Human muscle state prediction method, device and equipment

Abstract

The present invention provides a method, device, and apparatus for predicting human muscle state, relating to the technical field of muscle state prediction. The method comprises: obtaining target human posture data of a target object through motion capture technology; inputting the target human posture data into a posture-muscle state mapping model, and obtaining a predicted muscle state change sequence corresponding to the target object, output by the posture-muscle state mapping model; the posture-muscle state mapping model is trained based on electromyographic signal sample data and posture sample data corresponding to the electromyographic signal sample data. The present invention predicts muscle state using human posture data and a pretrained posture-muscle state mapping model. Compared to existing technologies, this method eliminates the need for real-time electromyographic signal acquisition, reduces human labor costs, and expands application scenarios and applicable environments.

Description

Technical Field

[0001] The present invention relates to the technical field of muscle state prediction, and in particular to a method, device and equipment for predicting human muscle state. Background Art

[0002] The assessment of human muscle status is widely used in intelligent perception scenarios such as rehabilitation assessment, motion analysis, human-computer interaction, and movement quality assessment.

[0003] Existing technologies primarily assess muscle status through real-time acquisition of electromyographic signals. This requires the subject to continuously wear electrodes, resulting in high costs. Furthermore, electromyographic signals are sensitive to factors such as electrode fit and environmental interference. They are susceptible to noise in non-laboratory environments, resulting in reduced signal quality. This significantly limits the environment in which they can be used, making widespread application difficult in both time and space. Summary of the Invention

[0004] The present invention provides a method, device and equipment for predicting human muscle state, which is used to solve the defects of the existing technology that the real-time acquisition of electromyographic signals is costly, has significant restrictions on the use environment, and is not convenient for wide application in time and space. The technical solution of the present invention predicts the muscle state through human posture data and a pre-trained posture-muscle state mapping model. Compared with the existing technology, there is no need to collect electromyographic signals in real time, which reduces human labor costs and expands the use scenarios and applicable environments.

[0005] The present invention provides a method for predicting human muscle state, comprising the following steps.

[0006] Acquire target human body posture data of the target object through motion capture technology;

[0007] The target human body posture data is input into a posture-muscle state mapping model to obtain a predicted muscle state change sequence corresponding to the target object output by the posture-muscle state mapping model; the posture-muscle state mapping model is trained based on electromyographic signal sample data and posture sample data corresponding to the electromyographic signal sample data.

[0008] According to a human muscle state prediction method provided by the present invention, the posture-muscle state mapping model is obtained by training through the following steps:

[0009] Acquiring the electromyographic signal sample data and the posture sample data;

[0010] The electromyographic signal sample data and the posture sample data are respectively segmented based on a preset action cycle to obtain target segment electromyographic signals and target segment posture data corresponding to a plurality of action cycles;

[0011] Based on the target segment electromyographic signals and target segment posture data corresponding to all the action cycles, supervised learning iterative training is performed on the untrained posture-muscle state mapping model to obtain the posture-muscle state mapping model.

[0012] According to a human muscle state prediction method provided by the present invention, the myoelectric signal sample data and the posture sample data are segmented based on a preset action cycle to obtain target segment myoelectric signals and target segment posture data corresponding to each of the multiple action cycles, including:

[0013] The electromyographic signal sample data and the posture sample data are respectively segmented based on a preset action cycle to obtain segment electromyographic signal and segment posture data corresponding to each action cycle;

[0014] For each of the action cycles, performing noise reduction processing on the segment electromyographic signal corresponding to the action cycle to obtain the target segment electromyographic signal corresponding to the action cycle;

[0015] Interpolation and completion processing is performed on the paragraph posture data corresponding to the action cycle to obtain target paragraph posture data corresponding to the action cycle.

[0016] According to a method for predicting human muscle state provided by the present invention, the untrained posture-muscle state mapping model is subjected to supervised learning iterative training based on the target segment electromyographic signals and target segment posture data corresponding to all the action cycles to obtain the posture-muscle state mapping model, including:

[0017] For each of the action cycles, performing sliding window segmentation on the target segment electromyographic signal data corresponding to the action cycle based on a preset unit window time to obtain a plurality of window electromyographic signals corresponding to the target segment electromyographic signal;

[0018] For each of the window electromyographic signals, determining a manual feature of the window electromyographic signal based on the window electromyographic signal;

[0019] Clustering the electromyographic signals of each window based on the manual features corresponding to all the electromyographic signals of the window to obtain an electromyographic symbol sequence corresponding to the action cycle;

[0020] The myoelectric symbol sequences and target segment posture data corresponding to all the action cycles are input into the untrained posture-muscle state mapping model for supervised learning iterative training to obtain the posture-muscle state mapping model.

[0021] According to a human muscle state prediction method provided by the present invention, the manual features include one or more of root mean square value, zero crossing rate, slope change, waveform length, average absolute value, Willison amplitude, autoregressive coefficient, average frequency, median frequency and power spectrum ratio.

