the hip joint data to be predicted through the acquired torque prediction model, and the hip joint torque corresponding to the hip joint data to be predicted is obtained. Among them, the torque prediction model is obtained by training a pre-constructed initial torque prediction model based on a hip joint torque training data set, and the hip joint torque training data set includes a plurality of hip joint torque training data, each hip joint torque training data includes body parameter information, hip joint motion information and actual hip joint torque, and the torque prediction model integrates the body parameter information and hip joint motion information in the hip joint torque training data set, thereby reducing the impact of human body differences on the torque prediction model, and improving the accuracy of hip joint torque prediction, and, through the above-mentioned torque prediction model, torque prediction can be performed on the hip joint data to be predicted in real time to obtain the corresponding hip joint torque, which is highly efficient and effectively solves the technical problems of low prediction accuracy and low efficiency of the existing hip joint torque estimation method. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0047] Figure 1 It is a flowchart of a method for predicting hip joint torque provided by an embodiment of the present application;

[0048] Figure 2 is a structural schematic diagram of a torque prediction model provided by another embodiment of the present application;

[0049] Figure 3 is a structural diagram of a feature estimator provided by another embodiment of the present application;

[0050] Figure 4 is a curve diagram of a prediction result of a hip joint torque provided by another embodiment of the present application;

[0051] Figure 5 is a schematic diagram of another prediction result of hip joint torque provided by another embodiment of the present application;

[0052] Figure 6 is a flow chart of a training method for a torque prediction model provided by another embodiment of the present application;

[0053] Figure 7 It is a structural schematic diagram of a hip joint torque prediction device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0054] In the following description, specific details such as specific system structures, technologies, etc. are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present application. However, it should be clear to those skilled in the art that the present application may also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from obstructing the description of the present application.

[0055] It should be understood that when used in the present specification and the appended claims, the term "comprising" indicates the presence of described features, wholes, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or combinations thereof.

[0056] It should also be understood that the term “and / or” used in the specification and appended claims refers to any and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0057] As used in the specification and appended claims of this application, the term "if" can be interpreted as "when" or "uponce" or "in response to determining" or "in response to detecting", depending on the context. Similarly, the phrase "if it is determined" or "if [described condition or event] is detected" can be interpreted as meaning "uponce it is determined" or "in response to determining" or "uponce [described condition or event] is detected" or "in response to detecting [described condition or event]", depending on the context.

[0058] In addition, in the description of the present application specification and the appended claims, the terms "first", "second", "third", etc. are only used to distinguish the descriptions and cannot be understood as indicating or implying relative importance.

[0059] References to "one embodiment" or "some embodiments" etc. described in the specification of this application mean that one or more embodiments of the present application include specific features, structures or characteristics described in conjunction with the embodiment. Therefore, the statements "in one embodiment", "in some embodiments", "in some other embodiments", "in some other embodiments", etc. that appear in different places in this specification do not necessarily refer to the same embodiment, but mean "one or more but not all embodiments", unless otherwise specifically emphasized in other ways. The terms "including", "comprising", "having" and their variations all mean "including but not limited to", unless otherwise specifically emphasized in other ways.

[0060] Hip torque refers to the rotational torque acting on the hip joint, which is a key parameter in the lower limb exoskeleton assistance system. When designing and controlling the lower limb exoskeleton assistance system, it is impossible to calculate the hip torque with the help of motion capture systems, force measurement treadmills and other equipment. How to predict the hip torque of the person wearing the lower limb exoskeleton assistance system based on the human body's motion information measured by the lower limb exoskeleton assistance system's own sensors (such as the motor encoder or the IMU inertial measurement unit at the thigh) is an important issue that needs to be solved in the design and control of the exoskeleton assistance system.

[0061] How to predict the hip joint torque of a person wearing a lower limb exoskeleton assistance system based on the measured human body motion information. The traditional method is to obtain the hip joint torque by establishing an approximate dynamic model. However, this type of method relies on complex sensor equipment in the lower limb exoskeleton assistance system, such as plantar pressure sensors. Complex sensor equipment reduces the stability of the lower limb exoskeleton assistance system and reduces the user's wearing comfort.

[0062] In order to make the lower limb exoskeleton power-assisting system more stable and comfortable for users to wear, the current methods for estimating hip joint torque include: using the IMU data of the torso and thigh measured by the inertial measurement unit (IMU) sensor to predict the hip joint torque based on the time domain convolutional neural network (TCN network), or based on the ensemble neural network (Ensemble network) to predict the lower limb joint torque. However, the above joint torque prediction methods have a long calculation cycle and low efficiency. They are suitable for offline joint torque estimation. The hip joint torque predicted for different users (human bodies) has low accuracy and low prediction efficiency.

[0063] The present application embodiment provides a method for predicting hip joint torque, see Figure 1 , Figure 1 : is a flowchart of a method for predicting hip joint torque provided by an embodiment of the present application, comprising:

[0064] Step S11, obtaining a torque prediction model, wherein the torque prediction model is obtained by training a pre-constructed initial torque prediction model based on a hip joint torque training data set, the hip joint torque training data set includes a plurality of hip joint torque training data, and each hip joint torque training data includes: body parameter information, hip joint motion information and actual hip joint torque;

[0065] Step S12: performing torque prediction on the hip joint data to be predicted by using a torque prediction model to obtain the hip joint torque corresponding to the hip joint data to be predicted.

[0066] Specifically, joint torque refers to the measured value of the torque applied to the joint at different angles during human motion. Joint torque can usually be obtained through multi-body dynamics analysis, electromyographic signal analysis, or using specialized force measurement equipment.

[0067] Hip joint torque refers to the torque generated at the hip joint, which is determined by the force acting on the hip joint and the lever arm (the distance between the point of force application and the center of the joint). The force and lever arm acting on the hip joint are affected by the height and weight of the human body. During a human gait cycle, the change in hip joint torque is closely related to the movement state of the hip joint, the height and weight of the human body, etc.

