Method for modeling human muscle, and human muscle force measuring device and measuring method

Abstract

This application provides a method for modeling human muscles, a device for measuring human muscle strength, a measurement method, and a method for creating a 3D model. The method includes a pre-designed process for performing a testing phase on the muscle being tested. The testing phase involves the following steps: under positional constraints provided by the human muscle strength measuring device, the subject is instructed to perform a specified action by a motion command; surface electromyography (EMG) signals are acquired through sensors and fed to a trained recognition model for action category identification; the identified action category is compared with the motion command to determine if it conforms to the command; if it does, the strength test result from the human muscle strength measuring device is obtained, and the strength test result and EMG signals are used as human muscle modeling data for the corresponding action; the human muscle modeling data obtained from the pre-designed testing phase is processed using a preset method to establish a mechanical model of the tested muscle. This application uses strength results and EMG signals as human muscle modeling data, improving the naturalness and synergy of human-computer interaction.

Description

Technical Field

[0001] This application relates to the field of human body measurement technology, specifically to a method for modeling human muscles, a human muscle force measuring device, a method for measuring human muscle force, and a method for creating a three-dimensional model. Background Technology

[0002] Exoskeleton robots are wearable intelligent devices that can provide movement assistance, strength enhancement, or rehabilitation training for the human body, and have broad application prospects in military, industrial, medical rehabilitation, and elderly and disabled assistance fields. In a human-exoskeleton system, by recognizing the user's movement patterns in real time, the exoskeleton can proactively provide corresponding assistance or movement guidance, thereby achieving smooth and natural human-computer interaction. However, most exoskeleton robots currently employ control strategies based on predefined motion patterns and trajectories. The system drives joint movements according to a preset program, and the user can only passively follow the exoskeleton's movements, unable to flexibly adjust their actions according to their own intentions. This human-machine interaction mode not only limits the user's active participation but also easily leads to problems such as human-machine incoordination, poor wearing comfort, and low task adaptability, especially in upper limb applications where high mobility is required, its limitations are even more pronounced.

[0003] The reason for the above problems lies in the lack of a dedicated human musculoskeletal model for exoskeleton design, specifically for the human musculoskeletal system (muscles and bones). Solving this fundamental problem and establishing an accurate human musculoskeletal model would greatly assist in the design of exoskeleton robots. Therefore, there is an urgent need for a method to accurately build a human musculoskeletal model to help achieve efficient and accurate recognition of user movement intentions, thereby improving the naturalness and synergy of human-computer interaction. Summary of the Invention

[0004] This application provides a method for human muscle modeling. Accurate muscle modeling lays the foundation for forming a human musculoskeletal model, further providing a basis for solving problems such as human-machine incoordination, poor wearability, and low task adaptability. This application also provides a human muscle strength measuring device, a method for measuring human muscle strength using the aforementioned human muscle strength measuring device, and a method for creating a three-dimensional model.

[0005] This application provides a method for modeling human muscles, including: According to a pre-designed standard procedure, a set testing phase is performed on the muscles being tested, and each testing phase executes the following steps: Under the positional constraints provided by the human muscle strength measuring device, the subject is instructed to perform a specified action by a motion command; the surface electromyography (EMG) signals of the corresponding muscles are acquired through installed sensors; the measurement results of the aforementioned EMG signals are provided to a trained recognition model to identify the subject's action category; the identified action category is compared with the motion command to determine whether it conforms to the motion command; if it conforms, the strength test results obtained by the human muscle strength measuring device are acquired, and the strength test results and the EMG signals are used as human muscle modeling data for the corresponding action; the human muscle modeling data obtained from each testing phase of the set standard procedure are processed in a preset manner to establish a mechanical model for each tested muscle.

[0006] This application also provides a human muscle strength measuring device for measuring the strength test results during the process of performing the human muscle modeling method described above, comprising: a platform, at least two supports, and a base; a first end of each support is mounted on the base and is vertically extendable relative to the base, a second end of each support is hinged to the bottom surface of the platform, the supports are spaced apart, and the platform is raised and lowered by the extension and retraction of each support, and the platform is positioned at the angle required for measuring the upper limb muscle strength by extending and retracting different heights of each support; a first measuring module is provided on the platform for measuring the upper limb muscle strength, and a second measuring module is provided on the base, located below the first measuring module, for measuring the lower limb muscle strength.

[0007] This application also provides a method for measuring human muscle strength, wherein the method measures human muscle strength using any of the aforementioned human muscle strength measuring devices; the method includes: adjusting the extension and retraction height of each support member of the human muscle strength measuring device to adjust the platform of the human muscle strength measuring device to a preset flip angle and a preset height; acquiring muscle strength of the upper limb and lower limb muscles of the human body through a first measuring module set on the platform and a second measuring module set on the base of the human muscle strength measuring device; determining the maximum muscle strength information of the human body's test area based on the acquired muscle strength; and determining the muscle strength distribution information of the human body based on the maximum muscle strength information of each test area.

[0008] This application also provides a method for creating a three-dimensional model, the method comprising: determining the muscle force distribution information of a human body through a human muscle force measurement method; determining the model parameters of the three-dimensional human body model to be created based on the muscle force distribution information, so as to obtain a three-dimensional human body model that conforms to the muscle force distribution information.

[0009] Compared with the prior art, this application has the following advantages: This application provides a method for human muscle modeling, comprising: performing a set test step for the muscle to be tested according to a pre-designed standard procedure, and performing the following steps in each test step: under the positional constraints provided by a human muscle force measuring device, requiring the subject to perform a specified action by a motion command; acquiring the surface electromyography (EMG) signal of the corresponding muscle through an installed sensor; providing the measurement result of the above-mentioned EMG signal to a trained recognition model to identify the action category of the subject; comparing the identified action category with the motion command to determine whether it conforms to the motion command; if it conforms, acquiring the strength test result obtained by the human muscle force measuring device, and using the strength test result and the surface EMG signal as human muscle modeling data for the corresponding action; processing the human muscle modeling data obtained from each test step of the set standard procedure in a preset manner to establish a mechanical model of the muscle to be tested.

[0010] This application provides a method for human muscle modeling. The method involves acquiring surface electromyography (EMG) signals of corresponding muscles using installed sensors; providing the measurement results of these EMG signals to a trained recognition model to identify the subject's movement category; comparing the identified movement category with a movement command to determine if it conforms to the command; if it does, acquiring the strength test results from a human muscle force measurement device; and using the strength test results and the EMG signals as human muscle modeling data for the corresponding movement; and processing the human muscle modeling data obtained from each test stage of the set process in a preset manner to establish a mechanical model of the tested muscle. In other words, this application uses the strength test results and the EMG signals as human muscle modeling data for the corresponding movement. Because it reflects the characteristics of the human movement system from two different dimensions, this method can improve the accuracy and robustness of movement pattern recognition, thereby improving the naturalness and synergy of human-computer interaction. Attached Figure Description

[0011] Figure 1 This is a schematic diagram of the subject's movement provided in the embodiments of this application.

[0012] Figure 2 This is a schematic diagram of a CNN-BiLSTM-Attention framework for motion classification based on surface electromyography signals provided in an embodiment of this application.

[0013] Figure 3 This is a schematic diagram of the cell structure of the LSTM provided in the embodiments of this application.

[0014] Figure 4 This is a schematic diagram of the loss curve during the training process of the CNN-BiLSTM-Attention model provided in the embodiments of this application.

[0015] Figure 5 This is a schematic diagram of the accuracy curve of the training process of the CNN-BiLSTM-Attention model provided in the embodiments of this application.

[0016] Figure 6 This is a schematic diagram of the CNN (model-free processing) processing features provided in the embodiments of this application.

[0017] Figure 7 This is a schematic diagram of the CNN (after CNN layer processing) processing features provided in the embodiments of this application.

[0018] Figure 8 This is a schematic diagram of sEMG feature visualization based on t-SNE (after Bi-LSTM layer) provided in an embodiment of this application.

[0019] Figure 9 This is a schematic diagram of sEMG feature visualization (after overall processing) based on t-SNE provided in the embodiments of this application.

[0020] Figure 10 This is a schematic diagram illustrating the accuracy of different model combinations provided in the embodiments of this application.

[0021] Figure 11 This is a schematic diagram of the confusion matrix of the CNN-BiLSTM-Attention model provided in the embodiments of this application.

[0022] Figure 12 This is a schematic diagram showing the accuracy distribution of different models validated using the LOOCV method, provided in an embodiment of this application.

[0023] Figure 13 This is a schematic diagram of the structure of the human muscle strength measuring device without a housing provided in the embodiments of this application.

[0024] Figure 14 This is a schematic diagram of the structure of the human muscle strength measuring device with a housing provided in the embodiments of this application.

