Method, device and readable storage medium for detecting motion pattern of lower limb exoskeleton

Abstract

The present application discloses a motion pattern detection method, device and readable storage medium for a lower limb exoskeleton, the method comprising: obtaining a motion signal generated by a human body; preprocessing the motion signal to obtain a preprocessed signal; extracting features from the preprocessed signal using a feature extraction network to obtain a first feature; predicting the motion pattern category based on the first feature to obtain a classification result; and reconstructing the first feature using a reconstruction network to obtain a reconstructed feature; and determining the credibility of the classification result based on the reconstructed feature. In the above manner, the present application can improve the stability and accuracy of motion intention detection.

Description

Technical Field

[0001] The present application relates to the field of lower limb exoskeleton control, and in particular to a motion pattern detection method, device and readable storage medium for a lower limb exoskeleton. Background Art

[0002] As my country's population grows and the aging population deepens, the number of people with lower limb disabilities will continue to grow. Relying on scientific and technological innovation to protect and improve the quality of life of people with lower limb disabilities has become a national strategic need. Robotics technology is developing rapidly, and lower limb exoskeleton robots have become an effective tool to help people with lower limb disabilities to conduct rehabilitation exercises and assist in walking. Lower limb exoskeleton robots are wearable, human-machine integrated mechatronic devices, and are intelligent human-machine interaction systems that intersect disciplines such as biomedical engineering, mechanics, robotics, and control. The system can control the robot by identifying the wearer's movement intentions, and use the robot to assist in driving human movement to achieve functions such as standing, walking, and crossing obstacles. Therefore, it is crucial to achieve accurate and stable detection of human motion patterns.

[0003] Existing motion pattern detection for lower limb exoskeletons mainly uses mechanical signals (such as inertial measurement units, accelerometers, plantar pressure, etc.), surface electromyography signals, electroencephalogram signals, etc., and then uses common pattern recognition methods (such as support vector machines, linear discriminant analysis, dynamic Bayesian networks, etc.) for classification. Summary of the invention

[0004] The present application mainly provides a method, device and readable storage medium for detecting the motion pattern of a lower limb exoskeleton, which solves the problem of unstable motion pattern detection of a lower limb exoskeleton in the prior art.

[0005] To solve the above technical problems, the first aspect of the present application provides a motion pattern detection method for a lower limb exoskeleton, comprising: acquiring a motion signal generated by a human body; preprocessing the motion signal to obtain a preprocessed signal; using a feature extraction network to extract features from the preprocessed signal to obtain a first feature; predicting a motion pattern category based on the first feature to obtain a classification result; and reconstructing the first feature using a reconstruction network to obtain a reconstructed feature; and determining the credibility of the classification result based on the reconstructed feature.

[0006] Optionally, determining the credibility of the classification result based on the reconstructed feature includes: determining a first loss based on the reconstructed feature and a real feature corresponding to the classification result; and determining that the classification result is unreliable in response to the first loss exceeding a preset threshold.

[0007] Optionally, the first loss is calculated by the following formula: Wherein, Loss1 represents the first loss result, and M represents the dimension of the first feature; represents the real feature corresponding to the classification result, and y represents the reconstructed feature.

[0008] Optionally, the acquiring of motion signals generated by a human body includes: acquiring electromyographic signals generated by the vastus medialis, semitendinosus, tibialis anterior and medial gastrocnemius of a single leg of the human body.

[0009] Optionally, the preprocessing of the signal includes: performing overlapping sliding window processing and filtering processing on the motion signal.

[0010] Optionally, before using the feature extraction network to extract features from the preprocessed signal, the method also includes: acquiring training data, the training data including multiple signal samples; using the feature extraction network to extract features from the signal samples to obtain a second feature; using the reconstruction network to reconstruct the second feature to obtain a third feature, and predicting the motion pattern category based on the second feature to obtain a prediction result; determining a second loss based on the prediction result and the corresponding true result, and obtaining a third loss based on the third feature; and adjusting the parameters of the feature extraction network and the reconstruction network based on the second loss and the third loss.

