Electromyographic signal processing method and device based on large model, and electronic device

Abstract

This application proposes a method, device, and electronic device for electromyography (EMG) signal processing based on a large model, which relates to the field of artificial intelligence technology. The method includes: determining the target application scenario of the user and collecting the user's initial vital signs information based on the target application scenario, the initial vital signs information including at least initial EMG signals; preprocessing the initial vital signs information to obtain target vital signs information; extracting features from the target vital signs information based on a target large model corresponding to the target application scenario to obtain feature vectors of the target vital signs information; and classifying and recognizing the feature vectors based on a target classification model to obtain the user's target action label. The large model can automatically learn richer and more abstract feature information from EMG signals, enabling the discovery of subtle feature differences. Furthermore, by combining different application scenarios, the processing of EMG signals becomes more extensive, and the fusion of feature extraction and classification improves the accuracy of classification.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular to a method and device for electromyography signal processing based on a large model, as well as an electronic device. Background Technology

[0002] Electromyography (EMG) signals are electrical signals that carry rich motion information generated during muscle contraction. However, due to the high noise levels and significant individual variations in EMG signals, extracting effective signals is extremely difficult, making it impossible to fully extract the complex information within them. This results in poor accuracy in human-computer interaction based on EMG. Summary of the Invention

[0003] The purpose of this application is to at least partially solve one of the technical problems in the related art.

[0004] Therefore, the first objective of this application is to propose a large-scale model-based electromyography (EMG) signal processing method for high-precision EMG feature extraction and recognition, thereby making EMG-based human-computer interaction more accurate.

[0005] The second objective of this application is to propose an electromyography signal processing device based on a large model.

[0006] The third objective of this application is to propose an electronic device.

[0007] The fourth objective of this application is to provide a computer-readable storage medium.

[0008] The fifth objective of this application is to provide a computer program product.

[0009] To achieve the above objectives, the first aspect of this application proposes a method for electromyographic signal processing based on a large model, comprising:

[0010] Determine the target application scenario in which the user is located, and collect the user's initial vital signs information based on the target application scenario. The initial vital signs information includes at least the initial electromyographic signal.

[0011] The initial vital signs information is preprocessed to obtain the target vital signs information;

[0012] Based on the target large model corresponding to the target application scenario, feature extraction is performed on the target vital signs information to obtain the feature vector of the initial vital signs information;

[0013] The feature vector is classified and identified based on the target classification model to obtain the user's target action label.

[0014] To achieve the above objectives, a second aspect of this application provides an electromyography signal processing device based on a large model, comprising:

[0015] The acquisition module is used to determine the target application scenario in which the user is located, and to acquire the user's initial vital signs information based on the target application scenario. The initial vital signs information includes at least initial electromyographic signals.

[0016] The preprocessing module is used to preprocess the initial vital sign information to obtain the target vital sign information;

[0017] The feature extraction module is used to extract features from the target vital signs information based on the target large model corresponding to the target application scenario, and obtain the feature vector of the initial vital signs information;

[0018] The classification and recognition module is used to classify and recognize the feature vector based on the target classification model to obtain the user's target action label.

[0019] To achieve the above objectives, a third aspect of this application provides an electronic device, comprising: a processor; and a memory communicatively connected to the processor; the memory storing computer-executable instructions; and the processor executing the computer-executable instructions stored in the memory to enable the processor to perform the large-model-based electromyography signal processing method described in the first aspect of the application.

[0020] To achieve the above objectives, a fourth aspect of this application provides a computer-readable storage medium having a computer program stored thereon, the computer instructions being used to cause the computer to execute the electromyography signal processing method based on a large model as described in the above aspect of the embodiment.

[0021] To achieve the above objectives, a fifth aspect of this application provides a computer program product, including a computer program that, when executed by a processor, implements the large-model-based electromyography signal processing method described in the above-mentioned aspect embodiment.

[0022] The electromyography (EMG) signal processing method and apparatus based on a large model provided in this application can, in conjunction with the user's target application scenario, call upon an adapted large model to extract features from the EMG signals collected in the target application scenario, and perform classification and recognition based on the extracted feature vectors to determine the user's current target action label. This application, through the large model, can automatically learn richer and more abstract feature information from EMG signals, thereby uncovering subtle feature differences, which is beneficial to improving the accuracy of classification and recognition. Furthermore, by incorporating different application scenarios, it makes the processing of EMG signals more comprehensive.