[0025] Figure 15 This is a schematic diagram of the connection between the first support member, the second support member, and the platform provided in the embodiments of this application.

[0026] Figure 16This is a schematic diagram of the structure provided in this application embodiment when the first measurement module is measuring and the platform is in a horizontal state.

[0027] Figure 17 This is a schematic diagram of the structure provided in this application embodiment when the first measurement module is measuring and the platform is in a vertical state.

[0028] Figure 18 This is a schematic diagram of the tibialis anterior muscle measurement module provided in this application embodiment during measurement.

[0029] Figure 19 This is a schematic diagram of the leg lift measurement module provided in this application embodiment during measurement.

[0030] Figure 20 This is a schematic diagram of the structure of the top cover plate provided in the embodiment of this application.

[0031] Figure 21 This is a schematic diagram of the structure of the support member provided in the embodiment of this application.

[0032] Figure 22 This is a schematic diagram of the structure of the first measurement module provided in the embodiment of this application.

[0033] Figure 23 This is a schematic diagram of the structure of the second measurement module provided in the embodiment of this application.

[0034] Figure 24 This application provides a flowchart for three-dimensional overall modeling.

[0035] Figure 25 This is a simplified zero-position attitude model provided in the embodiments of this application.

[0036] Figure 26 This is a simplified mathematical model of the platform in its normal posture as provided in the embodiments of this application.

[0037] Figure 27 This is a schematic diagram of the Opensim simulation provided in the embodiments of this application.

[0038] Figure label: 1: Platform; 11: Bottom surface of the platform; 12: Top surface of the platform; 2: Support component; 2-a: Lead screw; 2-b: Nut; 2-c: Telescopic rod; 21: First support component; 211: First end of the first support component; 212: Second end of the first support component; 213: First motor; 22: Second support component; 221: First end of the second support component; 222: Second end of the second support component; 223: Second motor; 3: Base; 31: Fixing plate; 311: Notch; 32: Mounting component ; 33: Cover plate; 331: Strip hole; 332: Through hole; 4: Outer shell; 5: First measuring module; 51: First measuring panel; 52: First pressure sensor; 53: First support plate; 6: Second measuring module; 61: Second measuring panel; 62: Second pressure sensor; 63: Second support plate; 64: Heightening part; 6-a: Tibialis anterior muscle measuring module; 6-b: Leg abduction measuring module; 6-c: Leg lift measuring module; 6-d: Standing abduction measuring module. Detailed Implementation

[0039] Many specific details are set forth in the following description to provide a full understanding of this application. However, this application can be implemented in many other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of this application; therefore, this application is not limited to the specific embodiments disclosed below.

[0040] In the description of this application, it should be understood that the terms "upper," "lower," "left," "right," etc., indicating orientation or positional relationships based on the orientation or positional relationships shown in the accompanying drawings, are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this application. Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0041] This application provides a method for human muscle modeling, comprising: performing a set test step for the muscle to be tested according to a pre-designed standard procedure, and performing the following steps in each test step: under the positional constraints provided by a human muscle force measuring device, requiring the subject to perform a specified action by an action command; acquiring the surface electromyography (EMG) signal of the corresponding muscle through an installed sensor; providing the measurement result of the above-mentioned EMG signal to a trained recognition model to identify the action category of the subject; comparing the identified action category with the action command to determine whether it conforms to the action command; if it conforms, acquiring the strength test result obtained by the human muscle force measuring device, and using the strength test result and the surface EMG signal as human muscle modeling data for the corresponding action; processing the human muscle modeling data obtained from each test step of the set standard procedure in a preset manner to establish a mechanical model of the muscle to be tested.

[0042] A standard procedure can be pre-designed according to specific needs. Its purpose is to obtain reliable, repeatable, and high-quality biological signals (such as electromyographic signals) through a pre-designed set of standardized experimental procedures. The standard procedure includes several testing steps, each targeting a specific muscle. In this application, the standard procedure typically includes, but is not limited to, the following elements: Muscle selection: pre-determining which key muscles to measure (e.g., the deltoid responsible for arm raising, the biceps brachii responsible for elbow flexion, etc.). Movement specifications: specifying the posture, speed, and range of motion the user must use to complete the movement. Timing: specifying the duration of each movement, the rest period, and the number of repetitions. Appropriate selection of target muscles is crucial for accurately decoding the subject's core upper limb movement intentions. This application primarily focuses on basic upper limb movement patterns in the sagittal and coronal planes, such as flexion / extension and abduction / adduction, which occur frequently in daily life and have significant functional importance. Based on this premise, this application selects three muscles that play a key role in the aforementioned movements as subjects for surface electromyography (sEMG) signal acquisition: the deltoid, biceps brachii, and triceps brachii. The rationale for selecting these three muscles is explained below from a biomechanical perspective.

[0043] The deltoid muscle (DT) is the primary executor of shoulder abduction movements and plays an irreplaceable role in arm-raising actions (such as abduction and flexion). Its anterior, middle, and posterior bundles of fibers each have a functional focus: the anterior bundle participates in shoulder flexion and internal rotation, the middle bundle dominates abduction, and the posterior bundle is responsible for extension and external rotation. Therefore, monitoring the electromyographic signals of the deltoid muscle can effectively obtain crucial information on the direction and range of motion of the shoulder joint.

[0044] The biceps brachii (BB) is the most important flexor muscle of the elbow joint, and its function is not limited to elbow flexion. Because the long head tendon crosses the shoulder joint and attaches to the supraglenoid tubercle, the muscle also plays a synergistic role during shoulder flexion. Simultaneously, this tendon is physiologically important for maintaining the dynamic stability of the glenohumeral joint, especially preventing excessive superior displacement of the humeral head during abduction. Therefore, the activity of the biceps brachii can reflect the intention to flex the elbow joint and indirectly reveal the characteristics of complex shoulder joint movements. The triceps brachii (TB) is the only extensor muscle of the elbow joint, forming a functional antagonistic pair with the biceps brachii. Its long head also crosses the shoulder joint, attaching to the subarticular tubercle of the scapula, allowing it to not only play a dominant role in elbow extension but also cooperate in shoulder extension and adduction. Acquiring electromyographic signals from the triceps brachii helps to accurately distinguish the flexion and extension states of the elbow joint and provides crucial information for identifying movement patterns involving shoulder extension.

[0045] In summary, the deltoid, biceps brachii, and triceps brachii muscles constitute the "agonist-antagonist" functional pair for the main movements of the upper limb. By synchronously monitoring their sEMG signals, a relatively complete biomechanical information foundation can be constructed, thereby more accurately identifying and predicting the user's intention to perform basic functional movements such as raising, lowering, flexing, and extending the elbow. Other elements of the standard procedure are not listed here; they can be determined based on specific work conditions.

[0046] Next, under the positional constraints provided by the human muscle strength measuring device, the subject needs to be instructed to perform a specified action via motion commands. The specific structure of the human muscle strength measuring device will be described later.

[0047] Before the test, a maximum voluntary isometric contraction (MVIC) test was conducted, simultaneously acquiring sMEG and mechanical signals. Specifically, maximum voluntary isometric contraction (MVIC) refers to the maximum force output that a subject can generate by subjectively contracting a target muscle with maximum effort while maintaining a constant joint angle. Surface electromyography (sEMG) and mechanical signals are recorded during this process. sEMG signal: records the maximum electrical activity amplitude (unit: μV) of motor unit recruitment when generating this maximum force; this is the electrophysiological evidence for the "maximum voluntary effort" in the MVIC definition. Mechanical signal: records the maximum torque or tension (unit: Nm or N) at the current joint angle; this is the quantitative representation of the "maximum force output" in the MVIC definition.

[0048] In this application, the selection of test subjects and the testing platform can be determined according to specific work conditions, and no specific restrictions are imposed. Each test subject completes one MVIC test using the MVIC testing platform. Then, according to the SENIAM guidelines for non-invasive assessment of muscles, sEMG sensors are placed at key locations, such as the middle of the deltoid (DT), biceps brachii (BB), and triceps brachii (TB), to acquire sEMG data for the corresponding muscles. These sensors can be sEMG electrodes.

[0049] like Figure 1 As shown, the test subject wears an exoskeleton and executes the following action instructions in sequence. The test subject is required to perform the specified actions according to the action instructions. The specified actions represent five states: Resting State (RS) - rest or no movement; Mild Activity (MA) - mild movement; Rapid Movement (RM) - rapid limb movement; Dynamic Load-Bearing (DLB) - dynamic load; and Static Load-Bearing (SLB) - static heavy load.