[0068] The torque prediction model obtained in step S11 is obtained by training a pre-constructed initial torque prediction model based on the hip joint torque training data set. The hip joint torque training data set includes a plurality of hip joint torque training data, and each hip joint torque training data includes: body parameter information, hip joint motion information and actual hip joint torque. Body parameter information refers to information that can be directly obtained by measuring the human body, for example, it may include: height information, weight information, etc.; hip joint motion information refers to the hip joint angle information of the human body during movement, which can be collected by a three-dimensional motion capture system in a laboratory environment; the actual hip joint torque is the hip joint torque value calculated by the motion capture data and the ground force plate data when the human body is walking.

[0069] It should be noted that the specific method for calculating the hip joint torque value can be: in a laboratory environment, the motion capture data of the human body when walking is collected by a three-dimensional motion capture system, and the ground force plate data of the human body when walking is obtained by a force plate device placed on the ground. The motion capture data is the kinematic information such as the angle and angular velocity of the joint obtained by capturing the motion trajectory of the human body marker points, including the hip joint angle information of the human body when moving. The ground force plate data is the information of the ground reaction force (GRF), including vertical force, front and back force and left and right force. Through special software, the motion capture data and the ground force plate data are received, stored and analyzed. Both the ground force plate data and the motion capture data carry timestamp information. The ground force plate data are combined with the synchronously recorded motion capture data, and the hip joint torque value is calculated by the inverse dynamics equation. The present invention does not specifically limit the method for calculating the hip joint torque value.

[0070] In an optional example, a hip joint torque training data includes 6-dimensional hip joint motion information X of human body r at time t t =[X t，1 , X t，2 , X t，3 , X t，4 , X t，5 , X t，6 ], 2D body parameter information Q r =[Q r，1 , Q r，2 ] and actual hip joint torque M t，r , this hip joint torque training data is [X t，1 , X t，2 , X t，3 , X t，4 , X t，5 , X t，6 , Q r，1 , Q r，2 , M t，r ].

[0071] Specifically, after obtaining the hip joint torque training data set, the hip joint torque training data set may be filtered. The filtering method includes filtering out the hip joint torque training data containing missing values ​​and invalid values, filtering out the hip joint torque training data with abnormal actual hip joint torque, or filtering out the actual hip joint torque with mismatched timestamp information between the ground force plate data and the motion capture data.

[0072] The pre-built initial torque prediction model is a deep learning network model built in advance, and the network parameters included in it are initial values. The initial torque prediction model is used to predict the hip joint torque based on body parameter information and hip joint motion information.

[0073] It should be noted that the initial torque prediction model may include other network models such as a long short-term memory network model (Long Short-Term Memory network model, LSTM for short) or a GRU (Gated Recurrent Unit) network model, and the present invention does not make specific limitations on this.

[0074] When the initial torque prediction model includes the GRU network model, since the GRU network model is a recurrent neural network architecture and a simplified version of LSTM, a gating mechanism is introduced, which only includes two gates, an update gate and a reset gate. This avoids the problems of gradient vanishing and gradient exploding in traditional recurrent neural networks (RNN neural networks) during data processing, making the initial torque prediction model more efficient in processing hip joint torque training data.

[0075] Specifically, the hip joint torque training data set obtained above is input into a pre-constructed initial torque prediction model, and the pre-constructed initial torque prediction model is continuously trained so that the initial torque prediction model learns the body parameter information of different human bodies, the hip joint movement information at different times and the corresponding actual hip joint torque, and continuously updates the network parameters in the initial torque prediction model to obtain the torque prediction model.

[0076] For the hip joint data to be predicted, the hip joint data to be predicted is input into the above-mentioned torque prediction model. Since the hip joint data to be predicted includes body parameter information and hip joint motion information, torque prediction is performed based on the body parameter information and hip joint motion information to obtain the hip joint torque corresponding to the hip joint data to be predicted.

[0077] It can be understood that the technical solution provided in this embodiment performs torque prediction on the predicted hip joint data by acquiring a torque prediction model, and obtains the hip joint torque corresponding to the predicted hip joint data. Among them, the torque prediction model is obtained by training a pre-constructed initial torque prediction model based on a hip joint torque training data set, and the hip joint torque training data set includes a plurality of hip joint torque training data, each hip joint torque training data includes body parameter information, hip joint motion information and actual hip joint torque, and the torque prediction model integrates the body parameter information and hip joint motion information in the hip joint torque training data set, thereby reducing the impact of human body differences on the torque prediction model, improving the accuracy of hip joint torque prediction, and, through the above-mentioned torque prediction model, torque prediction can be performed on the predicted hip joint data in real time to obtain the corresponding hip joint torque, thereby improving the prediction efficiency of torque prediction for the predicted hip joint data, and effectively solving the technical problems of low prediction accuracy and low efficiency of the existing hip joint torque estimation method.

[0078] In a possible implementation, in step S11, before obtaining the torque prediction model, the following steps are further included:

[0079] Step S111, obtaining a hip joint torque training data set, wherein the hip joint torque training data set includes a plurality of hip joint torque training data, and each hip joint torque training data includes: body parameter information, hip joint motion information and actual hip joint torque;

[0080] Step S112, extracting features of the hip joint torque according to the body parameter information and the actual hip joint torque in the hip joint torque training data set to obtain a generalized torque feature value;

[0081] Step S113, estimating the eigenvalue of the hip joint torque according to the hip joint motion information in the hip joint torque training data set to obtain an estimated torque eigenvalue;

[0082] Step S114, restoring the hip joint torque according to the body parameter information and the generalized torque eigenvalue in the hip joint torque training data set to obtain an estimated hip joint torque;

[0083] Step S115, optimizing the pre-constructed initial torque prediction model according to the generalized torque eigenvalue, the estimated torque eigenvalue, the estimated hip joint torque and the actual hip joint torque to obtain a torque prediction model.