[0022] According to a human muscle state prediction method provided by the present invention, clustering the myoelectric signals of each window based on the manual features corresponding to all the myoelectric signals of the window to obtain the myoelectric symbol sequence corresponding to the action cycle includes:

[0023] Using the K-means clustering algorithm, clustering the electromyographic signals of each window based on the manual features corresponding to all the electromyographic signals of the window, and obtaining the cluster category number corresponding to each electromyographic signal of the window;

[0024] The myoelectric symbol sequence corresponding to the action cycle is determined based on the cluster category numbers corresponding to all the windowed myoelectric signals.

[0025] According to a human muscle state prediction method provided by the present invention, the untrained posture-muscle state mapping model is specifically used for:

[0026] In each supervised learning iteration round, for each action cycle, performing sliding window segmentation on the target paragraph posture data corresponding to the action cycle based on the preset unit window time to obtain a plurality of window posture data corresponding to the target paragraph posture data;

[0027] For each of the window posture data, performing feature extraction on the window posture data to obtain a window posture feature corresponding to the window posture data;

[0028] Supervised learning iterative training is performed based on all the window posture features and all the myoelectric symbol sequences corresponding to the action cycle.

[0029] According to a human muscle state prediction method provided by the present invention, the supervised learning iterative training based on all the window posture features and all the myoelectric symbol sequences corresponding to the action cycle includes:

[0030] For each of the window posture features, a time attention mechanism is used in the time dimension to weight the window posture feature for a key time period to obtain a window time posture feature corresponding to the window posture feature;

[0031] In the spatial dimension, a spatial attention mechanism is used to weight the key limb regions of the window posture feature to obtain a window spatial posture feature corresponding to the window posture feature;

[0032] Performing spatiotemporal fusion processing on all the window time posture features and all the window space posture features to obtain a posture representation vector corresponding to the action cycle;

[0033] The supervised learning iterative training is performed based on the time-aligned posture representation vector and the myoelectric symbol sequence.

[0034] The present invention also provides a human muscle state prediction device, comprising the following modules:

[0035] An acquisition module, used for acquiring target human body posture data of a target object through motion capture technology;

[0036] A prediction module is used to input the target human body posture data into a posture-muscle state mapping model to obtain a predicted muscle state change sequence corresponding to the target object output by the posture-muscle state mapping model; the posture-muscle state mapping model is trained based on electromyographic signal sample data and posture sample data corresponding to the electromyographic signal sample data.

[0037] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, any of the above-described methods for predicting human muscle state is implemented.

[0038] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the method for predicting the human muscle state as described above is implemented.

[0039] The present invention also provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the computer program implements any of the above-mentioned methods for predicting human muscle state.

[0040] The human muscle state prediction method, device, and apparatus provided by the present invention use motion capture technology to obtain target human body posture data of a target object. The target human body posture data is input into a posture-muscle state mapping model to obtain a predicted muscle state change sequence corresponding to the target object, which is output by the posture-muscle state mapping model. The posture-muscle state mapping model is trained based on electromyographic signal sample data and posture sample data corresponding to the electromyographic signal sample data. The technical solution of the present invention predicts muscle state using human posture data and a pre-trained posture-muscle state mapping model. Compared to existing technologies, this solution eliminates the need for real-time electromyographic signal acquisition, reduces human labor costs, and expands application scenarios and applicable environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the technical solutions in the present invention or the prior art, a brief introduction is given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0042] Figure 1 It is a flow chart of the human muscle state prediction method provided by the present invention.

[0043] Figure 2 It is a flow chart of the posture-muscle state mapping model training provided by the present invention.

[0044] Figure 3 It is a structural schematic diagram of the human muscle state prediction device provided by the present invention.

[0045] Figure 4 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0046] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the embodiments described are only some of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0047] In view of the above problems in the prior art, the present invention provides a method for predicting human muscle state. Figure 1 This is one of the flow charts of the method for predicting human muscle state provided by the present invention, such as Figure 1 As shown, the method includes the following steps 110 to 120.

[0048] Step 110: Acquire target human body posture data of the target object through motion capture technology.

[0049] Specifically, the target object whose muscle state is to be detected can first be determined. After the target object is determined, target human posture data of the target object can be obtained using motion capture technology. The target human posture data is time-series data representing changes in the target object's posture. Motion capture technology records the motion information of an object or human body and converts it into digital data. This motion capture technology can include one or more of mechanical motion capture (e.g., using wearable sensors), electromagnetic motion capture, optical motion capture (e.g., using a camera to capture moving images of the target object), and inertial motion capture, which are not specifically limited in this embodiment of the present invention.

[0050] Step 120: Input the target human body posture data into a posture-muscle state mapping model to obtain a predicted muscle state change sequence corresponding to the target object output by the posture-muscle state mapping model; the posture-muscle state mapping model is trained based on electromyographic signal sample data and posture sample data corresponding to the electromyographic signal sample data.

[0051] Specifically, electromyographic signal sample data can be obtained in advance. After obtaining the electromyographic signal sample data and posture sample data, an untrained posture-muscle state mapping model can be trained based on the electromyographic signal sample data and posture sample data to obtain a trained posture-muscle state mapping model.