[0050] The movement categories include two dimensions: movement type and movement intensity; the movement commands include movement type commands and movement intensity commands. Specifically, movement type refers to the spatial pattern of movement or the manner of joint activity, i.e., "what movement to perform." For example, common upper limb movement types include shoulder flexion, extension, and abduction; elbow flexion and extension; and forearm pronation / supination. Figure 1 The specific values ​​of V1, V2, and V3 can be determined based on the specific task. Movement force refers to the intensity of muscle contraction or the magnitude of external load during exercise, i.e., "how much force is used." Figure 1 The F in SLB. For example: light activity, moderate exertion, maximum voluntary contraction, or static load, dynamic load, etc.

[0051] The movement patterns in this application are designed as follows: MA involves the normal arm swing during walking, while RM is designed to complete a rapid 90-degree shoulder extension within one second. To simulate static load conditions, SLB is defined as the subject holding a 25 kg dumbbell with maximum effort for 5 seconds without movement. DLB requires the subject to perform a front raise using a 7.5 kg dumbbell, with the duration depending on the subject's adaptation to the movement. During the experiment, each subject rested for 5 minutes between movements and 30 minutes between sets to prevent fatigue from affecting data quality. sEMG signals were recorded using an sEMG acquisition module. After data processing, the total recording time for each type of movement was normalized to 30 seconds. The raw signals acquired in this application are shown in Table 1: Table 1 The identified action category is compared with the action command to determine whether it conforms to the action command; if it does, the strength test result obtained by the human muscle strength measurement device is obtained, and the strength test result and the surface electromyography signal are used as human muscle modeling data for the corresponding action; the human muscle modeling data obtained from each test step of the set standard procedure are processed in a preset way to establish the mechanical model of each tested muscle.

[0052] In other words, the actual movement category of the subject (e.g., the model determines that the subject is performing "slight elbow flexion") is first compared with the expected movement instruction (e.g., the experimental procedure requires the subject to perform "slight elbow flexion"). Only when the two match (i.e., the subject completes the movement as required) is the data considered valid and data recording is triggered. It should be noted that two data streams are recorded simultaneously at this time: mechanical signal: force test results obtained from force sensors (such as dynamometers or exoskeleton joint torque sensors); electrophysiological signal: surface electromyography signals obtained from electrodes. This pair of synchronized "electro-force" data is combined and saved as the raw data for modeling the corresponding specific movement (e.g., slight elbow flexion).

[0053] It should be noted that this application uses the CNN-BiLSTM-Attention recognition model, such as... Figure 2 As shown, this recognition model utilizes the local temporal feature extraction capability of CNN (Convolutional Neural Network), the long-term temporal dependency processing capability of BiLSTM (Bidirectional Long Short-Term Memory Network), and the attention mechanism's focus on key information to achieve higher motion classification performance.

[0054] The step of providing the surface electromyography (EMG) signal measurement results to a trained recognition model to identify the subject's action category includes: processing the EMG signals obtained from multiple sensors into 100 values ​​using an overlapping sliding window mechanism. The N sEMG time-series signal matrix is ​​extracted in a predetermined manner to form a feature data set, which is then provided to the trained recognition model; where N is the number of sensors.

[0055] In this application, to process sEMG data acquired by sensors, an overlapping sliding window mechanism is first applied. This mechanism divides a long signal or data stream into multiple fixed-length continuous segments (windows) for processing. "Overlapping" means that adjacent windows share data points on the time axis. In this application, the window length is 500 milliseconds, with a 450-millisecond overlap, enhancing the inclusion of short-term dynamic information. Due to the sequential nature of sEMG data, information changes rapidly between adjacent time points. Using overlapping windows allows the model to capture more detailed changes while reducing the loss of sEMG information at window boundaries. Here, N is the number of sensors, i.e., N can be 3. Subsequently, the sEMG signal is processed into multiple 100×3 sEMG signal matrices, and then systematic feature engineering is performed to extract multi-domain discriminative patterns. Next, a 1D CNN layer is applied for local temporal feature extraction. The features extracted from the sEMG feature engineering are shown in Table 2. Table 2 Raw sEMG data contains many bias values, and directly inputting these signals into the network increases the complexity of model training. Furthermore, due to the limited size of the constructed dataset (N=10), the sEMG features of individual subjects may disproportionately affect the model. To address these challenges, this application implements systematic feature engineering to extract comprehensive time-domain, frequency-domain, and morphological features from the signals, as shown in Table 2. These features have been proven in multiple studies to effectively improve the performance of sEMG classification models. By utilizing these general features, the intrinsic properties of consistent sEMG signals across different subjects can be captured, thereby improving the model's ability to recognize motion states without being affected by individual physiological differences.

[0056] First, this application extracts time-domain features to characterize the amplitude properties of the sEMG signal. The root mean square (RMS) represents the average signal energy, reflecting the level of muscle activation, while the peak-to-peak value (PP) represents the range of muscle contraction intensity. Next, frequency-domain analysis is performed to capture the spectral characteristics of the sEMG signal. The mean frequency (MNF) and median frequency (MDF) are extracted to reveal muscle fiber recruitment patterns and potential fatigue indicators, which are crucial for identifying changes in motor unit action potentials under different movement states. Furthermore, this application calculates the waveform factor (SF) to provide waveform morphology information, and the root sum of squares (RSS) to provide a comprehensive measurement of signal intensity. RSS emphasizes peak values ​​more than RMS, making it particularly suitable for detecting transient but intense muscle activation during specific upper limb movements.

[0057] To capture subtle spectral features, this application employs Mel-frequency cepstral coefficients (MFCCs), specifically utilizing the first and third coefficients (MFCC1 and MFCC3) to detect subtle frequency distribution patterns in sEMG signals associated with different motion states. These eight features are calculated for each sEMG channel, creating a 24-dimensional feature vector. This multi-domain feature set serves as a comprehensive representation of the sEMG signal, laying a solid foundation for motion state analysis. The recognition model employs an LSTM network and incorporates an attention mechanism to weight key content, enabling the identification of action categories.

[0058] In this embodiment, eight features are calculated for each sEMG channel. These eight features are: root mean square, peak-to-peak value, shape factor, average frequency, median frequency, root sum of squares, Mel frequency cepstral coefficient (MFCC1), and Mel frequency cepstral coefficient (MFCC3). The sEMG channels are Channel 1, Channel 2, and Channel 3, and Channel 1, Channel 2, and Channel 3 can correspond to the deltoid, biceps brachii, and triceps brachii muscles, respectively. Therefore, 3 channels × 8 features = 24 dimensions, that is, a 24-dimensional feature vector is created.

[0059] For temporal feature extraction from sEMG data, this application considers recurrent neural networks (RNNs). However, RNNs may encounter vanishing or exploding gradient problems during training when processing sequential data. Long Short-Term Memory (LSTM) networks, however, can partially solve these problems. Figure 3 This refers to the data processing flow within an LSTM unit, which is the core component of the LSTM layer. The states of the LSTM layer include hidden states. (As the output of the LSTM layer at the current time step) and cell state (Responsible for transmitting information across time steps). Information conditioning is achieved through gates, which selectively allow specific information to pass through. LSTM includes three types of gates (such as...). Figure 3 (As shown in the dashed box): forget gate, input gate, and output gate. These gates are responsible for discarding irrelevant information, updating the LSTM cell state, and determining which information is included in the LSTM output state. Additionally, candidate cell states are processed through the tanh layer. Input gate The outputs are combined to determine the update of the cell state.

[0060] The parameters of an LSTM network include input weights W, recurrent weights Q, and bias b. The calculation of each gate and candidate cell state is as follows: in, This refers to the forget gate, which determines the state of a cell from the previous moment. Which information is discarded? Output value The value is between 0 and 1 (generated by the sigmoid function σ), where 0 represents "complete forgetting" and 1 represents "complete retention". Input: The input at the current time step. (e.g., the eigenvectors of an sEMG signal) and the hidden state at the previous time step. concatenated vector . Here is the weight matrix for the forget gate. This is a bias term.

[0061] This represents the input gate, which controls how much candidate information at the current time step needs to be added to the cell state. Output Also between 0 and 1, determining the update weight of candidate values. Input: Same as the forget gate, ... . Here is the weight matrix of the input gate. For bias.

[0062] This represents the output gate, whose function is to determine the current cell state. Which parts will be output to the hidden state? In the middle. Output A value between 0 and 1 is used to filter cell states. Input: Same as the forget gate, is... . Here is the weight matrix of the output gate. For bias.

[0063] This represents the candidate cell state. Its function is to generate a candidate cell state vector based on the current input and the previous hidden state. Its value range is between -1 and 1 (generated by the tanh function). It contains new information that may be added to long-term memory. Input: Same as the forget gate, ... . Here is the weight matrix of the output gate. For bias.

[0064] Cell state and hidden state The calculation is as follows: Cell status update: Hidden state output: Among them, the Gate of Oblivion Controlling the degree to which historical information is retained, input gate By controlling the degree of inclusion of candidate information and combining elements-by-element multiplication, new cell states are obtained. Output gate. The hidden output at the current time step is obtained by filtering the current cell state after tanh activation. .