[0084] Specifically, before obtaining the torque prediction model, the above hip joint torque prediction method further includes:

[0085] A hip joint torque training data set is obtained, wherein the hip joint torque training data set includes multiple hip joint torque training data, each hip joint torque training data includes: body parameter information, hip joint motion information and actual hip joint torque. The hip joint torque training data set includes filtered hip joint torque training data.

[0086] refer to Figure 2 , Figure 2 is a structural diagram of a torque prediction model provided by another embodiment of the present application. The pre-built initial torque prediction model includes an initial feature generator ( Figure 2 generatorG in ), initial feature estimator ( Figure 2 estimator E) and initial moment decoder ( Figure 2 decoder R in .

[0087] The hip joint torque is related to the body parameters of the human body. The initial feature generator can be one or more adaptive neural networks (such as convolutional layers, fully connected layers, etc.) to transform the body parameter information of the human body ( Figure 2 Q in) and hip joint torque ( Figure 2 The M in the figure is fused to extract the relationship characteristics between the body parameter information and the hip joint torque in the hip joint torque training dataset, and the generalized torque eigenvalue (GTF for short) corresponding to the hip joint torque is obtained.

[0088] In an optional example, the initial feature generator includes a fully connected network with 5 hidden layers, the number of nodes in each hidden layer is set to 32, and a neural network activation function (LeakyReLU) is connected after the fully connected network to achieve nonlinearity, fully utilize the output of each node, and improve the training speed and performance of the feature generator.

[0089] Specifically, the initial feature estimator is used to receive the generalized moment eigenvalue output by the initial feature generator, and according to the generalized moment eigenvalue and the previous internal state (hip joint motion information, Figure 2 The initial feature estimator can be composed of a recurrent neural network (such as GRU).

[0090] The initial torque decoder is used to receive the output of the feature generator and decode the generalized torque eigenvalue output by the feature generator and the actual hip joint torque ( Figure 2 The initial moment decoder can be composed of one or more neural network layers (such as fully connected layers, Figure 2 The output) in .

[0091] The hip joint torque training data set is input into the initial feature generator of the pre-built initial torque prediction model. According to the body parameter information and the actual hip joint torque in the hip joint torque training data set, the hip joint torque is feature extracted to obtain the generalized torque eigenvalue, and the generalized torque eigenvalue is output to the initial feature generator.

[0092] The hip joint torque training data set is input into the initial feature estimator, and the eigenvalue of the hip joint torque is estimated according to the hip joint motion information in the hip joint torque training data set to obtain the estimated torque eigenvalue, and the estimated torque eigenvalue is output to the initial feature estimator.

[0093] The hip joint torque training data set and the generalized torque eigenvalues ​​output by the above-mentioned initial feature generator are input into the initial torque decoder. The hip joint torque is restored according to the body parameter information and the generalized torque eigenvalues ​​in the hip joint torque training data set to obtain the estimated hip joint torque, and the estimated hip joint torque is output to the initial torque decoder.

[0094] According to the generalized torque eigenvalues ​​output by the above-mentioned initial feature generator, the estimated torque eigenvalues ​​output by the above-mentioned initial feature estimator, the estimated hip joint torque output by the above-mentioned initial torque decoder and the actual hip joint torque in the hip joint torque training data set, the initial feature generator, initial feature estimator and initial torque decoder in the pre-constructed initial torque prediction model are optimized to obtain a torque prediction model, which includes a feature generator, a feature estimator and a torque decoder.

[0095] Continuing with the above example, a hip joint torque training data is [X t，1 , X t，2 , X t，3 , X t，4 , X t，5 , X t，6 , Q r，1 , Q r，2 , M t，r ], the 2D body parameter information [Q r，1 , Q r，2 ] and the actual hip joint torque value M t，r Input the pre-built initial feature generator to extract the features of the hip joint torque and obtain the generalized torque eigenvalue M′ corresponding to the hip joint torque training data. t，r .

[0096] The 6-dimensional hip joint motion information in the above hip joint torque training data [X t，1 , X t，2 , X t，3 , X t，4 , X t，5 , X t，6] Input the pre-built initial feature estimator, estimate the eigenvalue of the hip joint torque, and obtain the estimated torque eigenvalue M″ corresponding to the hip joint torque training data t，r .

[0097] The generalized moment eigenvalue M′ t，r With 2D body parameter information [Q r，1 , Q r，2 ] Input the pre-built initial torque decoder, decode and restore the hip joint torque, and obtain the estimated hip joint torque corresponding to the hip joint torque training data

[0098] According to the generalized moment eigenvalue M′ output by the initial feature generator above t，r , the estimated moment eigenvalue M″ output by the initial eigenvalue estimator t，r , the estimated hip joint torque output by the initial torque decoder and the actual hip joint torque M in the hip joint torque training dataset t，r , the initial feature generator, the initial feature estimator and the initial moment decoder in the pre-built initial moment prediction model are optimized to obtain the moment prediction model, which includes the feature generator, the feature estimator and the moment decoder.

[0099] It should be noted that the hip joint torque training data set can be obtained multiple times, each time the hip joint torque training data set contains multiple hip joint torque training data, and the operation in the above step S11 is repeated for multiple rounds, and the pre-constructed initial torque prediction model is optimized multiple times to obtain a torque prediction model.

[0100] It can be understood that when optimizing the initial moment prediction model, the generalized moment eigenvalue output by the initial feature generator will be used as the output target value of the initial feature estimator ( Figure 2 label), the estimated torque eigenvalue output by the initial feature estimator is used as the output target value of the initial feature generator, the initial feature generator and the initial feature estimator cooperate and optimize together, the generalized torque eigenvalue output by the initial feature generator will be used as the input of the initial torque decoder, the initial feature generator and the initial torque decoder are trained and optimized together, the optimized feature generator aims to improve the prediction accuracy of the feature estimator, generates the optimal GTF, and improves the generalization performance of the torque prediction model in predicting the hip joint torque of different users (human bodies); the optimized feature generator is also used to ensure that the torque decoder can restore the GTF to the real joint torque with high precision, so that the trained torque decoder can decode and restore the generalized torque eigenvalue to the real torque.