[0023] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0024] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein:

[0025] Figure 1 A schematic flowchart illustrating a large-scale electromyography signal processing method provided in this application embodiment;

[0026] Figure 2 A schematic flowchart illustrating another electromyographic signal processing method based on a large model provided in this application embodiment;

[0027] Figure 3 A schematic flowchart illustrating another electromyographic signal processing method based on a large model provided in this application embodiment;

[0028] Figure 4 A schematic flowchart illustrating a large-scale electromyography signal processing method provided in this application embodiment;

[0029] Figure 5 A schematic flowchart illustrating a large-scale electromyography signal processing method provided in this application embodiment;

[0030] Figure 6 A preprocessing procedure for electromyographic signals provided in this application embodiment;

[0031] Figure 7 This is a schematic diagram of the structure of an electromyography signal processing device based on a large model, provided in an embodiment of this application. Detailed Implementation

[0032] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0033] The electromyography signal processing method and apparatus based on a large model according to embodiments of this application are described below with reference to the accompanying drawings.

[0034] Figure 1 This is a flowchart illustrating a large-model-based electromyography signal processing method provided in an embodiment of this application. Figure 1 As shown, this large-model-based electromyography signal processing method includes, but is not limited to, the following steps:

[0035] S101, determine the target application scenario in which the user is located, and collect the user's initial vital signs information based on the target application scenario.

[0036] In some embodiments, the user’s initial vital signs information includes at least initial electromyographic signals.

[0037] In some embodiments, different application scenarios may include, but are not limited to: patient rehabilitation training scenarios in medical rehabilitation, immersive experience scenarios in virtual reality interaction, and auxiliary assistance scenarios in emergency rescue.

[0038] In some embodiments, multimodal sensors can be used to acquire electromyographic signals, taking into account different application scenarios and individual differences. Optionally, the user's vital signs information may also include other motion information. Optionally, the multimodal sensor may include, but is not limited to, electromyographic sensors, accelerometers, gyroscopes, pressure sensors, etc.

[0039] In some embodiments, the user's initial electromyographic signals can be acquired using an electromyographic sensor.

[0040] In some embodiments, a combination of multiple sensors, such as electromyography (EMG) sensors, accelerometers, gyroscopes, and pressure sensors, can be used to collect the user's EMG signals and motion information to obtain more comprehensive vital signs information.

[0041] In some embodiments, appropriate sensors and acquisition strategies can be selected based on the target application scenario in which the user is located. Optionally, a set of sensors for acquisition is determined according to the target application scenario, and the sampling area of ​​each type of sensor in the sensor set is determined according to the target application scenario. Further, corresponding sensors can be deployed in the sampling area. The sensors can collect the user's vital signs information in the sampling area. The sensors can communicate with electronic devices and feed back relevant sampling data to the electronic devices to obtain the user's initial vital signs information.

[0042] Optionally, depending on the actual target application scenario, the upper or lower limbs of the human body can be determined as the sampling area for electromyography (EMG) signals. Furthermore, individual EMG signals can be collected based on surface electrode sensors.

[0043] In some embodiments, high-sensitivity electrodes are used, combined with bioelectric signal amplification and anti-interference technology, to ensure that weak electromyographic signals can be accurately captured, and relevant information collected, such as personnel information, data collection time, and collection parameters, can be recorded and stored to construct a dataset of vital signs information.

[0044] In some embodiments, in the fields of medical rehabilitation and emergency rescue, accelerometers and gyroscopes are combined to acquire human limb movement information. Furthermore, the time of data acquisition, acquisition parameters, and personnel information can be recorded and stored to construct a dataset of vital signs information.

[0045] S102, preprocess the initial vital signs information to obtain the target vital signs information.

[0046] In some embodiments, to improve the accuracy of electromyography signal extraction and recognition, it is necessary to preprocess the initial vital sign information to obtain the target vital sign information. Optionally, the acquired initial vital sign information can be preprocessed, including a series of related operations such as filtering, amplification, and noise reduction, and the vital sign information after the related operations can be segmented to obtain the target vital sign information.

[0047] S103, based on the target large model corresponding to the target application scenario, perform feature extraction on the target vital signs information to obtain the feature vector of the target vital signs information.

[0048] In some embodiments, a large model can be used to extract features from the target's vital signs information using a dedicated deep learning accelerator, resulting in a feature vector. It is understood that the feature vector includes at least the feature information of the electromyographic signal.

[0049] In some embodiments, different application scenarios can correspond to different large models. For each application scenario, the large model can be trained based on the training sample set of the application scenario to obtain the large model corresponding to that application scenario.

[0050] In some embodiments, the large models corresponding to different application scenarios may have different input dimensions and formats for the input and output layers. Before training, the input dimensions and formats of the large model's input and output layers can be adjusted to meet the needs of the application scenario.