[0065] Unlike traditional LSTMs, which only propagate information forward in time, Bi-LSTMs model time-series data in both forward and reverse directions, thus capturing dynamic changes in sEMG signals more comprehensively and improving the accuracy of motion pattern classification. Considering the varying importance of features under different motion states, especially in large-amplitude motions like RM and DLB, features extracted solely by CNNs and BiLSTMs may be insufficient to highlight key information closely related to specific motion patterns. Therefore, this application's recognition model introduces an attention mechanism. By adaptively allocating weights, the network focuses on the sEMG signal segment that contributes most to the current motion intent recognition. This mechanism not only enhances the robustness of classification results but also improves the interpretability of model decisions, significantly contributing to improved motion intent recognition performance. The model training process is as follows: Figure 4 and Figure 5As shown. Furthermore, to explore the mechanism by which this model distinguishes different motion states, this application first employs t-Distributed Stochastic Neighbor Embedding (t-SNE) for visualization analysis of the feature space. This dimensionality reduction method can intuitively present the data distribution of the five motion states at different processing stages, thereby revealing the model's ability to extract features layer by layer. By extracting and visualizing features after each major module, the evolution of signal representation can be systematically tracked, and the trend of its gradually increasing discriminative power can be observed.

[0066] Building upon this foundation, to further understand the focusing behavior of the attention mechanism in various motion states, this application presents the attention weight matrix in the form of a heatmap. Analysis of the attention heatmap allows for the assessment of the model's attention to each part of the input sequence and verification of its consistency with the physiological characteristics of the motion pattern. Regarding the comprehensive evaluation of classification performance, while a single accuracy metric can reflect the overall performance of the model, its limitations are particularly pronounced in multi-class tasks, especially when there is a risk of confusion between classes. For example, easily confused states such as resting state (RS) and mild activity (MA), or rapid movement (RM) and dynamic weight-bearing (DLB) cannot be clearly revealed by accuracy alone. Therefore, a confusion matrix is ​​introduced as an auxiliary evaluation tool, which can intuitively present the classification performance of each class, identify easily misclassified states, and thus provide a clear direction for subsequent optimization.

[0067] To evaluate the model's generalization ability across different individuals, the recognition model was validated using leave-one-subject-out cross-validation (LOOCV) during training. This method fully considers the inherent physiological differences among subjects, using data from each subject sequentially as the test set, with the remainder used for training, thus simulating the model's adaptability to unseen individuals. LOOCV results can reveal the model's strengths and weaknesses in cross-subject applications, particularly helpful in identifying classification weaknesses in easily confused states.

[0068] In the statistical comparison of model performance, given that the normality test results indicate that the cross-subject accuracy of some models does not conform to a normal distribution, a nonparametric Dunn test was used for multiple comparisons, combined with Bonferroni correction to control for the population error rate. This combined method provides robust significance criteria in pairwise comparisons among the five models, enhancing the statistical reliability of the conclusions.

[0069] Next, the visualization process of feature extraction will be explained. Figure 6The distribution of features in the 2D t-SNE space before deep learning model processing is shown. The visualization shows that after manual feature extraction, the sEMG signal exhibits preliminary separability between motion states. Figure 7 The data shows that although preliminary clustering patterns appear after convolution processing, there is still a significant overlap between different motion states. For example... Figure 8 As shown, introducing temporal modeling through BiLSTM layers significantly enhances the category structure, making different motion clusters more apparent. Figure 9 The results demonstrate that after processing with the complete network architecture, the motion states form well-defined and clearly separated clusters. This significant separation in the feature space verifies the effectiveness of the proposed recognition model in learning discriminative features for robust motion state classification. The impact of different module combinations on classification performance is shown in Table 3. Table 3 Through the above Table 3 and Figure 10 As can be seen, the CNN-BiLSTM-Attention recognition model proposed in this application outperforms all compared architectures on key evaluation metrics. Specifically, the recognition model in this application achieves 97.29% accuracy, 97.29% precision, 97.29% recall, and an F1 score of 0.9729, surpassing the performance of single CNN (96.00%), BiLSTM (90.30%), CNN+BiLSTM (96.64%), and CNN+Attention (94.96%) models. Furthermore, the recognition model in this application maintains a reasonable inference time of 57.20 ± 2.38 milliseconds, which is comparable to simpler architectures, demonstrating superior performance. Figure 11 The confusion matrix of the recognition model in this application shows that its clear diagonal dominance indicates high classification accuracy in all motion states. As shown in Table 4, the recognition model in this application achieves excellent performance in all motion states. The RS state performs best, with a precision of 98.39%, a recall of 98.87%, and an F1 score of 0.9863. Similarly, the SLB state also shows strong results, with a precision of 97.75% and a recall of 97.90% (F1: 0.9782). Although the RM and DLB metrics are slightly lower, with F1 scores of 0.9608 and 0.9655 respectively, they still maintain robust classification performance. MA also achieves impressive results, with a precision of 97.59% and a recall of 97.12% (F1: 0.9735), confirming the consistent performance of the model in all motion states. Specific details are shown in Table 4: Table 4 In current research on motion classification, especially upper limb motion, there is a lack of widely accepted standardized datasets, which contrasts with the mature benchmarks in the field of gesture recognition. To construct an effective evaluation system, this application selected several baseline models for comparison, including CNN+LSTM, cubic SVM (Support Vector Machine), LSTM, and DCNN (Deep Convolutional Neural Network), and compared their classification performance with the proposed model on a self-constructed dataset. Simultaneously, all algorithms were tested using leave-one-out-of-subjects cross-validation (LOOCV) to ensure a reliable assessment of generalization ability among test subjects. Table 5 summarizes the detailed comparison results of the classification accuracy of each method: Table 5 As shown in Table 5 and Figure 12 The experimental results comparing different models are shown. The recognition model of this application achieved the highest performance across all metrics, with an accuracy of 88.17±5.39%, precision of 88.76±4.97%, recall of 88.13±5.47%, and an F1 score of 0.8799±0.5555. Statistical analysis using the Dunn test with Bonferroni correction confirmed that the recognition model of this application significantly outperformed the LSTM model (p<0.001) and the DCNN model (p<0.001) in terms of F1 score. The second best was CNN-LSTM (accuracy: 77.96±10.38%), followed by the SVM model (accuracy: 77.65±6.42%). Notably, compared to CNN-LSTM (10.38%), the recognition model of this application maintained more consistent performance with a significantly lower standard deviation (5.39%), indicating better stability among different subjects. The DCNN model achieved moderate results (accuracy: 56.54 ± 24.20%), while the basic LSTM model performed the worst (accuracy: 46.71 ± 11.54%). These results demonstrate that the architecture proposed in this application provides robust sEMG signal classification and enhanced generalization ability for upper limb motion recognition in different subjects.

[0070] Based on the mechanical models of the various tested muscles established in this embodiment, the mechanical models of the various tested muscles can be further integrated to establish corresponding musculoskeletal models. For example, after establishing the mechanical models related to the various muscles of the upper limb, these mechanical models can be integrated to establish the musculoskeletal model of the entire upper limb. The specific method is not within the scope of this application and will not be described in detail here.

[0071] The human muscle strength measuring device will now be described in detail with reference to the accompanying drawings.

[0072] Figure 13 This is a schematic diagram of the structure of the human muscle strength measuring device without a housing provided in the embodiments of this application. Figure 14 This is a schematic diagram of the structure of the human muscle strength measuring device with a housing provided in the embodiments of this application. Figure 15 This is a schematic diagram of the connection between the first support member, the second support member, and the platform provided in the embodiments of this application. Figure 16 This is a schematic diagram of the structure provided in this application embodiment when the first measurement module is measuring and the platform is in a horizontal state. Figure 17 This is a schematic diagram of the structure provided in this application embodiment when the first measurement module is measuring and the platform is in a vertical state. Figure 18 This is a schematic diagram of the tibialis anterior muscle measurement module provided in this application embodiment during measurement. Figure 19 This is a schematic diagram of the leg lift measurement module provided in this application embodiment during measurement. Figure 20 This is a schematic diagram of the structure of the top cover plate provided in the embodiment of this application. Figure 21 This is a schematic diagram of the structure of the support member provided in the embodiment of this application. Figure 22 This is a schematic diagram of the structure of the first measurement module provided in the embodiment of this application. Figure 23 This is a schematic diagram of the structure of the second measurement module provided in the embodiment of this application. Figure 24 This application provides a flowchart for three-dimensional overall modeling. Figure 25 This is a simplified zero-position attitude model provided in the embodiments of this application. Figure 26 This is a simplified mathematical model of the platform in its normal posture as provided in the embodiments of this application. Figure 27 This is a schematic diagram of the Opensim simulation provided in the embodiments of this application.