[0011] Optionally, adjusting the parameters of the feature extraction network and the reconstruction network based on the second loss and the third loss includes: taking the sum of the second loss and the third loss setting multiples as the fourth loss; and adjusting the parameters of the feature extraction network and the reconstruction network according to the fourth loss.

[0012] Optionally, the third loss is calculated by the following formula: Wherein, Loss3 represents the third loss, X represents the set of training data, N represents the amount of training data, x represents the second feature of the signal sample, S x represents the set of third features corresponding to the training data with the same label as x, The third feature corresponding to other signal samples of the same category as x in the training data.

[0013] To solve the above-mentioned technical problems, the second aspect of the present application provides a motion pattern detection device for a lower limb exoskeleton, the device comprising a processor and a memory coupled to each other; a computer program is stored in the memory, and the processor is used to execute the computer program to implement the steps of the motion pattern detection method for a lower limb exoskeleton as provided in the first aspect above.

[0014] In order to solve the above technical problems, the third aspect of the present application provides a computer-readable storage medium, which stores program data. When the program data is executed by a processor, the steps of the lower limb exoskeleton motion pattern detection method provided in the first aspect are implemented.

[0015] The beneficial effects of the present application are as follows: Different from the prior art, the present application pre-processes the acquired motion signal generated by the human body to obtain a pre-processed signal, then uses a feature extraction network to extract features from the pre-processed signal to obtain a first feature, predicts the motion pattern category based on the first feature to obtain a classification result, and uses a reconstruction network to reconstruct the first feature to obtain a reconstructed feature, and determines the credibility of the classification result based on the reconstructed feature. The above method adds a reconstruction network, reconstructs the first feature to obtain a reconstructed feature, and judges the credibility of the classification result based on the reconstructed feature, which can determine whether the classification result is accurate or not, and improve the stability and accuracy of motion intention detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments 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 creative work.

[0017] Figure 1 This is a schematic flow chart of an embodiment of a method for detecting a motion pattern of a lower limb exoskeleton of the present application;

[0018] Figure 2 This is a schematic diagram of the position of a motion signal collection position according to an embodiment of the present application;

[0019] Figure 3 It is a structural diagram of an embodiment of a feature extraction network of the present application;

[0020] Figure 4 This is a schematic diagram of the topology structure of an embodiment of the network reconstruction of the present application;

[0021] Figure 5 This is a schematic flow chart of an embodiment of step S16 of the present application;

[0022] Figure 6 This is a schematic flow chart of an embodiment of network training of the present application;

[0023] Figure 7 This is a schematic flow chart of an embodiment of step S25 of the present application;

[0024] Figure 8It is a schematic structural block diagram of an embodiment of a motion pattern detection device for a lower limb exoskeleton of the present application;

[0025] Fig. 9 It is a schematic structural block diagram of another embodiment of the motion pattern detection device for lower limb exoskeleton of the present application;

[0026] Fig.10 It is a schematic block diagram of the structure of an embodiment of a computer-readable storage medium of the present application. DETAILED DESCRIPTION

[0027] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0028] The terms "first" and "second" in this application are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of indicated technical features. Thus, the features defined as "first" and "second" can explicitly or implicitly include at least one of the features. In the description of this application, the meaning of "multiple" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally also includes steps or units that are not listed, or optionally also includes other steps or units inherent to these processes, methods, products or devices.

[0029] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0030] See also Figure 1 , Figure 1 1 is a schematic flow chart of an embodiment of the present invention's lower limb exoskeleton motion pattern detection method. It should be noted that if there are substantially the same results, this embodiment does not necessarily refer to the above embodiment. Figure 1 The process sequence shown is limited. This embodiment includes the following steps:

[0031] Step S11: Acquire a motion signal generated by a human body.

[0032] This step uses signal acquisition equipment to collect motion signals generated by the human body in various motion scenarios. For example, an inertial measurement unit, accelerometer, plantar pressure sensor and other devices are installed on the human body to collect mechanical signals. Another example is to use a surface electromyography signal acquisition device to collect surface electromyography signals generated by the human legs. Another example is to collect the human brain electrical signals.

[0033] In one embodiment, Biometrics Datalog (surface electromyography signal acquisition device) is used to acquire human surface electromyography signals. The device has a total of 8 independent analog data acquisition channels. In this embodiment, 4 channels are used to acquire data at a sampling rate of 1000 Hz.