[0101] In a possible implementation, in step S115, the pre-constructed initial torque prediction model is optimized according to the generalized torque eigenvalue, the estimated torque eigenvalue, the estimated hip joint torque and the actual hip joint torque to obtain a torque prediction model, including:

[0102] According to the generalized moment eigenvalue and the estimated moment eigenvalue, a first loss value of the initial moment prediction model is calculated;

[0103] According to the estimated hip joint torque and the actual hip joint torque, a second loss value of the initial torque prediction model is calculated;

[0104] Performing a weighted summation on the first loss value and the second loss value of the moment prediction model to obtain a total loss value of the initial moment prediction model;

[0105] According to the total loss value of the initial torque prediction model, the initial torque prediction model is optimized to obtain the torque prediction model.

[0106] Specifically, according to the generalized torque eigenvalue, the estimated torque eigenvalue, the estimated hip joint torque and the actual hip joint torque, the pre-built initial torque prediction model is optimized to obtain the torque prediction model, including:

[0107] refer to Figure 2 For each hip joint torque training data, according to the above generalized torque eigenvalue and estimated torque eigenvalue, the first loss function formula is as follows: Figure 2 L1 in the figure), the first loss value L1 of the initial torque prediction model is calculated:

[0108] L1=||(E(X;θ E )-G(Q,M;θ G ))||2;

[0109] Among them, X represents the hip joint motion information, θ E represents the initial feature generator or the network parameters of the feature generator, E(X;θ E ) represents the estimated torque eigenvalue, Q represents the body parameter information, M represents the actual hip joint torque, θ G represents the initial feature estimator or the network parameters of the feature estimator, G(Q, M; θ G ) represents the generalized moment eigenvalue.

[0110] According to the estimated hip joint torque and the actual hip joint torque, the second loss function formula is as follows: Figure 2 L2 in the calculation), the second loss value L2 of the initial moment prediction model is calculated:

[0111] L2=||(R(M′,Q;θ R )-M)||2;

[0112] Where M′=G(Q,M;θ G ) represents the generalized moment eigenvalue, Q represents the body parameter information, θ R represents the initial moment decoder or the network parameters of the moment decoder, R(M′, Q; θ R ) represents the estimated hip joint torque, and M represents the actual hip joint torque.

[0113] The first loss value L1 and the second loss value L2 of the torque prediction model calculated above are weighted and summed according to the following total loss function formula to obtain the total loss value L of the initial torque prediction model:

[0114] L = w1*L1+w2*L2;

[0115] Among them, w1 represents the weight of the preset first loss function, w2 represents the weight of the preset second loss function, L1 represents the first loss value, and L2 represents the second loss function; w1+w2=1.

[0116] For multiple hip joint torque training data, repeat the operation in the above step S115, and optimize the initial torque prediction model according to the total loss value of the initial torque prediction model. When the total loss value tends to converge, the network parameters in the feature generator, feature estimator, and torque decoder obtain the optimal parameters. At this time, a torque prediction model including the feature generator, feature estimator, and torque decoder is obtained.

[0117] Among them, the network parameters θ in the feature generator obtained after optimization G and the network parameters θ in the feature estimator E , is the optimal parameter and

[0118]

[0119] The network parameters θ in the torque decoder obtained after optimization R , is the optimal parameter

[0120]

[0121] At this time, the feature generator, feature estimator, and moment decoder all reach the optimal state.

[0122] It should be noted that when multiple hip joint torque training data in the hip joint torque training data set are selected and synchronously input into the initial torque prediction model, the first loss value L1 is calculated according to the following formula:

[0123]

[0124] Among them, E i (X;θ E ) represents the estimated torque eigenvalue corresponding to the i-th hip joint torque training data, G i (Q, M; θ G ) represents the generalized torque eigenvalue corresponding to the i-th hip joint torque training data, N represents the number of hip joint torque training data, i = 1, 2, …, N.

[0125] The second loss value L2 is calculated according to the following formula:

[0126]

[0127] Among them, R i (G(Q, M; θ G ), Q; θ R ) represents the estimated hip joint torque corresponding to the i-th hip joint torque training data, M i represents the actual hip joint torque in the i-th hip joint torque training data, N represents the number of hip joint torque training data, i = 1, 2, …, N.

[0128] The first loss value L1 and the second loss value L2 of the torque prediction model calculated above are weighted and summed to obtain the total loss value L of the torque prediction model. According to the total loss value, the pre-built initial feature generator, initial feature estimator and initial moment decoder are optimized at the same time. After repeated multiple rounds of training, the total loss value L tends to converge, and the network parameters in the feature generator, feature estimator and moment decoder reach the optimal value. At this time, the torque prediction model including the feature generator, feature estimator and moment decoder is obtained.

[0129] In a possible implementation, in step S113, the hip joint torque training data further includes time information and user information, and the hip joint torque is estimated based on the hip joint motion information in the hip joint torque training data set to obtain an estimated torque characteristic value, including:

[0130] According to the user information in the hip joint torque training data, the target user's hip joint torque training data is selected from the hip joint torque training data set;

[0131] Selecting target historical hip joint torque training data from the target user's hip joint torque training data according to time information in the hip joint torque training data;

[0132] According to the hip joint motion information in the hip joint torque training data and the hip joint motion information in the target historical hip joint torque training data, the eigenvalue of the hip joint torque is estimated to obtain the estimated torque eigenvalue corresponding to the hip joint torque training data.