[0073] This application provides a human muscle strength measuring device, including: a platform 1, at least two support members 2 and a base 3, the first end of the support member 2 is mounted on the base 3, and the support member 2 is vertically extendable relative to the base 3, the second end of the support member 2 is hinged to the bottom surface 11 of the platform, and the support members 2 are spaced apart.

[0074] like Figures 13-15As shown, the human muscle strength measuring device provided in this application includes: a platform 1, at least two support members 2, and a base 3. The first end of each support member 2 is mounted on the base 3, and the support member 2 can extend and retract vertically relative to the base 3. The second end of each support member 2 is hinged to the bottom surface 11 of the platform. The platform 1 is positioned above the support member 2, and the platform 1 is hinged to the second end of the support member 2 below it. Specifically, the second end of each support member 2 is provided with a second connecting member, which is hinged to the platform 1. It can be understood that the second end of each support member 2 is hinged to the bottom surface 11 of the platform via the second connecting member, allowing the platform 1 to rotate relative to the second end of the support member 2. Furthermore, the human muscle strength measuring device includes at least two support members 2, therefore, at least two support members 2 are spaced apart on the bottom surface 11 of the platform.

[0075] It should be noted that the platform is a square-shaped flat plate. However, the platform can also be a circular or hexagonal flat plate. Furthermore, the number of supporting components can be selected according to the platform's structure. There are no specific restrictions on the specific structure of the platform or the number of supporting components, as long as the operational requirements are met. Moreover, the overall structure of the base 3 is a frame shape, formed by interconnecting several vertical and horizontal bars to create the frame-shaped base 3. Figure 14 and Figure 17 As shown, a housing 4 is also provided on the outside of the base 3, and the housing 4 can be fixed to the outside of the base 3.

[0076] The first end of the support member 2 is mounted on the base 3, and the second end of the support member 2 is hinged to the bottom surface 11 of the platform. That is, the platform 1 and the base 3 can be indirectly connected through the support member 2, and the support member 2 can extend and retract vertically relative to the base 3 to meet the needs of people of different heights. At least two support members include a first support member 21 and a second support member 22; the first end 211 of the first support member is hinged to the base, and the first end 221 of the second support member is fixed to the base.

[0077] like Figure 15As shown, the at least two supporting members include a first supporting member 21 and a second supporting member 22. The first end 211 of the first supporting member is hinged to the base, meaning the first end 211 of the first supporting member is provided with a first connecting member. The first end 211 of the first supporting member is hinged to the base via the first connecting member, allowing the first end 211 of the first supporting member to rotate relative to the base. The first end 221 of the second supporting member is fixed to the base, and can be fixed to the base by welding, bolts, or other means, meaning the first end 211 of the second supporting member cannot rotate relative to the base.

[0078] A first motor 213 is provided at the first end 211 of the first support member. The output end of the first motor 213 is connected to the first support member to drive the first support member to extend and retract. The end of the first motor 213 opposite to the output end is hinged to the base so that the first end of the first support member is hinged to the base. A second motor is provided at the first end of the second support member. The output end of the second motor is connected to the second support member to drive the second support member to extend and retract. The end of the second motor opposite to the output end is fixed to the base so that the first end of the second support member is fixed to the base.

[0079] like Figure 15 As shown, a first motor 213 is provided at the first end 211 of the first support member. The output end of the first motor 213 is connected to the first support member 21 for transmission. When the first motor 213 rotates, it can drive the first support member 21 to rotate, that is, when the first motor 213 rotates, it drives the first support member 21 to extend and retract. The other end of the first motor 213 is hinged to the base, that is, the end opposite to the output end of the first motor is hinged to the base. It can be understood that the end opposite to the output end of the first motor 213 is hinged to the base through a first connecting member, thereby realizing that the first end of the first support member is hinged to the base.

[0080] A second motor 223 is provided at the first end 221 of the second support member. The output end of the second motor 223 is connected to the second support member 22 for transmission. When the second motor 223 rotates, it can drive the second support member 22 to rotate, that is, when the second motor 223 rotates, it drives the second support member 22 to extend and retract. The other end of the second motor 223 is fixed to the base, that is, the end opposite to the output end of the second motor 223 is fixed to the base. It can be understood that the end opposite to the output end of the second motor 223 can be fixed to the base by welding or bolts, so that the first end of the second support member 22 is fixed to the base.

[0081] In this embodiment, each support member is equipped with a corresponding motor, which provides a power source for the corresponding support member. That is, the extension and retraction of each support member are independent and do not affect each other. Specifically, a first motor 213 is provided at the first end 211 of the first support member, and a second motor 223 is provided at the first end 221 of the second support member. The first motor 213 and the second motor 223 respectively control the extension and retraction of the corresponding first support member 21 and second support member 22. Furthermore, the opposite end of the output shaft of the first motor 213 is hinged to the base via a first connector, allowing the first end 211 of the first support member to rotate relative to the base. The opposite end of the output shaft of the second motor is fixed to the base, meaning the first end 221 of the second support member cannot rotate relative to the base.

[0082] like Figures 15-16 As shown, the at least two support members include two first support members 21 and two second support members 22. The two first support members 21 and the two second support members 22 are distributed in a rectangular array on the bottom surface of the platform, and the two first support members 21 are adjacent to each other, and the two second support members 22 are adjacent to each other. In this embodiment, the number of the at least two support members can be four, namely two first support members 21 and two second support members 22. The second ends 212 of the two first support members and the second ends 222 of the two second support members are hinged to the bottom surface of the platform, and the four support members are distributed in a rectangular array on the bottom surface of the platform. The two first support members 21 are adjacent to each other, and the two second support members 22 are adjacent to each other in the width direction of the rectangular array on the bottom surface of the platform. That is, the two first support members 21 are distributed in the width direction of the rectangular array, and the two second support members 22 are also distributed in the width direction of the rectangular array.

[0083] Of course, it is not impossible that the number of the first support member 21 and the number of the second support member 22 are different. Furthermore, the number of the first support member 21 and the number of the second support member 22 can also be three, or other numbers, etc. Therefore, there is no limitation on the number of the first support member 21 and the number of the second support member 22, as long as the working requirements are met. Figure 21 As shown, the support component includes a lead screw 2-a, a nut 2-b, and a telescopic rod 2-c. The nut 2-b is threaded onto the outside of the lead screw 2-a. One end of the telescopic rod 2-c is fixed to the nut 2-b, and the telescopic rod 2-c is fitted onto the outside of the lead screw 2-a. The lead screw 2-a is used to drive the nut 2-b to move axially along the lead screw 2-a, thereby causing the nut 2-b to drive the telescopic rod 2-c to extend and retract along the lead screw 2-a, thereby raising and lowering the platform.

[0084] Specifically, the support component includes a lead screw 2-a, a nut 2-b, and a telescopic rod 2-c. The nut 2-b is threadedly connected to the outside of the lead screw 2-a, meaning the nut 2-b and the lead screw 2-a are threaded together to fit the outer side of the lead screw 2-a. Furthermore, one end of the telescopic rod 2-c is fixed to the nut 2-b, meaning the nut 2-b and the telescopic rod 2-c are fitted onto the outside of the lead screw 2-a. In other words, one end of the nut 2-b is fixed to one end of the telescopic rod 2-c, and both the nut 2-b and the telescopic rod 2-c are fitted onto the outside of the lead screw 2-a.

[0085] When the lead screw 2-a moves, the nut 2-b on the lead screw 2-a moves along with the lead screw 2-a. That is, when the lead screw 2-a rotates, it drives the nut 2-b to move axially along the lead screw 2-a, and the rotational motion can be converted into linear motion. In other words, the nut 2-b drives the telescopic rod 2-c to extend and retract along the lead screw 2-a, so that the support member can drive the platform to rise and fall through extension and retraction. The platform 1 is raised and lowered by the extension and retraction of each of the support members 2, and the platform 1 is positioned at the angle required for measuring the muscle strength of the human upper limb by extending and retracting the support members 2 to different heights. Specifically, when the lead screw 2-a moves, the lead screw 2-a drives the nut 2-b to move along the axis of the lead screw 2-a, which in turn causes the nut 2-b to drive the telescopic rod 2-c to extend and retract along the lead screw 2-a. That is, the support frame 2 raises and lowers the platform 1 above it by extending and retracting. The extension and retraction of the support members 2 can raise and lower the platform 1 to accommodate people of different heights. In other words, the extension and retraction of the support members 2 can be adjusted according to the specific height of the people being measured so that the platform reaches the height required for the measurement.