[0052] Furthermore, the target human posture data can be input into the posture-muscle state mapping model. The posture-muscle state mapping model can predict the changes in muscle state during the corresponding action process based on the mapping relationship between the human posture data and the muscle state, and then the predicted muscle state change sequence corresponding to the target object output by the posture-muscle state mapping model can be obtained. The predicted muscle state change sequence represents the muscle state changes of the target object during the execution of the action corresponding to the target human posture data.

[0053] The human muscle state prediction method provided by the present invention uses motion capture technology to obtain target human body posture data of a target object. The target human body posture data is input into a posture-muscle state mapping model to obtain a predicted muscle state change sequence corresponding to the target object, which is output by the posture-muscle state mapping model. The posture-muscle state mapping model is trained based on electromyographic signal sample data and posture sample data corresponding to the electromyographic signal sample data. The technical solution of the present invention predicts muscle state using human posture data and a pre-trained posture-muscle state mapping model. Compared with existing technologies, this method does not require real-time electromyographic signal collection, reduces human labor costs, and expands application scenarios and applicable environments.

[0054] In one embodiment, the posture-muscle state mapping model is trained by the following steps:

[0055] Acquiring the electromyographic signal sample data and the posture sample data;

[0056] The electromyographic signal sample data and the posture sample data are respectively segmented based on a preset action cycle to obtain target segment electromyographic signals and target segment posture data corresponding to a plurality of action cycles;

[0057] Based on the target segment electromyographic signals and target segment posture data corresponding to all the action cycles, supervised learning iterative training is performed on the untrained posture-muscle state mapping model to obtain the posture-muscle state mapping model.

[0058] Specifically, Figure 2 This is a flow chart of the posture-muscle state mapping model training process provided by the present invention, such as Figure 2 As shown, the electromyographic signal sample data may include electromyographic signals corresponding to multiple actions of multiple objects. It is easy to understand that the posture data in the posture sample data and the electromyographic signals in the electromyographic signal sample data correspond to the same action of the same object. The richer the objects and actions corresponding to the electromyographic signal sample data and the posture sample data, the better the effect of the posture-muscle state mapping model obtained by subsequent training may be.

[0059] After acquiring the electromyographic signal sample data and the posture sample data, the electromyographic signal sample data and the posture sample data can be segmented based on a preset action cycle to extract individual action segments from the electromyographic signal sample data and the posture sample data, thereby obtaining target segment electromyographic signals and target segment posture data corresponding to multiple action cycles. The preset action cycle may include pre-marked action start and end points.

[0060] Furthermore, the untrained posture-muscle state mapping model can be subjected to supervised learning iterative training based on the target segment electromyographic signals and target segment posture data corresponding to all action cycles to obtain the posture-muscle state mapping model.

[0061] In the above embodiment, the myoelectric signal sample data and the posture sample data are segmented by preset action cycles, so as to facilitate subsequent feature extraction and segmented modeling for the data corresponding to each action cycle.

[0062] In one embodiment, the segmenting of the electromyographic signal sample data and the posture sample data based on the preset action cycle to obtain target segment electromyographic signals and target segment posture data corresponding to multiple action cycles includes:

[0063] The electromyographic signal sample data and the posture sample data are respectively segmented based on a preset action cycle to obtain segment electromyographic signal and segment posture data corresponding to each action cycle;

[0064] For each of the action cycles, performing noise reduction processing on the segment electromyographic signal corresponding to the action cycle to obtain the target segment electromyographic signal corresponding to the action cycle;

[0065] Interpolation and completion processing is performed on the paragraph posture data corresponding to the action cycle to obtain target paragraph posture data corresponding to the action cycle.

[0066] Specifically, if Figure 2 As shown, the electromyographic signal sample data and the posture sample data can be segmented based on the preset action cycle, and the segment electromyographic signal and segment posture data corresponding to each action cycle can be obtained.

[0067] Because low-frequency baseline drift and high-frequency electrical interference are common in raw EMG signals, noise reduction can be performed on segmented EMG signals to suppress interference from non-physiological signal components. For each movement cycle, noise reduction can be performed on the segmented EMG signals corresponding to that movement cycle. For example, a bandpass filter with a frequency range of 20 Hz to 450 Hz can be introduced to reduce the noise of the segmented EMG signals to obtain the target segmented EMG signals corresponding to that movement cycle.

[0068] Because raw pose data often suffers from missing keypoint frames due to occlusion or loss of reflective landmarks, a temporal interpolation strategy can be employed to ensure the continuity and spatial structural integrity of keypoint trajectories (e.g., upper limb elbow joints, which characterize human motion). This strategy interpolates and repairs the positions of missing keypoints in the frames, referencing the temporal trends of adjacent frames and the spatial trajectories of the same group of keypoints. Specifically, for each motion cycle, the corresponding segment pose data can be interpolated and completed to obtain the target segment pose data for that motion cycle.

[0069] In the above embodiment, the paragraph electromyographic signal is subjected to noise reduction processing to suppress the interference of non-physiological signal components, and the paragraph posture data is interpolated and completed to ensure the continuity of the key point trajectory. The obtained target paragraph electromyographic signal data and target paragraph posture data are of higher quality and stronger consistency, laying the foundation for subsequent model training.