[0133] Specifically, according to the hip joint motion information in the hip joint torque training data set, the eigenvalue of the hip joint torque is estimated to obtain the estimated torque eigenvalue, including:

[0134] The hip joint torque training data also includes time information and user information. The time information indicates the collection time of the hip joint motion information in the hip joint torque training data, and the user information indicates the information of the user to whom the hip joint torque training data belongs. One user corresponds to one body parameter information, one user may correspond to multiple hip joint motion information, and one hip joint motion information corresponds to one time information.

[0135] Since the hip joint torque training data includes user information, the target user's hip joint torque training data is selected from the hip joint torque training data set according to the user information in the hip joint torque training data. The target user's hip joint torque training data is the target user's hip joint torque training data, which may include the target user's hip joint torque training data collected at different times.

[0136] The hip joint torque training data also includes time information, which is the collection time of the user's hip joint torque training data. According to the time information in the hip joint torque training data, the above selected target user hip joint torque training data are sorted to obtain the target user hip joint torque training data arranged in chronological order. For a target user hip joint torque training data at the current moment, the target user hip joint torque training data at multiple adjacent moments before it are selected as the target historical hip joint torque training data.

[0137] Continuing with the above example, the target user’s hip joint motion information at time t is X t , this hip joint torque training data is [X t，1 , X t，2 , X t，3 , X t，4 , X t，5 , X t，6 , Q r，1 , Q r，2 , M t，r ], obtain the hip joint torque training data of the target user at ΔT moments before moment t as the target historical hip joint torque training data, where ΔT is the interception window size, which can be set according to the specific situation.

[0138] Get the target user's hip joint motion information X at time t-1 t-1 , the target user’s hip joint motion information X at time tj t-j , the target user’s hip joint motion information X at time t-ΔT t-ΔT. Where j is 1, 2, …, ΔT. t-1 …X t-j …X t-ΔT ] as the target historical hip joint torque training data.

[0139] The target user's hip joint motion information X at time t t and target historical hip joint torque training data [X t-1 …X t-j …X t-ΔT ], recorded as: A T .

[0140] A T =(X t , X t-1 , …X t-j， …, X t-ΔT )∈R ΔT×D ;

[0141] Wherein, D is the dimension of hip joint motion information, which can be set according to specific circumstances.

[0142] According to the hip joint motion information in the hip joint torque training data and the hip joint motion information in the target historical hip joint torque training data, the eigenvalue of the hip joint torque is estimated to obtain the estimated torque eigenvalue corresponding to the hip joint torque training data.

[0143] It should be noted that when estimating the eigenvalue of the hip joint torque, it is done through an initial feature estimator, which is a recurrent neural network structure. Figure 3 , Figure 3 is a structural diagram of a feature estimator provided by another embodiment of the present application. The initial feature estimator consists of a GRU block. The number of hidden neurons in the GRU block is set to 16, including h 0 ,h 1 …h 15 The 16*1 fully connected network output is used to estimate the moment eigenvalue. The GRU block obtains two different modes at each moment, including the hidden input information H of the previous moment. t-1 , and the hip joint motion information X at the current time t t One of the hidden neurons and the hip joint motion information X at the current time t t After calculation, the information flow is adjusted by updating the gate and resetting the gate, and the output The same principle applies to all hidden neurons. to Respectively with the hip joint motion information X at the current time t t After calculation, the output is to As Ht, after passing through the fully connected network, the estimated moment eigenvalue is obtained.

[0144] It can be understood that the recurrent neural network structure is based on memory information transmission. The feature estimator adopts the recurrent neural network structure, which can inherit the hidden information of the previous moment without processing within the network, which can effectively reduce the calculation time and improve the calculation efficiency.

[0145] In a possible implementation, the hip joint motion information includes: hip joint angle information, hip joint angular velocity information, and hip joint angular acceleration information;

[0146] According to the hip joint motion information in the hip joint torque training data and the hip joint motion information in the historical target hip joint torque training data, the characteristic value of the hip joint torque is estimated to obtain the estimated torque characteristic value corresponding to the hip joint torque training data, including:

[0147] According to the hip joint angle information, hip joint angular velocity information and hip joint angular acceleration information in the hip joint torque training data, as well as the hip joint angle information, hip joint angular velocity information and hip joint angular acceleration information in the historical target hip joint torque training data, the eigenvalue of the hip joint torque is estimated to obtain the estimated torque eigenvalue corresponding to the hip joint torque training data.

[0148] Specifically, according to the hip joint motion information in the hip joint torque training data and the hip joint motion information in the historical target hip joint torque training data, the eigenvalue of the hip joint torque is estimated to obtain the estimated torque eigenvalue corresponding to the hip joint torque training data, including:

[0149] The hip joint motion information includes hip joint angle information, hip joint angular velocity information and hip joint angular acceleration information. The hip joint angle information includes the left leg hip joint angle and the right leg hip joint angle, the hip joint angular velocity information includes the left leg hip joint angular velocity and the right leg hip joint angular velocity, and the hip joint angular acceleration information includes the left leg hip joint angular acceleration and the right leg hip joint angular acceleration.

[0150] The left leg hip joint angle, right leg hip joint angle, left leg hip joint angular velocity, right leg hip joint angular velocity, left leg hip joint angular acceleration and right leg hip joint angular acceleration in the hip joint torque training data, as well as the left leg hip joint angle, right leg hip joint angle, left leg hip joint angular velocity, right leg hip joint angular velocity, left leg hip joint angular acceleration and right leg hip joint angular acceleration in the historical target hip joint torque training data are used to estimate the eigenvalue of the hip joint torque through an initial feature estimator to obtain the estimated torque eigenvalue corresponding to the hip joint torque training data.

[0151] Continuing with the above example, the hip joint torque training data [Xt，1 , X t，2 , X t，3 , X t，4 , X t，5 , X t，6 , Q r，1 , Q r，2 , M t，r ] contains 6-dimensional hip joint motion information: X t，1 , X t，2 , X t，3 , X t，4 , X t，5 , X t，6 They are the left leg hip joint angle, right leg hip joint angle, left leg hip joint angular velocity, right leg hip joint angular velocity, left leg hip joint angular acceleration and right leg hip joint angular acceleration of the target user at time t.