[0034] Optionally, in one embodiment of the present application, surface electromyographic signals of multiple muscle groups of the lower limbs of the human body are collected, for example, electromyographic signals generated by the vastus medialis, semitendinosus, tibialis anterior and medial gastrocnemius of the lower limbs of the human body can be collected. Figure 2 , Figure 2 It is a schematic diagram of the position of a motion signal collection position according to an embodiment of the present application.

[0035] Generally speaking, most of the existing motion pattern detection methods based on surface electromyography signals use common pattern recognition methods, but most of these pattern recognition methods require the collection of surface electromyography signals of more than eight channels. This embodiment only needs to collect four-channel surface electromyography signals of the human body to detect motion patterns and perform misclassification detection at the same time. For details, please refer to the description of each embodiment below.

[0036] Step S12: pre-process the motion signal to obtain a pre-processed signal.

[0037] The motion signal preprocessing may be performing overlapping sliding window processing and filtering processing on the motion signal.

[0038] Among them, first, the collected electromyographic signal can be subjected to sliding window processing to obtain multiple motion signal data segments. Optionally, in one embodiment of the present application, the collected motion signal is subjected to overlapping sliding window processing, and the signal data is divided into multiple data segments in time sequence to perform sparse processing on the dense signal data, wherein the window length of the overlapping sliding window is set to 500 milliseconds (500 samples), and the sliding step size can be set to 10 milliseconds (10 samples).

[0039] After obtaining the plurality of signal data segments, the plurality of signal data segments are filtered to obtain a preprocessed signal.

[0040] Optionally, in one embodiment of the present application, a notch filter is used to perform 50 Hz power frequency notch filtering to remove 50 Hz power frequency interference signals in multiple signal data segments, and a Butterworth high-pass filter is used to perform 10 Hz high-pass filtering to obtain a preprocessed signal. Optionally, the Butterworth high-pass filter can be second-order, third-order, or fourth-order.

[0041] Step S13: Use a feature extraction network to extract features from the preprocessed signal to obtain a first feature.

[0042] In this step, the preprocessed signal is input into the feature extraction network, and the feature extraction network performs feature extraction and outputs the first feature.

[0043] Optionally, the first feature obtained is a 7-dimensional feature.

[0044] See also Figure 3 , Figure 3 This is a structural diagram of an embodiment of the feature extraction network of the present application. The feature extraction network of this embodiment is a temporal convolutional network, which specifically includes multiple stacked dilated causal convolutional layers, weight normalization layers, activation layers, and dropout layers. The specific structure is shown in the figure. Among them, the activation layer uses the Relu function as the activation function, and the feature extraction network contains six dilation coefficients.

[0045] Optionally, the ratio of the dropout layer is set to 0.2.

[0046] In other embodiments, those skilled in the art may also select other deep learning networks as feature extraction networks to extract features from motion signals without being limited to the above-mentioned methods.

[0047] Step S14: Predict the motion pattern category based on the first feature to obtain a classification result.

[0048] This step can predict multiple motion patterns, and the motion pattern categories may include but are not limited to: standing, sitting, walking on flat ground, going up stairs, going down stairs, going up a slope, and going down a slope.

[0049] In one implementation scenario, this step can use LSTM (Long Short-Term Memory) to predict the motion pattern category of the first feature. This step first inputs the feature vector into the LSTM neural network and outputs the motion pattern category prediction result, wherein the motion pattern category prediction result includes the predicted motion pattern category and the corresponding probability, and the motion pattern category with the highest probability is used as the classification result.

[0050] By predicting the motion pattern category, this step can learn the intended movements of the human body based on the motion signal, thereby accurately controlling the lower limb exoskeleton device to assist walking and improving the stability of continuous motion intention recognition.

[0051] In other implementation scenarios, this step may also utilize other types of prediction networks to predict the motion pattern type, not limited to the above method.

[0052] Step S15: Reconstruct the first feature using the reconstruction network to obtain a reconstructed feature.