[0070] In one embodiment, the untrained posture-muscle state mapping model is subjected to supervised learning iterative training based on the target segment electromyographic signals and target segment posture data corresponding to all the action cycles to obtain the posture-muscle state mapping model, including:

[0071] For each of the action cycles, performing sliding window segmentation on the target segment electromyographic signal data corresponding to the action cycle based on a preset unit window time to obtain a plurality of window electromyographic signals corresponding to the target segment electromyographic signal;

[0072] For each of the window electromyographic signals, determining a manual feature of the window electromyographic signal based on the window electromyographic signal;

[0073] Clustering the electromyographic signals of each window based on the manual features corresponding to all the electromyographic signals of the window to obtain an electromyographic symbol sequence corresponding to the action cycle;

[0074] The myoelectric symbol sequences and target segment posture data corresponding to all the action cycles are input into the untrained posture-muscle state mapping model for supervised learning iterative training to obtain the posture-muscle state mapping model.

[0075] Specifically, myoelectric signals are high-dimensional, continuous, and non-stationary time series with strong individual differences and muscle group specificity. Their data patterns are complex and difficult to directly map with posture data. To solve this problem, it is necessary to convert myoelectric signals into discrete, stable, and interpretable myoelectric symbol sequences, which can effectively compress while retaining key information. Figure 2 As shown, for each action cycle, the target segment electromyographic signal data corresponding to the action cycle can first be subjected to sliding window segmentation based on the preset unit window time to obtain multiple window electromyographic signals corresponding to the target segment electromyographic signals. Among them, the overlap rate corresponding to the sliding window segmentation can be set as needed, for example, it can be 50%, and the preset unit window time can also be set as needed. The embodiment of the present invention does not make specific limitations here. For example, the preset unit window time can be 50 milliseconds, and the time period corresponding to the single window electromyographic signal is also 50 milliseconds.

[0076] Furthermore, for each window EMG signal, manual features of the window EMG signal can be determined based on the window EMG signal. It should be noted that the manual features can include one or more features, for example, classic manual features such as root mean square value, zero crossing rate, and slope change. Furthermore, the EMG signals of each window can be clustered based on the manual features corresponding to all window EMG signals to obtain the EMG symbol sequence corresponding to the motion cycle.

[0077] After obtaining the myoelectric symbol sequences corresponding to each action cycle, the myoelectric symbol sequences corresponding to all action cycles and the target segment posture data can be input into the untrained posture-muscle state mapping model for supervised learning iterative training to obtain the trained posture-muscle state mapping model.

[0078] In the above embodiment, the electromyographic signal is segmented by sliding windows to ensure that the subsequent electromyographic signal can be converted into a discrete, stable and interpretable electromyographic symbol sequence, so that the mapping relationship between the electromyographic symbol sequence and the target paragraph posture data can be established when training the posture-muscle state mapping model.

[0079] In one embodiment, the manual features include one or more of root mean square value, zero crossing rate, slope change, waveform length, mean absolute value, Willison amplitude, autoregressive coefficient, mean frequency, median frequency and power spectrum ratio.

[0080] Specifically, the manual features include one or more of the following classic manual features: (1) Root mean square root (RMS): used to measure the energy level of the electrical signal and reflect the strength of muscle contraction; (2) Zero crossing rate: used to measure the number of times the EMG signal passes through the zero point and reflect the frequency of EMG signal fluctuations; (3) Slope change: used to count the number of times the slope sign of the EMG signal changes and reflect the change in the direction of the EMG signal waveform; (4) Waveform length: used to measure the complexity of the EMG signal waveform and reflect the degree of "fluctuation" of muscle activity in the EMG signal; (5) Mean absolute value: represents the average value of the absolute value of the EMG signal and reflects the overall amplitude of the EMG signal; (6) Wil =Lison amplitude: used to evaluate the frequency of amplitude changes in the electromyographic signal, and to count the number of amplitude changes in the electromyographic signal that exceed a preset threshold. The preset threshold can be set as needed; (7) Autoregressive coefficient: describes the time series characteristics of the electromyographic signal through an autoregressive model, reflecting the linear relationship between the electromyographic signal and its past values; (8) Average frequency: represents the weighted average of all frequency components in the electromyographic signal spectrum; (9) Median frequency: represents the frequency point corresponding to half of the total power in the electromyographic signal power spectrum; (10) Power spectrum ratio: represents the ratio of the power of the high frequency band to the low frequency band in the electromyographic signal spectrum, and is used to evaluate the fatigue state of the muscle. It should be noted that the manual features may also include other manual features in addition to the above features, and the embodiments of the present invention do not make specific limitations.

[0081] In the above embodiment, one or more characteristics of the root mean square value, zero crossing rate, slope change, waveform length, average absolute value, Willison amplitude, autoregressive coefficient, average frequency, median frequency and power spectrum ratio can be used to take into account both time domain and frequency domain information, and comprehensively reflect the amplitude, frequency and dynamic characteristics of muscle contraction.