[0086] It should also be noted that the human muscle strength measuring device includes at least two support members 2, and the at least two support members 2 are spaced apart on the bottom surface 11 of the platform. Each support member 2 operates independently and does not affect the others. Each support member 2 can extend or retract to different heights, so that there is a flip angle between the platform 1 and the second end of the support member 2 in the horizontal direction, and the flip angle is 0°~90°. That is, each support member 2 can extend or retract to different heights. When each support member 2 extends or retracts to different heights, the platform 1 above the support member 2 will have a flip angle between the platform 1 and the second end of the support member 2 in the horizontal direction, and the flip angle α is 0°~90°.

[0087] like Figure 16 As shown, when the platform 1 is parallel to the horizontal direction of the second end of the support member, the flip angle is 0°. Figure 17 As shown, when the platform 1 is perpendicular to the horizontal direction of the second end of the support member, the rotation angle is 90°. Of course, the rotation angle between the platform 1 and the horizontal direction of the second end of the support member can also be other values. The telescopic height of each support member can be adjusted according to specific needs, providing measurement personnel with different heights, angles, and movements for accurate measurement and better meeting the needs of different groups. Figure 14 and Figure 20 As shown, the base 3 has a cover plate 33 on top, and the cover plate 33 has a strip hole 331. The second end of the first support member passes through the strip hole 331 and is hinged to the bottom surface of the platform. The strip hole 331 provides space for the rotation of the first support member 21.

[0088] Specifically, a cover plate 33 is provided on the top of the base 3. The cover plate 33 is fixed to the top of the base 3 by means of bolts or clips. It should be noted that the cover plate 33 is fixed to the top of the base 3, and the platform is located above the cover plate 33. There is a certain distance between the platform and the cover plate 33. This distance provides space for the platform to rotate. An appropriate distance can be set according to the specific working conditions. No specific limitation is made on the distance here.

[0089] The cover plate 33 has a strip-shaped hole 331. The second end of the first support member passes through the strip-shaped hole 331 and is hinged to the platform above. The strip-shaped hole 331 provides space for the rotation of the first support member 21. It can be understood that the second end of the first support member passes through the strip-shaped hole 331 on the cover plate 33 and is hinged to the platform above. Since the first end of the first support member is hinged to the base 3 and the second end is hinged to the bottom surface 11 of the platform, when the first motor drives the first support member 21 to rotate, the first support member 21 will rotate within the strip-shaped hole 331. In this embodiment, the number of strip-shaped holes 331 can be provided on the cover plate 33 according to the number of first support members 21, that is, the number of strip-shaped holes 331 is the same as the number of first support members 21. Moreover, the strip-shaped hole 331 can further improve the stability of the rotation of the first support member 21, allowing the first support member 21 to rotate smoothly within the strip-shaped hole 331.

[0090] Furthermore, the cover plate 33 is also provided with a through hole 332, which is located on one side of the width direction of the strip hole 331. The first end of the second support member passes through the through hole 332 and is hinged to the bottom surface 11 of the platform. That is, the first end of the second support member can be hinged to the bottom surface 11 of the platform through the through hole 332, and the number of through holes 332 is the same as the number of second support members 22. Figure 14 and Figure 16 As shown, a first measuring module 5 is provided on the platform 1, which is used to measure the muscle strength of the upper limbs. A second measuring module is provided on the base 3, located below the first measuring module 5, and is used to measure the muscle strength of the lower limbs. Specifically, the first measuring module 5 is fixed to the top surface 12 of the platform and is used to measure the muscle strength of the upper limbs. For example, according to the height of the person being tested and the requirements for upper limb measurement, the extension and tilting angles of the support are adjusted so that the height and tilting angle of the platform 1 can meet the measurement needs of the person being tested. The person being tested places the part to be measured, such as... Figure 16 and Figure 17 As shown, if a hand is placed on the first measuring module 5 and a certain force F is applied to the first measuring module 5, the first measuring module 2 can measure the force of the upper limb of the person being tested. It should be noted that the person being tested needs to perform the measurement according to the requirements, such as the hand needing to be in contact with the first measuring module 5, the application of a certain force F to the first measuring module 5, and the magnitude of F, etc. These requirements are not listed here; anything that meets industry standards is acceptable.

[0091] like Figure 22 As shown, the first measurement module 5 includes a first measurement panel 51, a first pressure sensor 52, and a first support plate 53. One side of the first support plate 53 is fixed to the top surface of the platform, the bottom of the first pressure sensor 52 is fixed to the other side of the first support plate 53, and the first measurement panel 51 is fixed to the top of the first pressure sensor 52. The first measurement panel 51 is used to provide contact with the upper limb of the human body to measure the force.

[0092] like Figure 14 and Figure 18 As shown, the first support plate 53 is fixed to the top surface 12 of the platform. Specifically, one side of the first support plate 53 is fixed to the center of the top surface 12 of the platform by bolts or welding. Alternatively, the first support plate 53 can also be fixed to other positions on the top surface 12 of the platform, such as a lower center position, to facilitate the measurement of upper limb muscle strength. A first pressure sensor 52 is fixed to the other side of the support plate 53. The bottom of the first pressure sensor 52 is fixed to the other side of the first support plate 53, and the top of the first pressure sensor 52 is fixed to one side of the first measuring panel 51. The first measuring panel 51 is used to provide contact with the upper limb of the human body for force measurement. The second measuring module is fixed to the base 3 and is located below the first measuring module 5. Multiple second measuring modules are used to measure the muscle strength of the lower limbs. These multiple modules can measure the muscle strength of different parts of the lower limbs, allowing for multi-directional measurement of lower limb muscle strength, with multiple measurement points and comprehensive measurement angles.

[0093] like Figure 23As shown, the second measuring module 6 includes a second measuring panel 61, a second pressure sensor 62, a second support plate 63, and a heightening section 64. One side of the second support plate 63 is fixed to the base, the bottom of the second pressure sensor 62 is fixed to the other side of the second support plate 63, and the second measuring panel 61 is fixed to the top of the second pressure sensor 62. The second measuring panel 61 is used to provide contact with the lower limb of the human body for force measurement. The heightening section 64 is fixed between the second support plate 63 and the second pressure sensor 62 to increase the height of the second measuring module 6. It should be noted that, compared with the first measuring module 5, the second measuring module 6 adds the heightening section 64, and the heightening section 64 is disposed between the second support plate 63 and the second pressure sensor 62 to increase the height of the second measuring module 6. Of course, if the second measuring module 6 does not need the heightening section 64, the heightening section 64 can be omitted. In this case, the structure of the second measuring module 6 is the same as that of the first measuring module 5, and the structure of the second measuring module 6 can be selected according to specific working requirements.

[0094] like Figure 13 and Figure 18 As shown, the second measurement module includes a tibialis anterior muscle measurement module 6-a. A fixing plate 31 is provided on the base 3. The first end of the support member 2 is mounted on the fixing plate 31. The fixing plate 31 and the bottom end of the base 3 have a preset distance. The tibialis anterior muscle measurement module 6-a is mounted on the bottom surface of the fixing plate 31. Specifically, a fixing plate 31 is provided on the base 3. Because the base 3 is frame-shaped, the fixing plate 31 is fixed to the inner side of the base 3. Furthermore, the first end of the support member 2 is mounted on the fixing plate 31, that is, the first end of the first support member is hinged to the fixing plate 31, and the first end of the second support member is fixed to the fixing plate 31.

[0095] It should be noted that the second measuring module 6 includes a tibialis anterior muscle measuring module 6-a. The fixing plate 31 is fixed to the inner side of the base 3 and has a preset distance from the bottom end of the base 3. This preset distance is for fixing the tibialis anterior muscle measuring module 6-a to the bottom surface of the fixing plate 31, and the tibialis anterior muscle measuring module 6-a faces the bottom end of the base 3. Furthermore, the preset distance can be determined according to specific working conditions, and no specific limitations are imposed here.

[0096] like Figure 18As shown, the person being measured raises their toes (performing a dorsiflexion movement), bending their foot and toes upwards towards their lower leg, so that the instep contacts the tibialis anterior muscle measurement module 6-a, applying a force F to it. This allows the force of the tibialis anterior muscle to be measured. It should be noted that the tibialis anterior muscle is located on the front of the lower leg. When standing and raising the toes (performing a dorsiflexion movement), the contraction of the tibialis anterior muscle can be seen and felt on the front of the lower leg. This is because the tibialis anterior muscle is one of the most prominent muscles on the front of the lower leg. Measuring the muscle strength of the tibialis anterior muscle is very important for athletes, physical therapists, and anyone concerned about lower limb health.

[0097] like Figure 13 As shown, the second measurement module also includes a leg abduction measurement module 6-b, which is higher than the fixed plate 31. The fixed plate 31 has a notch 311 for the leg to be inserted to contact and measure with the leg abduction measurement module 6-b.