[0053] See also Figure 4 , Figure 4 1 is a schematic diagram of the topological structure of an embodiment of the reconstruction network of the present application. The reconstruction network includes an input layer, multiple hidden layers and an output layer. The network adopts a fully connected mode, wherein the dimensions of the input layer and the output layer are both 7 dimensions. The input layer of the reconstruction network of this embodiment can be connected to the output layer of the feature extraction network. The first feature output by the feature extraction network is input by the input layer, and the reconstructed feature is output by the output layer after processing.

[0054] In one embodiment, the reconstruction network includes three hidden layers, namely a first hidden layer, a second hidden layer and a third hidden layer, and the dimensions are 4, 2, and 4 respectively.

[0055] Step S16: Determine the credibility of the classification result based on the reconstructed features.

[0056] Step S14 predicts the classification result and preliminarily determines the category of the motion mode. This step determines the credibility of the classification result.

[0057] See also Figure 5 , Figure 5 is a flowchart of an embodiment of step S16 of the present application. It should be noted that if there is substantially the same result, this embodiment does not necessarily Figure 5 The process sequence shown is limited. This embodiment includes the following steps:

[0058] Step S161: Determine a first loss based on the reconstructed features and the real features corresponding to the classification results.

[0059] The real feature may be preset, or may be the first feature mean of the signal samples corresponding to the classification results in the training data.

[0060] Optionally, the first loss is calculated by the following formula:

[0061]

[0062] Among them, Loss1 represents the first loss result, and M represents the dimension of the first feature; represents the real features corresponding to the classification results, y represents the reconstructed features, is the expression of the two-norm operation.

[0063] Step S162: Determine whether the first loss exceeds a preset threshold.

[0064] The preset threshold can be determined according to actual conditions and is not limited here.

[0065] If it is determined in this step that the first loss exceeds the preset threshold, the first loss exceeds the preset threshold, indicating that the difference between the reconstructed feature and the real feature corresponding to the classification result is large, the motion mode classification is deviated, and the classification result is not credible enough, and step S163 is executed; if the first loss does not exceed the preset threshold, it is determined that the classification result is credible, and step S164 is executed.

[0066] Step S163: Determine that the classification result is unreliable.

[0067] Step S164: Determine whether the classification result is credible.

[0068] Optionally, after determining that the classification result is unreliable, the method may further include: re-performing the motion pattern category prediction operation, or correcting the classification result to obtain a new classification result.

[0069] After determining that the classification result is credible, the method may further include: controlling the lower limb exoskeleton device to perform corresponding operations according to the classification result.

[0070] Different from the prior art, the present application obtains the first feature of the motion signal through a feature extraction network, predicts the motion pattern category corresponding to the motion signal based on the first feature, further reconstructs the first feature using a reconstruction network to obtain a reconstructed feature, and makes a credibility judgment on the classification result based on the loss between the reconstructed feature and the real feature corresponding to the classification result to determine whether the classification result is credible, thereby improving the stability and accuracy of the prediction of human action intentions, improving the adaptive control of the lower limb exoskeleton device, and reducing the error rate.

[0071] See also Figure 6 , Figure 6 is a flowchart of an embodiment of network training of the present application. It should be noted that if there are substantially the same results, this embodiment does not Figure 6 The process sequence shown is limited. This embodiment includes the following steps:

[0072] Step S21: Acquire training data, where the training data includes a plurality of signal samples.

[0073] The training data may include a large number of signal samples, wherein each group of signal samples may include 4 channels of surface electromyographic signals, for example, electromyographic signals generated by the vastus medialis, semitendinosus, tibialis anterior and medial gastrocnemius of the subject's lower limbs.

[0074] Each group of signal samples is labeled with a corresponding motion mode category, for example, the motion mode categories may include but are not limited to: standing, sitting, walking on flat ground, climbing stairs, descending stairs, climbing a slope, and descending a slope. In this embodiment, the samples related to climbing a slope and descending a slope are all data collected when the subject exercises at a slope of 10°.

[0075] Step S22: extract features from the signal sample using a feature extraction network to obtain a second feature.

[0076] Before this step, the signal sample can be preprocessed using the method of step S12, and the preprocessed sample is subjected to feature extraction through a feature extraction network to obtain a second feature.