[0082] In one embodiment, clustering the myoelectric signals of each window based on the manual features corresponding to all the myoelectric signals of the window to obtain the myoelectric symbol sequence corresponding to the action cycle includes:

[0083] Using the K-means clustering algorithm, clustering the electromyographic signals of each window based on the manual features corresponding to all the electromyographic signals of the window, and obtaining the cluster category number corresponding to each electromyographic signal of the window;

[0084] The myoelectric symbol sequence corresponding to the action cycle is determined based on the cluster category numbers corresponding to all the windowed myoelectric signals.

[0085] Specifically, the K-means clustering algorithm is an unsupervised learning algorithm that aims to partition a dataset into K clusters, with each cluster represented by its nearest mean (i.e., cluster center). "Unsupervised" here means that the algorithm does not require pre-labeled training data to guide the clustering process; instead, it discovers the inherent structure of the data based entirely on its own characteristics.

[0086] In this embodiment, a K-means clustering algorithm can be used to cluster the EMG signals of each window based on the manual features corresponding to the EMG signals of all windows. The objective function corresponding to the K-means clustering algorithm can be expressed by the following formula:

[0087]

[0088] in, represents minimization, Indicates the clusters, represents the number of clusters, Indicates the The cluster centers of the clusters, represents the Euclidean distance, Represents window electromyographic signal The corresponding hand-crafted features.

[0089] In the above K-means clustering algorithm, in order to determine a reasonable number of clusters , the elbow method can be used, that is, different The Sum of Squared Errors (SSE) under the value is plotted, and the SSE change curve is drawn. When the SSE increases gradually, it will decrease rapidly, but after a certain point, the decrease of SSE will slow down significantly. The inflection point corresponds to This is the optimal number of clusters.

[0090] After clustering the EMG signals of each window, the cluster category numbers corresponding to the EMG signals of each window can be obtained. For example, the cluster category numbers corresponding to the EMG signals of all windows may include 0, 1, 2, ... K -1, this cluster category number is the myoelectric symbol corresponding to the window EMG signal. The EMG symbol sequence corresponding to the action cycle can then be determined based on the cluster category numbers corresponding to all window EMG signals. It is easy to understand that this EMG symbol sequence can represent the muscle state corresponding to the action cycle. At this point, the continuous signal of each muscle group is converted into a discrete symbol sequence, forming a temporal trajectory of muscle activation.

[0091] In the above example, the K-means clustering algorithm is used to symbolize the EMG signals. This symbolization process constructs a stable and interpretable intermediate representation, facilitating subsequent semantic fusion with posture data. Furthermore, while retaining discriminative state information from muscle activity, it significantly reduces the input dimensionality, improving the stability and transfer efficiency of subsequent models.

[0092] In one embodiment, the untrained posture-muscle state mapping model is specifically used to:

[0093] In each supervised learning iteration round, for each action cycle, performing sliding window segmentation on the target paragraph posture data corresponding to the action cycle based on the preset unit window time to obtain a plurality of window posture data corresponding to the target paragraph posture data;

[0094] For each of the window posture data, performing feature extraction on the window posture data to obtain a window posture feature corresponding to the window posture data;

[0095] Supervised learning iterative training is performed based on all the window posture features and all the myoelectric symbol sequences corresponding to the action cycle.

[0096] Specifically, since posture data is usually collected through motion capture and contains the coordinates of three-dimensional key points at multiple locations, while electromyographic signals are muscle electrical activity signals that change in time steps, in each supervised learning iteration, in order to ensure the consistency of posture data and electromyographic signals in the time dimension, for each action cycle, it is necessary to perform sliding window segmentation on the target paragraph posture data corresponding to the action cycle based on the above-mentioned preset unit window time, and obtain multiple window posture data corresponding to the target paragraph posture data. It is easy to understand that the number of window posture data and window electromyographic signals corresponding to the same action cycle is the same.

[0097] Furthermore, for each window posture data, feature extraction can be performed on the window posture data to obtain the window posture features corresponding to the window posture data. Then, supervised learning iterative training can be performed based on all window posture features and all myoelectric symbol sequences corresponding to the action cycle.

[0098] In the above embodiment, the target segment posture data corresponding to the action cycle is segmented using a sliding window based on a preset unit window time, ensuring time alignment between the posture data and the electromyographic signal. This allows for a one-to-one correspondence between the window posture data and the window electromyographic signal, and subsequent fusion analysis can also effectively predict muscle activity based on data from the same time period.

[0099] In one embodiment, the supervised learning iterative training based on all the window posture features and all the myoelectric symbol sequences corresponding to the action cycle includes:

[0100] For each of the window posture features, a time attention mechanism is used in the time dimension to weight the window posture feature for a key time period to obtain a window time posture feature corresponding to the window posture feature;

[0101] In the spatial dimension, a spatial attention mechanism is used to weight the key limb regions of the window posture feature to obtain a window spatial posture feature corresponding to the window posture feature;

[0102] Performing spatiotemporal fusion processing on all the window time posture features and all the window space posture features to obtain a posture representation vector corresponding to the action cycle;

[0103] The supervised learning iterative training is performed based on the time-aligned posture representation vector and the myoelectric symbol sequence.