[0152] In a possible implementation, in step S112, the body parameter information includes: human height information and human weight information;

[0153] According to the body parameter information and actual hip joint torque in the hip joint torque training data set, the hip joint torque is feature extracted to obtain the generalized torque eigenvalues, including:

[0154] According to the human height information, human weight information and actual hip joint torque in the hip joint torque training data, the hip joint torque is feature extracted to obtain the generalized torque eigenvalue corresponding to the hip joint torque training data.

[0155] Specifically, according to the body parameter information and actual hip joint torque in the hip joint torque training data set, the hip joint torque is feature extracted to obtain the generalized torque eigenvalue, including:

[0156] The body parameter information includes body height information and body weight information, wherein the body height information is the height of the user, and the body weight information is the weight of the user.

[0157] According to the human height information, human weight information and actual hip joint torque in the hip joint torque training data, the hip joint torque is feature extracted through the initial feature generator according to the following formula to obtain the generalized torque eigenvalue M′ corresponding to the hip joint torque training data:

[0158] M′=G(Q,M;θ G );

[0159] Among them, Q represents body parameter information, M represents actual hip joint torque, θ G is the initial feature generator or the network parameters of the feature generator.

[0160] Continuing from the above example, in the hip joint torque training data, Q r，1 and Q r，2 are the height and weight of user r, Q r，1 and Q r，2 Composition body parameter information Q.

[0161] In a possible implementation, in step S12, the torque prediction model includes a feature estimator and a torque decoder. The torque prediction model is used to perform torque prediction on the hip joint data to be predicted, and the hip joint torque corresponding to the hip joint data to be predicted is obtained, including:

[0162] Inputting hip joint motion information in the to-be-predicted hip joint data into a feature estimator, performing feature value estimation on the hip joint torque, and obtaining an estimated torque feature value corresponding to the to-be-predicted hip joint data;

[0163] The body parameter information in the hip joint data to be predicted and the estimated torque characteristic value corresponding to the hip joint data to be predicted are input into the torque decoder to restore the hip joint torque to obtain the hip joint torque corresponding to the hip joint data to be predicted.

[0164] Specifically, refer to Figure 2 The optimized torque prediction model includes a feature generator, a feature estimator and a torque decoder. The feature generator is used to train the torque prediction model and does not participate in the torque prediction of the hip joint data.

[0165] Through the torque prediction model, the torque prediction of the hip joint data to be predicted is performed ( Figure 2 Momentpredicting in , obtain the hip joint torque corresponding to the hip joint data to be predicted, including:

[0166] The hip joint data to be predicted ( Figure 2 Hip joint motion information in Walking Figure 2 Hipkinematics) input feature estimator ( Figure 2 The fast estimator in the figure) contains the hip joint motion information of a user at a certain moment. The feature estimator is used to quickly estimate the eigenvalue of the hip joint torque, and the estimated torque eigenvalue corresponding to the hip joint data to be predicted is output ( Figure 2 *GTF in ).

[0167] The hip joint data to be predicted also includes a user's body parameter information ( Figure 2 The body parameter information in the to-be-predicted hip joint data and the estimated torque feature value corresponding to the to-be-predicted hip joint data are input into the torque decoder ( Figure 2The decoder in the decoder is used to restore the hip joint torque to obtain the hip joint torque corresponding to the hip joint data to be predicted.

[0168] In an optional example, the hip joint data to be predicted is [X a , X b , X c , X d , X e , X f , Q u , Q v ], where X a , X b , X c , X d , X e , X f The 6-dimensional hip joint motion information of the user is input into the optimized feature estimator to obtain the estimated torque feature value, and the estimated torque feature value is compared with the user's 2-dimensional body parameter information Q u , Q v The optimized torque decoder is input to restore the hip joint torque, and the hip joint torque corresponding to the hip joint data to be predicted is predicted.

[0169] It should be noted that the reference Figure 4 , Figure 4 It is a curve diagram of a prediction result of a hip joint torque provided by another embodiment of the present application. Hip joint torque data of 10 users are collected respectively, and different torque prediction models are trained using different information in the hip joint torque data. Then, the hip joint torque data of the 10 users are used as test sets, and the trained torque prediction model is used to perform hip joint torque test. The torque prediction model composed of a feature generator, a feature estimator, and a torque decoder used in this embodiment is trained and tested with hip joint torque data containing body parameter information. The predicted hip joint torque is superior to torque prediction models trained by other methods (for example, Figure 4 The torque prediction model Modelwithout bodyparameters obtained by training without using human body information parameters and the torque prediction model Fusionmodel without GMF that does not generate generalized moment eigenvalues).

[0170] refer to Figure 5 , Figure 5This is a schematic diagram of another prediction result of hip joint torque provided by another embodiment of the present application. It can be seen that different torque prediction models are trained using hip joint torque training data sets of 10 users, and then the hip joint torque data of 10 users are used as test sets, and the trained torque prediction models are used to perform hip joint torque tests. The torque prediction model composed of the recurrent neural network structure adopted in this embodiment, after training and testing, is superior to other network structures (such as CNN) in efficiency in estimating the generalized torque features of the hip joint motion information in the predicted hip joint data. Figure 5 The feature estimator in the torque prediction model composed of the fully connected network structure FC-based, the convolutional neural network structure CNN-based and the time domain convolutional network structure TCN-based. Moreover, for the hip joint torque obtained by predicting the hip joint data to be predicted, the torque prediction model in this embodiment is also more efficient than other network structures (such as Figure 5 The torque prediction model consists of a fully connected network structure FC-based, a convolutional neural network structure CNN-based and a time domain convolutional network structure TCN-based.