[0077] Step S23: Reconstruct the second feature using the reconstruction network to obtain the third feature, and perform category prediction based on the second feature to obtain a prediction result.

[0078] In this step, the second feature is input into the reconstruction network, the third feature is output, and the second feature is input into the prediction network to output the prediction result. The prediction result includes the predicted motion mode category and the corresponding probability, and the motion mode category with the highest probability is taken as the prediction result.

[0079] Step S24: Determine the second loss according to the predicted result and the corresponding true result, and obtain the third loss based on the third feature.

[0080] Among them, the label carried by the signal sample is the true result of the signal sample, and the second loss is calculated according to the difference between the predicted result and the label.

[0081] The second loss is the loss caused by the prediction operation, such as the cross entropy loss; the third loss is the loss caused by feature reconstruction.

[0082] In one embodiment, the third loss is calculated by the following formula:

[0083]

[0084] Among them, Loss3 represents the third loss, X represents the set of training data, N represents the amount of training data, x represents the second feature of the signal sample, S x represents the set of third features corresponding to the training data with the same label as x, Represents the third feature corresponding to other signal samples of the same category as x in the training data.

[0085] Step S25: Adjusting parameters of the feature extraction network and the reconstruction network based on the second loss and the third loss.

[0086] This step adjusts the network parameters based on the second loss and the third loss.

[0087] See also Figure 7 , Figure 7 is a flowchart of an embodiment of step S25 of the present application. It should be noted that if there is substantially the same result, this embodiment does not necessarily use Figure 7 The process sequence shown is limited. This embodiment includes the following steps:

[0088] Step S251: taking the sum of the second loss and the third loss setting multiple as the fourth loss.

[0089] Specifically, the fourth loss can be calculated by the following formula:

[0090] Loss total =Loss2+λLoss1

[0091] Among them, Loss total is the fourth loss, Loss2 is the second loss, and the range of λ is 0.001-0.01, for example, it can be 0.001, 0.005, and 0.01. λ is used to balance the loss caused by the prediction operation and the loss caused by the feature reconstruction.

[0092] Step S252: adjusting parameters of the feature extraction network and the reconstruction network according to the fourth loss.

[0093] When one of the fourth loss, the second loss, and the third loss is less than a preset loss threshold or the cumulative number of training times is greater than a set number, stop adjusting the network parameters and determine that the network pre-training is completed; otherwise, continue training until the preliminary training is completed.

[0094] See also Figure 8 , Figure 8 It is a structural schematic block diagram of an embodiment of the motion pattern detection device of the lower limb exoskeleton of the present application. The motion pattern detection device 100 of the lower limb exoskeleton includes: an acquisition module 110, a preprocessing module 120, a feature extraction module 130, a prediction module 140, a reconstruction module 150 and a credibility judgment module 160, wherein the acquisition module 110 is used to acquire the motion signal generated by the human body; the preprocessing module 120 is used to preprocess the motion signal to obtain a preprocessed signal; the feature extraction module 130 is used to extract features from the preprocessed signal using a feature extraction network to obtain a first feature; the prediction module 140 is used to predict the motion pattern category based on the first feature to obtain a classification result; the reconstruction module 150 is used to reconstruct the first feature using a reconstruction network to obtain a reconstructed feature; the credibility judgment module 160 is used to determine the credibility of the classification result based on the reconstructed feature.

[0095] For the specific methods of executing each step of each processing, please refer to the description of each step in the above-mentioned embodiment of the motion pattern detection method for the lower limb exoskeleton of the present application, which will not be repeated here.

[0096] See also Fig. 9 , Fig. 9 1 is a schematic block diagram of another embodiment of the motion pattern detection device for lower limb exoskeleton of the present application. The motion pattern detection device 200 for lower limb exoskeleton includes a processor 210 and a memory 220 coupled to each other, the memory 220 stores a computer program, and the processor 210 is used to execute the computer program to implement the motion pattern detection device method for lower limb exoskeleton described in the above embodiments.

[0097] For the description of each step of the processing execution, please refer to the description of each step of the embodiment of the motion pattern detection method for the lower limb exoskeleton of the present application, which will not be repeated here.