[0104] Specifically, the temporal attention mechanism can focus on the temporal characteristics of key posture changes within the action cycle. Within each time window and between windows, weighted attention can be applied to key time periods that are important for action classification, ensuring that the posture-muscle state mapping model effectively captures the key stages of the action. Specifically, for each window posture feature, the temporal attention mechanism can be used to weight the key time periods of the window posture feature in the temporal dimension to obtain the window temporal posture feature corresponding to that window posture feature.

[0105] Posture data usually includes multiple key points, and the spatial distribution and mutual relationship of these key points are crucial for the recognition of movements. The spatial attention mechanism can effectively analyze these spatial correlations, enhance the modeling ability of inter-joint coordination, and thus improve the prediction ability of posture data for myoelectric symbols. Through the spatial attention mechanism, the posture-muscle state mapping model can comprehensively analyze the posture changes and spatial relationships at each moment in the posture data, ensuring that the posture-muscle state mapping model not only considers the changes in the time dimension, but also fully utilizes the spatial connection between key points. That is, for each window posture feature, the spatial attention mechanism can be used to weight the key limb area of ​​the window posture feature in the spatial dimension to obtain the window spatial posture feature corresponding to the window posture feature. The key limb area can include, for example, the left upper limb, right upper limb, torso, left lower limb and right lower limb and other limb areas.

[0106] Furthermore, corresponding spatiotemporal fusion processing can be performed on all window time posture features and all window space posture features to obtain the posture representation vector corresponding to the action cycle. Then, supervised learning iterative training can be performed based on the time-aligned posture representation vector and the electromyographic symbol sequence. In the supervised learning iterative training process, the electromyographic symbol sequence is the label corresponding to the posture data, that is, the supervision information. It should be noted that in step 120, the posture-muscle state mapping model converts the target human posture data into a posture representation vector according to the above steps, and then determines the predicted muscle state change sequence corresponding to the target human posture data according to the mapping relationship between the posture representation vector and the electromyographic symbol sequence.

[0107] In the above embodiment, the window posture features are weighted in the time dimension and the space dimension respectively, and further spatiotemporal fusion is performed, so that the posture data at each moment can establish an effective mapping relationship with the corresponding electromyographic symbol label, further improving the prediction accuracy of muscle activity status. At the same time, it ensures that in scenarios where real-time electromyographic signals cannot be collected, such as in high-cost, space-constrained, or equipment-inconvenient environments, muscle status can still be accurately predicted based on posture data, thereby filling the information gap caused by the lack of electromyographic signals.

[0108] The human muscle state prediction device provided by the present invention is described below. The human muscle state prediction device described below and the human muscle state prediction method described above can be referenced to each other.

[0109] Figure 3 Schematic diagram of the structure of the human muscle state prediction device provided by the present invention. Figure 3 As shown, the human muscle state prediction device 300 includes the following modules:

[0110] An acquisition module 310 is configured to acquire target human body posture data of a target object through motion capture technology;

[0111] The prediction module 320 is used to input the target human body posture data into the posture-muscle state mapping model to obtain the predicted muscle state change sequence corresponding to the target object output by the posture-muscle state mapping model; the posture-muscle state mapping model is trained based on the electromyographic signal sample data and the posture sample data corresponding to the electromyographic signal sample data.

[0112] The human muscle state prediction device provided by the present invention uses motion capture technology to obtain target human posture data of a target object. The target human posture data is input into a posture-muscle state mapping model, and the posture-muscle state mapping model outputs a predicted muscle state change sequence corresponding to the target object. The posture-muscle state mapping model is trained based on electromyographic signal sample data and posture sample data corresponding to the electromyographic signal sample data. The technical solution of the present invention predicts muscle state using human posture data and a pre-trained posture-muscle state mapping model. Compared to existing technologies, this solution eliminates the need for real-time electromyographic signal acquisition, reduces human labor costs, and expands application scenarios and applicable environments.

[0113] In one embodiment, the human muscle state prediction device further includes a model training module, which is specifically used to:

[0114] Acquiring the electromyographic signal sample data and the posture sample data;

[0115] The electromyographic signal sample data and the posture sample data are respectively segmented based on a preset action cycle to obtain target segment electromyographic signals and target segment posture data corresponding to a plurality of action cycles;

[0116] Based on the target segment electromyographic signals and target segment posture data corresponding to all the action cycles, supervised learning iterative training is performed on the untrained posture-muscle state mapping model to obtain the posture-muscle state mapping model.

[0117] In one embodiment, the model training module is further configured to:

[0118] The electromyographic signal sample data and the posture sample data are respectively segmented based on a preset action cycle to obtain segment electromyographic signal and segment posture data corresponding to each action cycle;

[0119] For each of the action cycles, performing noise reduction processing on the segment electromyographic signal corresponding to the action cycle to obtain the target segment electromyographic signal corresponding to the action cycle;

[0120] Interpolation and completion processing is performed on the paragraph posture data corresponding to the action cycle to obtain target paragraph posture data corresponding to the action cycle.