[0098] Specifically, the second measuring module further includes a leg abduction measuring module 6-b, which is fixed to the inner side of the base 3 and is higher than the fixing plate 31. In other words, the leg abduction measuring module 6-b is fixed to the inner side of the base 3 above the fixing plate 31. In this embodiment, there are two leg abduction measuring modules 6-b, which are symmetrically arranged on the inner side of the base 3 above the fixing plate 31, and are horizontally oriented.

[0099] Furthermore, the fixing plate 31 is provided with a notch 311, which is used for the leg to extend into and contact the leg abduction measuring module 6-b for measurement. That is, the leg of the person being tested extends into the inner side of the base 3 through the notch 311 on the fixing plate 31, so that both sides of the person being tested's leg contact the two leg abduction measuring modules 6-b respectively, and apply a certain force to the two leg abduction measuring modules 6-b to measure the force of the person being tested's leg abduction.

[0100] In this embodiment, the fixing plate 31 can be a U-shaped plate. However, it is not excluded that the fixing plate 31 can also be other structures, such as a C-shaped plate or a V-shaped plate, as long as the notch 311 on the fixing plate 31 allows the legs of the person being tested to extend into it and contact the leg abduction measurement module 6-b for measurement. Figure 13 and Figure 19As shown, the second measurement module also includes a leg raising measurement module 6-c. A mounting component 32 is fixed on the base 3. The mounting component 32 is located above the leg abduction measurement module 6-b. The leg raising measurement module 6-c is installed on the bottom surface of the mounting component 32 and faces downward. The leg raising measurement module 6-c is located above the leg abduction measurement module 6-b.

[0101] Specifically, a mounting component 32 is fixed to the base 3. The mounting component 32 is fixed to the inner side of the base 3 and located above the leg abduction measuring module 6-b. That is, the mounting component 32 is positioned higher than the leg abduction measuring module 6-b on the inner side of the base 3. The leg lift measuring module 6-c is mounted on the bottom surface of the mounting component 32 and faces downwards. Simultaneously, the leg lift measuring module 6-c is located above the leg abduction measuring module 6-b, meaning the leg lift measuring module 6-c is higher than the leg abduction measuring module 6-b, so that the measurement of the leg lift measuring module 6-c does not affect the measurement of the leg abduction measuring module 6-b. It should be noted that the mounting component 32 can be a horizontal bar on the inner side of the base, or an I-shaped structure composed of horizontal and vertical bars, etc.

[0102] For example, such as Figure 19 As shown, the subject sits with hip flexion, maintaining a neutral body position with the thigh parallel to the ground. One thigh is raised upwards and brought into contact with the leg raise measurement module 6-c, with a certain force applied. This allows the leg raise measurement module 6-c to measure the muscle strength of the front of the thigh, specifically the quadriceps femoris or quadriceps muscle strength. The quadriceps femoris has four parts: Rectus femoris: Located in the middle of the front thigh, extending from the hip joint to the knee joint. The rectus femoris participates in both knee extension and hip flexion. Vastus lateralis: Located on the outer side of the thigh, it is the largest of the four muscles and is primarily responsible for knee extension. Vastus medialis: Located on the inner side of the thigh, it has a prominent protrusion near the knee and also participates in knee extension, playing an important role in patellar stability. Vastus intermedius: Located below the rectus femoris, close to the front of the femur. This muscle is not directly visible because it is covered by the rectus femoris, but it also participates in knee extension. Figure 14 and Figure 17As shown, the second measurement module includes a standing abduction measurement module 6-d, which is mounted on the outer surface of the base and faces outwards. Specifically, the standing abduction measurement module 6-d is fixed to the outer side of the base and faces outwards, and is used to measure the muscle strength of the gluteus medius. For example, the person being measured uses one leg as the supporting leg and performs an abduction movement with the other leg so that the side of the foot performing the abduction movement contacts the standing abduction measurement module 6-d, and a certain force is applied to measure the muscle strength of the gluteus medius.

[0103] The above describes the specific structure of a human muscle strength measuring device provided in this application. The device includes a platform 1, at least two support members 2, and a base 3. The first end of each support member 2 is mounted on the base 3, and the support member 2 is vertically extendable relative to the base 3. The second end of each support member 2 is hinged to the bottom surface 11 of the platform. The support members 2 are spaced apart, and the platform 1 is raised and lowered by the extension and retraction of each support member 2. The support members 2 are used to adjust the platform 1 to the angle required for measuring the muscle strength of the upper limbs by extending and retracting at different heights. This structure allows for the extension and retraction of the support members 2 and the flipping of the platform 1 to accommodate different heights and measurement needs of different individuals. Furthermore, a first measuring module 5 is provided on the platform 1 for measuring the muscle strength of the upper limbs, and a second measuring module 6 is provided on the base 3, located below the first measuring module 5, for measuring the muscle strength of the lower limbs. Moreover, there are multiple second measurement modules 6. Through the first measurement module 5 and multiple second measurement modules 6, multi-part measurements can be performed, and the measurement angles are relatively comprehensive. The structure is simple, the cost is low, and it has great advantages in terms of promotion and large-scale deployment.

[0104] This application also provides a method for measuring human muscle strength. The method uses the aforementioned human muscle strength measuring device to measure human strength. The method includes: adjusting the extension and retraction height of each support member of the human muscle strength measuring device to adjust the platform of the human muscle strength measuring device to a preset flip angle and a preset height; collecting muscle strength data of the upper limb and lower limb muscles through a first measuring module on the platform and a second measuring module on the base of the human muscle strength measuring device; determining the maximum muscle strength information of the body part to be measured based on the collected muscle strength; and determining the muscle strength distribution information of the body based on the maximum muscle strength information of each body part to be measured.

[0105] The human muscle strength measuring device described above is used to measure human strength. The structure of this device can be referenced from the embodiment described above, and will not be described in detail here. Before measuring the person being measured, the human muscle strength measuring device needs to be adjusted, specifically the extension and retraction height of each support member, so that the platform above the support member is adjusted to a preset flip angle and preset height to meet the measurement needs of the person being measured. The preset flip angle and preset height can be determined based on the height of the person being measured and the requirements of the body part to be measured, and are not limited here.

[0106] After adjusting the platform to the required height and rotation angle for the person being tested, a first measurement module and a second measurement module can be selected to perform measurements based on the body part to be measured. The first measurement module can measure the muscle strength of the upper limbs, such as the upper arm and shoulder joint, while the second measurement module can measure the muscle strength of the lower limbs, such as the thigh, calf, and hip joint. It should be noted that the first and second measurement modules can measure the muscle strength of their respective body parts individually or in combination. Furthermore, there can be multiple second measurement modules, and these modules can also work together. Based on the muscle strength collected by the first and second measurement modules, the collected muscle strength is analyzed to determine the maximum muscle strength information of the body part to be measured. In this embodiment, the collected muscle strength can be the magnitude of the muscle force. The maximum muscle strength information of the body part to be measured is then input into a muscle-skeleton model (OPensim) for reverse engineering to obtain the distribution of human muscle strength.

[0107] This application also provides a method for creating a three-dimensional model, the method comprising: determining the muscle force distribution information of a human body using a human muscle force measurement method; determining the model parameters of the three-dimensional human body model to be created based on the muscle force distribution information, so as to obtain a three-dimensional human body model that conforms to the muscle force distribution information.

[0108] like Figure 24 The diagram shown is a flowchart of the overall 3D modeling process. The method for creating the 3D model is as follows: An OSIM model similar to the human body shape to be measured is created in OpenSim for reverse engineering testing. The OSIM musculoskeletal model is a musculoskeletal system model created by OpenSim.

[0109] 1. Adjust the force measurement platform's posture. 2. Adjust the model's posture to match the human body's movement to simulate real-world measurement actions, i.e., the maximum force measurement experiment for each human movement. 3. Perform quantitative processing on the collected data, using the maximum force of the human body under that movement as input. 4. Establish the OPensim human body model, i.e., perform OPensim simulation to obtain the muscle exertion under specific movements and force states of the adjusted model. 5. Adjust the model's posture to match the real experiment. 6. Simulate the input of maximum mechanical information. 7. Perform muscle mechanics simulation testing. 8. Calculate and process the data; based on human biomechanics, use the maximum muscle force of a specific muscle group from the simulation results as the maximum muscle force under this musculoskeletal model. 9. Adjust and repeatedly debug the model parameters, i.e., modify the OPensim model parameters. 10. Complete the musculoskeletal model establishment.

[0110] Furthermore, the model's quality and motion were adjusted to simulate human movement in experimental scenarios, and pressure was applied to fully simulate real-world conditions. Under these conditions, the maximum force of specific muscles could be modified based on the back-calculation results. Following the same method, the maximum muscle force back-calculation model for other movements was completed, and an OSIM musculoskeletal model based on real experimental data was established. Based on this established model, assistive simulation tests were conducted using OpenIM.