[0051] In some embodiments, based on the target application scenario, a target large model corresponding to the target application scenario is determined from multiple large models. Further, target characteristic information is input into the target large model, which then performs feature extraction to obtain a feature vector.

[0052] In some embodiments, the preprocessed vital signs information is input into a pre-trained target large model, and the powerful feature extraction capability of the target large model is used to automatically extract deep features from the electromyographic signals and output feature vectors.

[0053] S104, classify and identify the feature vectors based on the target classification model to obtain the user's target action label.

[0054] In some embodiments, multi-model fusion technology can be employed, combining different types of classifiers to classify and identify the extracted feature vectors, outputting classification results, which are the user's target action labels. Optionally, after obtaining the classification results, deep learning-based anomaly detection technology can be used to detect and correct outliers. Time series analysis technology can be used to smooth the classification results, improving their stability.

[0055] In some embodiments, in medical rehabilitation and emergency rescue scenarios, a deep neural network classifier combined with transfer learning techniques is used to quickly adapt to electromyography signal classification tasks. In virtual scene interaction scenarios, a convolutional neural network and support vector machine (SVM) classifier, combined with model compression techniques, are used to improve classification speed.

[0056] In some embodiments, information such as the acquired electromyographic signals, extracted feature vectors, classification results, and system parameters can be stored.

[0057] In some embodiments, distributed database technology is employed to store the classification results according to a pre-designed storage structure and selected storage method. Data mining techniques are then used to analyze and mine the stored classification results, providing users with personalized services and suggestions.

[0058] The electromyography (EMG) signal processing method based on a large model provided in this application leverages the powerful feature extraction capabilities of large models to automatically learn deep features in EMG signals, breaking the limitations of relying on manual feature extraction. Compared with other deep learning algorithms, it has significant advantages in feature extraction, technical indicators, and application scenario expansion. For example, in feature extraction, it can extract more representative and discriminative features, resulting in a significant improvement in classification accuracy; in terms of technical indicators, both processing speed and classification accuracy far surpass traditional methods and other deep learning methods; and in terms of application scenario expansion, it is not only applicable to the traditional medical rehabilitation field but can also play an important role in virtual reality interaction, emergency rescue, and other fields.

[0059] Furthermore, the feature extraction and classification processes are organically integrated. Features extracted from a large model are directly input into the classifier for classification, greatly simplifying the processing flow and improving efficiency. Moreover, the rapid conversion from raw signal to classification result reduces error accumulation in intermediate steps, enhancing system stability and reliability.

[0060] Figure 2 This is a schematic flowchart illustrating another electromyographic signal processing method based on a large model, provided as an embodiment of this application. Figure 2 As shown, this large-model-based electromyography signal processing method includes, but is not limited to, the following steps:

[0061] S201, Determine the target application scenario in which the user is located.

[0062] S202, Based on the target application scenario, determine the set of sensors used for signal acquisition, and the sampling area of ​​each type of sensor in the sensor set.

[0063] S203 receives the signal collected in the sampling area from the sensor feedback to obtain initial vital sign information, wherein the initial vital sign information includes at least the initial electromyographic signal.

[0064] For a detailed description of steps S201 to S203, please refer to the description of step S101 in the above embodiments. The steps are not repeated here.

[0065] S204 performs noise reduction, filtering, and segmentation on the initial vital signs information to obtain the target vital signs information.

[0066] In some embodiments, hardware noise reduction is performed on the initial vital signs information to obtain noise-reduced vital signs information. Optionally, the latest low-noise amplifiers and electromagnetic shielding technologies are employed to effectively reduce noise interference.

[0067] Furthermore, based on the frequency characteristics of the initial electromyography (EMG) signal and the target application scenario, the filter parameters are configured, and the noise-reduced vital sign information is filtered based on the configured filter to obtain filtered vital sign information. Optionally, the filter can be a Butterworth bandpass filter, and the filter parameters of the Butterworth bandpass filter can be dynamically adjusted according to the frequency characteristics of the EMG signal under different application scenarios.

[0068] Furthermore, the energy value of the filtered vital sign information is determined, and the filtered vital sign information is segmented based on the energy value to obtain the target vital sign information. Optionally, the energy value of the filtered vital sign information can be determined, and a signal segmentation deep learning algorithm based on the energy value can be used to segment the filtered vital sign information to improve the accuracy of segmentation.