[0171] It can be understood that the technical solution provided by the present embodiment obtains a torque prediction model, and performs torque prediction on the hip joint data to be predicted through the obtained torque prediction model, so as to obtain the hip joint torque corresponding to the hip joint data to be predicted. Among them, the torque prediction model is obtained by training a pre-constructed initial torque prediction model based on a hip joint torque training data set, and the hip joint torque training data set includes a plurality of hip joint torque training data, each hip joint torque training data includes body parameter information, hip joint motion information and actual hip joint torque, and the torque prediction model integrates the body parameter information and hip joint motion information in the hip joint torque training data set, thereby reducing the influence of human body differences on the torque prediction model, and improving the accuracy of hip joint torque prediction, and the network structure in the above-mentioned torque prediction model adopts a recurrent neural network, with fewer parameters, and can perform torque prediction on the hip joint data to be predicted in real time to obtain the corresponding hip joint torque, with high efficiency, and effectively solves the technical problems of low prediction accuracy and low efficiency of the existing hip joint torque estimation method.

[0172] Another embodiment of the present application provides a method for training a torque prediction model, see Figure 6 , Figure 6 FIG. 1 is a flow chart of a training method for a torque prediction model provided by another embodiment of the present application, comprising:

[0173] Step S21, obtaining a hip joint torque training data set, wherein the hip joint torque training data set includes a plurality of hip joint torque training data, and each hip joint torque training data includes: body parameter information, hip joint motion information and actual hip joint torque;

[0174] Step S22, extracting features of the hip joint torque according to the body parameter information and the actual hip joint torque in the hip joint torque training data set to obtain a generalized torque eigenvalue;

[0175] Step S23, estimating the eigenvalue of the hip joint torque according to the hip joint motion information in the hip joint torque training data set to obtain an estimated torque eigenvalue;

[0176] Step S24, restoring the hip joint torque according to the body parameter information and the generalized torque eigenvalue in the hip joint torque training data set to obtain an estimated hip joint torque;

[0177] Step S25, optimizing the pre-constructed initial torque prediction model according to the generalized torque eigenvalue, the estimated torque eigenvalue, the estimated hip joint torque and the actual hip joint torque to obtain a torque prediction model.

[0178] It should be noted that the training method of the torque prediction model, its specific method, function and technical effects can be found in the embodiment of the prediction method of hip joint torque, and will not be repeated here.

[0179] It should be understood that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0180] Corresponding to the method of the above embodiment, Figure 7 A schematic structural diagram of a hip joint torque device provided in an embodiment of the present application is shown. For ease of explanation, only the parts related to the embodiment of the present application are shown.

[0181] Reference Figure 7 , the device comprises:

[0182] A torque prediction model acquisition module 31 is used to acquire a torque prediction model, wherein the torque prediction model is obtained by training a pre-constructed initial torque prediction model based on a hip joint torque training data set, wherein the hip joint torque training data set includes a plurality of hip joint torque training data, and each hip joint torque training data includes: body parameter information, hip joint motion information and actual hip joint torque;

[0183] The torque prediction module 32 is used to perform torque prediction on the hip joint data to be predicted through a torque prediction model to obtain the hip joint torque corresponding to the hip joint data to be predicted.

[0184] It should be noted that the information interaction, execution process, etc. between the above-mentioned devices / units are based on the same concept as the embodiment of the method of the present application. Their specific functions and technical effects can be found in the embodiment of the method for predicting hip joint torque, and will not be repeated here.

[0185] The technicians in the relevant field can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In practical applications, the above-mentioned function allocation can be completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated in a processing unit, or each unit can exist physically separately, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, which will not be repeated here.

[0186] An embodiment of the present application also provides a device for predicting hip joint torque, which includes: at least one processor, a memory, and a computer program stored in the memory and executable on at least one processor, and when the processor executes the computer program, the steps in any of the above-mentioned method embodiments are implemented.

[0187] The embodiment of the present application further provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments can be implemented.

[0188] An embodiment of the present application provides a computer program product. When the computer program product runs on a mobile terminal, the mobile terminal can implement the steps in the above-mentioned method embodiments when executing the computer program product.

[0189] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present application implements all or part of the processes in the above-mentioned embodiment method, which can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may at least include: any entity or device that can carry the computer program code to the camera / terminal device, a recording medium, a computer memory, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), an electric carrier signal, a telecommunication signal, and a software distribution medium. For example, a USB flash drive, a mobile hard disk, a magnetic disk or an optical disk. In some jurisdictions, according to legislation and patent practice, computer-readable media cannot be electric carrier signals and telecommunication signals.

[0190] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0191] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0192] In the embodiments provided in the present application, it should be understood that the disclosed devices / network equipment and methods can be implemented in other ways. For example, the device / network equipment embodiments described above are only schematic. For example, the division of modules or units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0193] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0194] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. A method for predicting hip joint torque, characterized in that: include: Obtaining a torque prediction model, wherein the torque prediction model is obtained by training a pre-constructed initial torque prediction model based on a hip joint torque training data set, the hip joint torque training data set includes a plurality of hip joint torque training data, each of the hip joint torque training data includes: body parameter information, hip joint motion information and actual hip joint torque, the body parameter information includes: human height information and human weight information, the hip joint motion information includes: hip joint angle information, hip joint angular velocity information and hip joint angular acceleration information; Using the torque prediction model, torque prediction is performed on the hip joint data to be predicted to obtain the hip joint torque corresponding to the hip joint data to be predicted; Before acquiring the torque prediction model, the method further includes: Obtain hip joint torque training data set; Extracting features of the hip joint torque according to the body parameter information and the actual hip joint torque in the hip joint torque training data set to obtain a generalized torque eigenvalue; According to the hip joint motion information in the hip joint torque training data set, eigenvalue estimation is performed on the hip joint torque to obtain an estimated torque eigenvalue; Reducing the hip joint torque according to the body parameter information and the generalized torque eigenvalue in the hip joint torque training data set to obtain an estimated hip joint torque; Optimizing a pre-constructed initial torque prediction model according to the generalized torque eigenvalue, the estimated torque eigenvalue, the estimated hip joint torque, and the actual hip joint torque to obtain a torque prediction model; The step of optimizing the pre-built initial torque prediction model according to the generalized torque eigenvalue, the estimated torque eigenvalue, the estimated hip joint torque and the actual hip joint torque to obtain the torque prediction model comprises: Calculating a first loss value of the initial moment prediction model according to the generalized moment eigenvalue and the estimated moment eigenvalue; Calculating a second loss value of the initial torque prediction model according to the estimated hip joint torque and the actual hip joint torque; Performing a weighted summation on the first loss value and the second loss value of the torque prediction model to obtain a total loss value of the initial torque prediction model; According to the total loss value of the initial torque prediction model, the initial torque prediction model is optimized to obtain a torque prediction model.