[0121] In one embodiment, the model training module is further configured to:

[0122] For each of the action cycles, performing sliding window segmentation on the target segment electromyographic signal data corresponding to the action cycle based on a preset unit window time to obtain a plurality of window electromyographic signals corresponding to the target segment electromyographic signal;

[0123] For each of the window electromyographic signals, determining a manual feature of the window electromyographic signal based on the window electromyographic signal;

[0124] Clustering the electromyographic signals of each window based on the manual features corresponding to all the electromyographic signals of the window to obtain an electromyographic symbol sequence corresponding to the action cycle;

[0125] The myoelectric symbol sequences and target segment posture data corresponding to all the action cycles are input into the untrained posture-muscle state mapping model for supervised learning iterative training to obtain the posture-muscle state mapping model.

[0126] In one embodiment, the manual features include one or more of root mean square value, zero crossing rate, slope change, waveform length, mean absolute value, Willison amplitude, autoregressive coefficient, mean frequency, median frequency and power spectrum ratio.

[0127] In one embodiment, the model training module is further configured to:

[0128] Using the K-means clustering algorithm, clustering the electromyographic signals of each window based on the manual features corresponding to all the electromyographic signals of the window, and obtaining the cluster category number corresponding to each electromyographic signal of the window;

[0129] The myoelectric symbol sequence corresponding to the action cycle is determined based on the cluster category numbers corresponding to all the windowed myoelectric signals.

[0130] In one embodiment, the untrained posture-muscle state mapping model is specifically used to:

[0131] In each supervised learning iteration round, for each action cycle, performing sliding window segmentation on the target paragraph posture data corresponding to the action cycle based on the preset unit window time to obtain a plurality of window posture data corresponding to the target paragraph posture data;

[0132] For each of the window posture data, performing feature extraction on the window posture data to obtain a window posture feature corresponding to the window posture data;

[0133] Supervised learning iterative training is performed based on all the window posture features and all the myoelectric symbol sequences corresponding to the action cycle.

[0134] In one embodiment, the untrained posture-muscle state mapping model is further used to:

[0135] For each of the window posture features, a time attention mechanism is used in the time dimension to weight the window posture feature for a key time period to obtain a window time posture feature corresponding to the window posture feature;

[0136] In the spatial dimension, a spatial attention mechanism is used to weight the key limb regions of the window posture feature to obtain a window spatial posture feature corresponding to the window posture feature;

[0137] Performing spatiotemporal fusion processing on all the window time posture features and all the window space posture features to obtain a posture representation vector corresponding to the action cycle;

[0138] The supervised learning iterative training is performed based on the time-aligned posture representation vector and the myoelectric symbol sequence.

[0139] Figure 4 An example of a physical structure diagram of an electronic device is shown below. Figure 4 As shown, the electronic device may include: a processor 410, a communications interface 420, a memory 430, and a communication bus 440, wherein the processor 410, the communications interface 420, and the memory 430 communicate with each other via the communication bus 440. The processor 410 may call the logic instructions in the memory 430 to execute the human muscle state prediction method, which includes:

[0140] Acquire target human body posture data of the target object through motion capture technology;

[0141] The target human body posture data is input into a posture-muscle state mapping model to obtain a predicted muscle state change sequence corresponding to the target object output by the posture-muscle state mapping model; the posture-muscle state mapping model is trained based on electromyographic signal sample data and posture sample data corresponding to the electromyographic signal sample data.

[0142] Furthermore, the logic instructions in the aforementioned memory 430 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product, stored in a storage medium, includes instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, a mobile hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0143] On the other hand, the present invention further provides a computer program product, comprising a computer program, which may be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can perform the human muscle state prediction method provided by the above methods, which includes:

[0144] Acquire target human body posture data of the target object through motion capture technology;

[0145] The target human body posture data is input into a posture-muscle state mapping model to obtain a predicted muscle state change sequence corresponding to the target object output by the posture-muscle state mapping model; the posture-muscle state mapping model is trained based on electromyographic signal sample data and posture sample data corresponding to the electromyographic signal sample data.

[0146] In another aspect, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the method for predicting human muscle state provided by the above methods is implemented, and the method includes:

[0147] Acquire target human body posture data of the target object through motion capture technology;

[0148] The target human body posture data is input into a posture-muscle state mapping model to obtain a predicted muscle state change sequence corresponding to the target object output by the posture-muscle state mapping model; the posture-muscle state mapping model is trained based on electromyographic signal sample data and posture sample data corresponding to the electromyographic signal sample data.

[0149] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0150] Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented using software plus a necessary general-purpose hardware platform, or of course, hardware. Based on this understanding, the essence of the above technical solution, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, or an optical disk, and includes a number of instructions for causing a computer device (such as a personal computer, server, or network device) to execute the methods described in each embodiment or certain portions of the embodiments.

[0151] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A method for predicting human muscle state, characterized in that: include: Acquire target human body posture data of the target object through motion capture technology; Inputting the target human body posture data into a posture-muscle state mapping model to obtain a predicted muscle state change sequence corresponding to the target object output by the posture-muscle state mapping model; the posture-muscle state mapping model is trained based on electromyographic signal sample data and posture sample data corresponding to the electromyographic signal sample data; The posture-muscle state mapping model is trained by the following steps: Acquiring the electromyographic signal sample data and the posture sample data; The electromyographic signal sample data and the posture sample data are respectively segmented based on a preset action cycle to obtain target segment electromyographic signals and target segment posture data corresponding to a plurality of action cycles; Based on the target segment electromyographic signals and target segment posture data corresponding to all the action cycles, supervised learning iterative training is performed on the untrained posture-muscle state mapping model to obtain the posture-muscle state mapping model.