[0069] In other words, based on conventional denoising, filtering, and segmentation operations, adaptive noise cancellation technology is introduced. This automatically adjusts filter parameters according to the noise characteristics of different application scenarios, achieving more efficient denoising. A deep learning-driven signal segmentation algorithm, combined with transfer learning technology, utilizes models pre-trained in other related fields to quickly and accurately segment continuous electromyographic signals into individual motion segments, improving segmentation accuracy and efficiency.

[0070] S205, based on the target large model corresponding to the target application scenario, feature extraction is performed on the target characteristic information to obtain the feature vector.

[0071] S206, classify and identify feature vectors based on target classification models to obtain the user's target action labels.

[0072] For a detailed description of steps S205 to S206, please refer to the description of steps S103 to S104 in the above embodiments. The steps are not repeated here.

[0073] The electromyography signal processing method based on a large model provided in this application can perform noise reduction, filtering, and segmentation on the sampled vital signs information to improve the accuracy of subsequent feature extraction, thereby obtaining better vital signs information to be input into the large model for feature extraction.

[0074] Based on the above embodiments, Figure 3 This is a schematic diagram illustrating the training process of another large model provided in an embodiment of this application. For example... Figure 3 As shown, the training process of this large model includes, but is not limited to, the following steps:

[0075] S301 collects sample vital signs information for different application scenarios.

[0076] In some embodiments, the sample vital signs information includes at least the sample electromyography signal.

[0077] In some embodiments, vital sign information is collected from samples in different sampling areas according to different application scenarios. Optionally, surface electrode sensors are used to collect electromyographic signals from different individuals, depending on the application scenario. In the fields of medical rehabilitation and emergency rescue, accelerometers and gyroscopes are combined to acquire limb movement information of the human body. Then, the data acquisition time, acquisition parameters, and personnel information are recorded and stored to construct an electromyographic signal dataset.

[0078] S302, preprocess the sample vital signs information to obtain the target vital signs information of the sample.

[0079] Employing the latest low-noise amplifiers and electromagnetic shielding technology, noise interference is effectively reduced. The Butterworth bandpass filter dynamically adjusts its parameters based on the frequency characteristics of electromyography (EMG) signals in different scenarios. A deep learning algorithm for signal segmentation based on energy thresholding improves segmentation accuracy.

[0080] S303, obtain the action labels matching the sample target's physical characteristics information, and construct a training sample set based on the sample target's physical characteristics information and action labels.

[0081] In some embodiments, preprocessed sample target vital sign information can be matched with corresponding action labels. For example, semi-supervised learning and active learning techniques can be used to reduce the annotation workload and improve annotation quality. During the annotation process, expert knowledge and experience, combined with machine learning algorithms, are used to verify and correct the annotation results.

[0082] In some embodiments, after obtaining the action labels that match the target's physical characteristics, a training sample set can be constructed using the target's physical characteristics and the action labels.

[0083] In some embodiments, to increase the sample size and enhance the generalization ability of large models, data augmentation of sample vital sign information can be performed based on generative adversarial networks (GANs) to obtain augmented vital sign information. A training sample set can then be constructed based on the sample vital sign information and the augmented vital sign information. In other words, a generative adversarial network (GAN) using an improved GAN architecture generates high-quality augmented electromyographic signals through continuous adversarial training to expand the dataset and improve the model's generalization ability.

[0084] Understandably, the enhanced vital signs information can be preprocessed to obtain the enhanced target vital signs information, and action labels can be added to the enhanced target vital signs information.

[0085] S304 adjusts the input dimensions and format of the input and output layers of the pre-trained large model for each application scenario.

[0086] In some embodiments, the selected large model can be optimized and configured as needed in terms of attention mechanism, number of network layers, parameter configuration, etc.

[0087] Optionally, the large model can be optimized for the characteristics of EMG signals, enabling it to more accurately focus on key features in the EMG signal and ignore noise interference. Through extensive experiments and optimizations, the most suitable architecture for EMG signal processing was determined regarding the number of network layers and parameter configurations. This architecture can automatically learn richer and more abstract EMG signal features and uncover more subtle feature differences.

[0088] In some embodiments, a spatiotemporal convolutional module based on an attention mechanism is introduced into the network architecture to enhance the model's ability to capture the spatiotemporal features of electromyography (EMG) signals. The input layer dynamically adjusts the input dimension and format according to the characteristics of EMG signals in different scenarios. The hidden layer employs an adaptive neuron number adjustment technique, automatically adjusting the number of hidden layer neurons based on feedback during training. An improved ReLU function, such as Leaky ReLU or PReLU, is used as the activation function to improve the model's convergence speed and performance. DropConnect technology is used for regularization to further prevent overfitting. The output layer adjusts the dimension and format of the output feature vector according to the needs of different application scenarios.