2. The method for predicting hip joint torque according to claim 1, characterized in that: The hip joint torque training data also includes time information and user information. The eigenvalue estimation of the hip joint torque is performed according to the hip joint motion information in the hip joint torque training data set to obtain the estimated torque eigenvalue, including: Selecting target user's hip joint torque training data from the hip joint torque training data set according to the user information in the hip joint torque training data; Selecting target historical hip joint torque training data from the target user's hip joint torque training data according to the time information in the hip joint torque training data; According to the hip joint motion information in the hip joint torque training data and the hip joint motion information in the target historical hip joint torque training data, the eigenvalue of the hip joint torque is estimated to obtain the estimated torque eigenvalue corresponding to the hip joint torque training data.

3. The method for predicting hip joint torque as claimed in claim 2, characterized in that: The method of estimating the characteristic value of the hip joint torque according to the hip joint motion information in the hip joint torque training data and the hip joint motion information in the historical target hip joint torque training data to obtain the estimated torque characteristic value corresponding to the hip joint torque training data includes: According to the hip joint angle information, hip joint angular velocity information and hip joint angular acceleration information in the hip joint torque training data, as well as the hip joint angle information, hip joint angular velocity information and hip joint angular acceleration information in the historical target hip joint torque training data, the eigenvalue of the hip joint torque is estimated to obtain the estimated torque eigenvalue corresponding to the hip joint torque training data.

4. The method for predicting hip joint torque according to claim 1, characterized in that: The step of extracting features of the hip joint torque according to the body parameter information and the actual hip joint torque in the hip joint torque training data set to obtain a generalized torque feature value includes: According to the human body height information, human body weight information and the actual hip joint torque in the hip joint torque training data, feature extraction is performed on the hip joint torque to obtain a generalized torque feature value corresponding to the hip joint torque training data.

5. The method for predicting hip joint torque according to claim 1, characterized in that: The torque prediction model includes: a feature estimator and a torque decoder; The step of performing torque prediction on the hip joint data to be predicted by using the torque prediction model to obtain the hip joint torque corresponding to the hip joint data to be predicted includes: Inputting the hip joint motion information in the to-be-predicted hip joint data into the feature estimator, performing feature value estimation on the hip joint torque, and obtaining the estimated torque feature value corresponding to the to-be-predicted hip joint data; The body parameter information in the hip joint data to be predicted and the estimated torque characteristic value corresponding to the hip joint data to be predicted are input into the torque decoder to restore the hip joint torque to obtain the hip joint torque corresponding to the hip joint data to be predicted.

6. A method for training a torque prediction model, characterized in that: include: Acquire a hip joint torque training data set, wherein the hip joint torque training data set includes a plurality of hip joint torque training data, each of the hip joint torque training data includes: body parameter information, hip joint motion information and actual hip joint torque, the body parameter information includes: human height information and human weight information, the hip joint motion information includes: hip joint angle information, hip joint angular velocity information and hip joint angular acceleration information; Extracting features of the hip joint torque according to the body parameter information and the actual hip joint torque in the hip joint torque training data set to obtain a generalized torque eigenvalue; According to the hip joint motion information in the hip joint torque training data set, eigenvalue estimation is performed on the hip joint torque to obtain an estimated torque eigenvalue; Reducing the hip joint torque according to the body parameter information and the generalized torque eigenvalue in the hip joint torque training data set to obtain an estimated hip joint torque; According to the generalized torque eigenvalue, the estimated torque eigenvalue, the estimated hip joint torque and the actual hip joint torque, a pre-constructed initial torque prediction model is optimized to obtain a torque prediction model; The step of optimizing the pre-built initial torque prediction model according to the generalized torque eigenvalue, the estimated torque eigenvalue, the estimated hip joint torque and the actual hip joint torque to obtain the torque prediction model comprises: Calculating a first loss value of the initial moment prediction model according to the generalized moment eigenvalue and the estimated moment eigenvalue; Calculating a second loss value of the initial torque prediction model according to the estimated hip joint torque and the actual hip joint torque; Performing a weighted summation on the first loss value and the second loss value of the torque prediction model to obtain a total loss value of the initial torque prediction model; Optimizing the initial torque prediction model according to the total loss value of the initial torque prediction model to obtain a torque prediction model; The hip joint torque training data also includes time information and user information. The eigenvalue estimation of the hip joint torque is performed according to the hip joint motion information in the hip joint torque training data set to obtain the estimated torque eigenvalue, including: Selecting target user's hip joint torque training data from the hip joint torque training data set according to the user information in the hip joint torque training data; Selecting target historical hip joint torque training data from the target user's hip joint torque training data according to the time information in the hip joint torque training data; According to the hip joint motion information in the hip joint torque training data and the hip joint motion information in the target historical hip joint torque training data, the eigenvalue of the hip joint torque is estimated to obtain the estimated torque eigenvalue corresponding to the hip joint torque training data.

7. A device for predicting hip joint torque, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 6 is implemented.

8. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images