2. The human muscle state prediction method according to claim 1, wherein: The step of segmenting the electromyographic signal sample data and the posture sample data based on a preset action cycle to obtain target segment electromyographic signals and target segment posture data corresponding to a plurality of action cycles includes: The electromyographic signal sample data and the posture sample data are respectively segmented based on a preset action cycle to obtain segment electromyographic signal and segment posture data corresponding to each action cycle; For each of the action cycles, performing noise reduction processing on the segment electromyographic signal corresponding to the action cycle to obtain the target segment electromyographic signal corresponding to the action cycle; Interpolation and completion processing is performed on the paragraph posture data corresponding to the action cycle to obtain target paragraph posture data corresponding to the action cycle.

3. The human muscle state prediction method according to claim 1, wherein: The untrained posture-muscle state mapping model is subjected to supervised learning iterative training based on the target segment electromyographic signals and target segment posture data corresponding to all the action cycles to obtain the posture-muscle state mapping model, including: For each of the action cycles, performing sliding window segmentation on the target segment electromyographic signal data corresponding to the action cycle based on a preset unit window time to obtain a plurality of window electromyographic signals corresponding to the target segment electromyographic signal; For each of the window electromyographic signals, determining a manual feature of the window electromyographic signal based on the window electromyographic signal; Clustering the electromyographic signals of each window based on the manual features corresponding to all the electromyographic signals of the window to obtain an electromyographic symbol sequence corresponding to the action cycle; The myoelectric symbol sequences and target segment posture data corresponding to all the action cycles are input into the untrained posture-muscle state mapping model for supervised learning iterative training to obtain the posture-muscle state mapping model.

4. The method for predicting human muscle state according to claim 3, wherein: The manual features include one or more of root mean square value, zero crossing rate, slope change, waveform length, average absolute value, Willison amplitude, autoregressive coefficient, average frequency, median frequency and power spectrum ratio.

5. The method for predicting human muscle state according to claim 3, wherein: The step of clustering the electromyographic signals of each window based on the manual features corresponding to all the electromyographic signals of the window to obtain the electromyographic symbol sequence corresponding to the action cycle includes: Using the K-means clustering algorithm, clustering the electromyographic signals of each window based on the manual features corresponding to all the electromyographic signals of the window, and obtaining the cluster category number corresponding to each electromyographic signal of the window; The myoelectric symbol sequence corresponding to the action cycle is determined based on the cluster category numbers corresponding to all the windowed myoelectric signals.

6. The method for predicting human muscle state according to any one of claims 3 to 5, characterized in that: The untrained posture-muscle state mapping model is specifically used for: In each supervised learning iteration round, for each action cycle, performing sliding window segmentation on the target paragraph posture data corresponding to the action cycle based on the preset unit window time to obtain a plurality of window posture data corresponding to the target paragraph posture data; For each of the window posture data, performing feature extraction on the window posture data to obtain a window posture feature corresponding to the window posture data; Supervised learning iterative training is performed based on all the window posture features and all the myoelectric symbol sequences corresponding to the action cycle.

7. The method for predicting human muscle state according to claim 6, wherein: The supervised learning iterative training based on all the window posture features and all the myoelectric symbol sequences corresponding to the action cycle includes: For each of the window posture features, a time attention mechanism is used in the time dimension to weight the window posture feature for a key time period to obtain a window time posture feature corresponding to the window posture feature; In the spatial dimension, a spatial attention mechanism is used to weight the key limb regions of the window posture feature to obtain a window spatial posture feature corresponding to the window posture feature; Performing spatiotemporal fusion processing on all the window time posture features and all the window space posture features to obtain a posture representation vector corresponding to the action cycle; The supervised learning iterative training is performed based on the time-aligned posture representation vector and the myoelectric symbol sequence.

8. A human muscle state prediction device, characterized in that: include: An acquisition module, used for acquiring target human body posture data of a target object through motion capture technology; A prediction module is configured to input the target human body posture data into a posture-muscle state mapping model to obtain a predicted muscle state change sequence corresponding to the target object output by the posture-muscle state mapping model; the posture-muscle state mapping model is trained based on electromyographic signal sample data and posture sample data corresponding to the electromyographic signal sample data; Model training module, used to: Acquiring the electromyographic signal sample data and the posture sample data; The electromyographic signal sample data and the posture sample data are respectively segmented based on a preset action cycle to obtain target segment electromyographic signals and target segment posture data corresponding to a plurality of action cycles; Based on the target segment electromyographic signals and target segment posture data corresponding to all the action cycles, supervised learning iterative training is performed on the untrained posture-muscle state mapping model to obtain the posture-muscle state mapping model.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the human muscle state prediction method according to any one of claims 1 to 7 is implemented.

Images