[0098] The memory 220 can be used to store program data and modules. The processor 210 executes various functional applications and data processing by running the program data and modules stored in the memory 220. The memory 220 may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function (such as a signal preprocessing function, a feature processing function, a prediction function, etc.), etc.; the data storage area may store data created according to the use of the motion pattern detection device 200 of the lower limb exoskeleton (such as motion signal data, feature data, network parameters, etc.), etc. In addition, the memory 220 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other volatile solid-state storage devices. Accordingly, the memory 220 may also include a memory controller to provide the processor 210 with access to the memory 220.

[0099] In each embodiment of the present application, the disclosed method and device can be implemented in other ways. For example, the various embodiments of the lower limb exoskeleton motion pattern detection device 200 described above are merely schematic. For example, the division of the 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 an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0100] 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 present embodiment.

[0101] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0102] 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 technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium.

[0103] See also Fig.10 , Fig.10 This is a schematic block diagram of the structure of an embodiment of a computer-readable storage medium of the present application. The computer-readable storage medium 300 stores program data 310. When the program data 310 is executed, the steps of each embodiment of the motion pattern detection method for a lower limb exoskeleton as described above are implemented.

[0104] For the description of each step of the processing execution, please refer to the description of each step of the embodiment of the motion pattern detection method for the lower limb exoskeleton of the present application, which will not be repeated here.

[0105] The computer-readable storage medium 300 may be a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, etc., which may store program codes.

[0106] The above descriptions are merely embodiments of the present application and are not intended to limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the present application specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A method for detecting a motion pattern of a lower limb exoskeleton, characterized in that: The method comprises: Acquire motion signals generated by the human body; Preprocessing the motion signal to obtain a preprocessed signal; Using a feature extraction network to extract features from the preprocessed signal to obtain a first feature; Predicting the motion pattern category based on the first feature to obtain a classification result; and reconstructing the first feature using a reconstruction network to obtain a reconstructed feature; Determining the credibility of the classification result based on the reconstructed feature includes: Based on the reconstructed features and the real features corresponding to the classification results, a first loss is determined; the first loss is calculated by the following formula: Wherein, Loss1 represents the first loss, and M represents the dimension of the first feature; represents the real feature corresponding to the classification result, and y represents the reconstructed feature; In response to the first loss exceeding a preset threshold, it is determined that the classification result is unreliable.

2. The method according to claim 1, characterized in that The step of obtaining a motion signal generated by a human body comprises: The electromyographic signals generated by the vastus medialis, semitendinosus, tibialis anterior and medial gastrocnemius muscles of a single leg of the human body were obtained.

3. The method according to claim 1, characterized in that: The preprocessing of the motion signal comprises: The motion signal is subjected to overlapping sliding window processing and filtering processing.

4. The method according to claim 1, characterized in that: Before extracting features from the preprocessed signal using a feature extraction network, the method further includes: Acquire training data, wherein the training data includes a plurality of signal samples; Using the feature extraction network to extract features from the signal sample to obtain a second feature; Reconstructing the second feature using the reconstruction network to obtain a third feature, and predicting the motion mode category based on the second feature to obtain a prediction result; Determine a second loss according to the predicted result and the corresponding true result, and obtain a third loss based on the third feature; Parameters of the feature extraction network and the reconstruction network are adjusted based on the second loss and the third loss.

5. The method according to claim 4, characterized in that The adjusting the parameters of the feature extraction network and the reconstruction network based on the second loss and the third loss includes: The sum of the second loss and the set multiple of the third loss is taken as the fourth loss; Parameters of the feature extraction network and the reconstruction network are adjusted according to the fourth loss.

6. The method according to claim 4, characterized in that The third loss is calculated by the following formula: Wherein, Loss3 represents the third loss, X represents the set of training data, N represents the amount of training data, x represents the second feature of the signal sample, S x represents the set of third features corresponding to signal samples with the same label as x, The third feature corresponding to other signal samples of the same category as x in the training data.

7. A motion pattern detection device for a lower limb exoskeleton, characterized in that: The device comprises a processor and a memory coupled to each other; a computer program is stored in the memory, and the processor is used to execute the computer program to implement the steps of the method according to any one of claims 1 to 6.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores program data, and when the program data is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images