[0111] Therefore, to control the 3D human body model (i.e., the robot), its kinematics need to be mathematically modeled. Based on the design model, it can be converted into a parallel mechanism, and its simplified zero-position attitude model is as follows: Figure 25 As shown, ac and bd are electric telescopic rods (i.e., ac is the second support and bd is the first support), whose length can be adjusted according to the position of the motor (first motor and second motor); cd is the base (i.e., the fixed plate), which always remains stationary; ab is the platform, whose position always changes with the length of the electric telescopic rod; ac and cd always remain perpendicular.

[0112] According to the formula for calculating degrees of freedom, we can obtain: Where n is the number of moving components, which is 4; specifically: ac inner rod (used for telescopic movement), bd outer cylinder (movable due to the rotating joint at point D), bd inner rod (used for telescopic movement), and ab platform.

[0113] PL represents the lower pair constraint number, which is 5. Specifically: Inside ac: Between the fixed base and the inner rod of ac: a sliding joint, assuming the sliding joint is at fixed point c. At point d: Between the fixed base and the outer cylinder of bd: a revolute joint. Inside bd: Between the outer cylinder of bd and the inner rod of bd: a sliding joint. At point a: Between the inner rod of ac and the platform: a revolute joint. At point b: Between the inner rod of bd and the platform: a revolute joint.

[0114] PH is the higher pair constraint number, which is 0.

[0115] Therefore, the platform has 2 degrees of freedom, and to ensure platform stability, the number of actuators should be 2. The model can control its attitude using two electrically operated telescopic rods. When the platform is in a normal attitude, its simplified mathematical model is as follows: Figure 26 As shown: A coordinate system oxy is established with point c as the origin, that is, the cd direction is the x-axis and the ca direction is the y-axis. The attitude of platform ab can be represented by the tilt angle α and the coordinates of the midpoint m of ab.

[0116] Wherein, the tilt angle α of platform ab is: The size of ∠cad is: The coordinates of point m at the midpoint of platform ab are: Therefore, the platform attitude can be represented by the tilt angle α and the height of the platform midpoint m: By solving the inverse equation, we can obtain the result when the platform target attitude (α, my) is given. Length is: Length is: The machine attitude control equation is then: in, This is the original length, which is the zero-position attitude length. According to this equation, given the target attitude, precise attitude control can be achieved for a 3D human body model.

[0117] Based on the target pose (α, my) of the platform mentioned above, the following is calculated: The length, i.e. The length is: exist Based on, combined Calculate .

[0118] Similarly, it can be calculated that .

[0119] Therefore, based on the above calculation process, it can be calculated that , The length.

[0120] Therefore, referring to the above calculation process, the extension and retraction heights of each support component of the human muscle strength measuring device can be adjusted to adjust the platform of the human muscle strength measuring device to a preset flip angle and preset height. This can be achieved by referring to the following steps: Given the target pose (a, My) of platform 1, calculate the length of the second support component. The length of the second support component is... The length, i.e. The length is: exist Based on, combined Calculate .

[0121] Similarly, the length of the first support member is calculated. The length of the first support member is... The length of can be calculated from this. .

[0122] like Figure 27 The image shows a simulation performed using Opensim, illustrating its actions and... Figure 16 The action is the same, but a reaction force is added to it. The maximum reaction force is obtained from the experimental data of human maximum force measurement. For example, data from several (e.g., n=50) maximum force experiments on the measuring device can be collected. The collected data is processed (e.g., filtering algorithm processing, noise processing, outlier removal, etc.) to obtain processed data. The processed data is then input into OPensim to obtain the maximum reaction force, thereby obtaining the true maximum mechanical information of the human body. Subsequently, the muscle force information is reversed, that is, the maximum muscle force mechanical information of the main force-generating muscles under this action is constructed through the muscle mechanical reverse deduction algorithm.

[0123] It should be noted that although several structures, components, or units for implementing the relevant functions have been mentioned in the detailed description above, this division is not mandatory. In fact, according to the specific embodiments of this application, the features and functions of two or more structures, components, or units described above can be embodied in one structure, component, or unit. Conversely, the features and functions of one structure, component, or unit described above can be further divided and embodied by multiple components, structures, or units.

[0124] Furthermore, although the various components of the components or apparatus in this application and the mounting methods between the components are described in a specific order in the accompanying drawings, this does not require or imply that the components or apparatus must be designed according to that specific component or mounting method, or that all the components shown must be included to achieve the desired result. Additional or alternative components may be omitted, multiple components may be combined into one component to achieve the corresponding function, and / or one component may be decomposed into multiple components to achieve the corresponding function, etc. Although this application has been disclosed above with reference to preferred embodiments, it is not intended to limit this application. Any person skilled in the art can make possible changes and modifications without departing from the spirit and scope of this application. Therefore, the scope of protection of this application should be determined by the scope defined in the claims of this application.

Claims

1. A method for modeling human muscles, characterized in that, include: According to a pre-designed standard procedure, the test steps are performed on the muscle being tested, and the following steps are performed for each test step: Under the positional constraints provided by the human muscle strength measuring device, the subject is required to perform a specified action by action commands; By installing sensors, surface electromyographic signals of the corresponding muscles are acquired; The measurement results of the above surface electromyography signals are provided to the trained recognition model to analyze the... Identify the types of actions taken by the tester; The identified action category is compared with the action instruction to determine whether it conforms to the action instruction; If the conditions are met, the strength test results obtained by the human muscle strength measurement device are acquired, and the strength test results and the surface electromyography signal are used as human muscle modeling data for the corresponding action. The human muscle modeling data obtained from each test stage of the established standard procedure are processed in a preset manner to establish the mechanical model of each tested muscle.

2. The method for modeling human muscles according to claim 1, characterized in that, The step of providing the above-mentioned surface electromyography signal measurement results to a trained recognition model to identify the subject's action category includes: The surface electromyography (EMG) signals obtained from multiple sensors were processed into 100 using an overlapping sliding window mechanism. The N sEMG time-series signal matrix is ​​extracted in a predetermined manner to form a feature data set, which is then provided to the trained recognition model; where N is the number of sensors.

3. The method for modeling human muscles according to claim 1, characterized in that, The action category includes two dimensions: action type and action intensity; the action command includes action type command and action intensity command.

4. A human muscle strength measuring device, used to measure the strength test results during the process of performing the human muscle modeling method according to any one of claims 1-3, characterized in that, include: Platform, at least two support components, and base; The first end of the support member is mounted on the base, and the support member is telescopically arranged relative to the base. The second end of the support member is hinged to the bottom surface of the platform. The support members are spaced apart. The platform is raised and lowered by telescopically extending and retracting the support members. The platform is positioned at the angle required for measuring the muscle strength of the human upper limb by telescopically extending and retracting the support members to different heights. The platform is provided with a first measurement module for measuring the muscle strength of the human upper limbs, and the base is provided with a second measurement module located below the first measurement module for measuring the muscle strength of the human lower limbs.

5. The human muscle strength measuring device according to claim 4, characterized in that, The at least two support members include a first support member and a second support member; The first end of the first support member is hinged to the base, and the first end of the second support member is fixed to the base.

6. The human muscle strength measuring device according to claim 4, characterized in that, The at least two support members include two first support members and two second support members. The two first support members and the two second support members are distributed in a rectangular array on the bottom surface of the platform, and the two first support members are adjacent to each other and the two second support members are adjacent to each other.

7. The human muscle strength measuring device according to claim 4, characterized in that, The second measurement module includes a tibialis anterior muscle measurement module. A fixing plate is provided on the base. The first end of the support is mounted on the fixing plate. The fixing plate is at a preset distance from the bottom end of the base. The tibialis anterior muscle measurement module is mounted on the bottom surface of the fixing plate.

8. The human muscle strength measuring device according to claim 7, characterized in that, The second measurement module also includes a leg abduction measurement module, which is higher than the fixed plate. The fixed plate has a notch for the leg to be inserted to contact and measure with the leg abduction measurement module.

9. A method for measuring human muscle strength, characterized in that, The method measures human muscle strength using the human muscle strength measuring device according to any one of claims 4-8; the method includes: Adjust the telescopic height of each support component of the human muscle strength measuring device to adjust the platform of the human muscle strength measuring device to a preset flip angle and preset height; The muscle strength of the human body is collected by the first measurement module set on the platform and the second measurement module set on the base of the human muscle strength measuring device. Based on the collected muscle strength, information on the maximum muscle strength of the body part to be tested is determined. The distribution of muscle strength in the human body is determined based on the information of the maximum muscle strength of each part of the body to be tested.

10. A method for creating a three-dimensional model, characterized in that, The method includes: The method for measuring human muscle strength according to claim 9 determines the distribution information of human muscle strength. The model parameters of the human body 3D model to be created are determined based on the muscle strength distribution information, so as to obtain a human body 3D model that conforms to the muscle strength distribution information.

Images