[0089] S305, fine-tuning the large model based on the training sample set to obtain the target large model corresponding to the application scenario.

[0090] In some embodiments, a large training set can be used to train a large model, enabling it to learn effective features of electromyography (EMG) signals. An adaptive learning rate adjustment strategy, such as Cosine Annealing, is employed to dynamically adjust the learning rate based on the number of training epochs. During training, model fusion techniques are used to combine multiple models with different initializations, improving the model's stability and generalization ability. Simultaneously, adversarial training techniques are employed to enhance the model's robustness.

[0091] In some embodiments, after the large model has been trained, its performance can be evaluated. For example, it can be evaluated using metrics such as accuracy, recall, and F1 score. Furthermore, information theory-based evaluation metrics, such as mutual information and information gain, can be introduced to more comprehensively evaluate the model's performance.

[0092] If the evaluation results of the large model meet the requirements, there is no need to retrain the large model, and it can be deployed for subsequent inference. If the evaluation results of the large model do not meet the requirements, it is necessary to retrain the large model to obtain a large model with satisfactory evaluation results.

[0093] The electromyography signal processing method based on a large model provided in this application utilizes the powerful feature extraction capability of the large model to automatically learn deep features in electromyography signals and uncover more subtle feature differences, which is beneficial for subsequent classification and recognition.

[0094] Based on the above embodiments, Figure 4 This is a schematic flowchart illustrating another electromyographic signal processing method based on a large model, provided as an embodiment of this application. Figure 4 As shown, this electromyography signal processing method based on a large model includes, but is not limited to, the following steps:

[0095] S401, preprocessed electromyographic signal of the sample.

[0096] S402, Build a large model.

[0097] S403, Obtain training samples.

[0098] S404, a large model trained based on samples.

[0099] S405, Evaluate the large model.

[0100] S406, determine whether the large model meets the requirements.

[0101] If the requirements are not met, return to step S404; if the requirements are met, continue to step S407.

[0102] S407, input the preprocessed electromyographic signal to be predicted.

[0103] S408 outputs the feature vector.

[0104] S409, based on feature vectors, obtains classification results.

[0105] The electromyography signal processing method based on a large model provided in this application utilizes the powerful feature extraction capability of the large model to automatically learn deep features in electromyography signals and uncover more subtle feature differences, which is beneficial for subsequent classification and recognition.

[0106] Based on the above embodiments, Figure 5 This is a schematic flowchart illustrating another electromyographic signal processing method based on a large model, provided as an embodiment of this application. Figure 5 As shown, this large-model-based electromyography signal processing method includes, but is not limited to, the following steps:

[0107] S501, determine the classification training samples of electromyographic signals.

[0108] Electromyography (EMG) signals were collected from the samples, and feature vectors were extracted from them. Furthermore, the action labels of the EMG signals were associated with the feature vectors to construct classification training samples.

[0109] S502, Construct a classification model.

[0110] In medical rehabilitation and emergency rescue scenarios, a deep neural network classifier combined with transfer learning techniques is used to quickly adapt to the classification task of electromyography signals. In virtual scene interaction scenarios, a convolutional neural network and support vector machine (SVM) classifier, combined with model compression techniques, are used to improve classification speed.

[0111] S503, training the classification model.

[0112] Based on the characteristics of different classifiers, appropriate loss functions and optimizers are selected. For deep neural network classifiers, the cross-entropy loss function and AdamW optimizer are used; for SVM classifiers, the hinge loss function and stochastic gradient descent (SGD) optimizer are used. During training, data augmentation techniques, such as rotation, scaling, and translation, are used to expand the training dataset and improve the model's generalization ability.

[0113] S504, Evaluate the classification model.

[0114] The classification model was comprehensively evaluated using methods such as leave-one-out cross-validation and K-fold cross-validation. Simultaneously, visualization techniques, such as confusion matrices and ROC curves, were used to visually demonstrate the model's performance.

[0115] S505, determine whether the classification model meets the requirements.

[0116] If the requirements are not met, return to step S504; if the requirements are met, continue to step S506.

[0117] S506, Input the feature vector extracted by the large model.

[0118] The new electromyographic signal feature vector, extracted from the target large model, is input into a trained and evaluated classification model. Based on the decision rules of different classifiers, the predicted class result is output.

[0119] S507, Classification and Recognition Results.

[0120] S508, Post-processing of classification results.

[0121] Anomaly detection technology based on deep learning is used to detect and correct outliers in the classification results. Time series analysis is used to smooth the classification results, improving their stability.

[0122] S509, Store the classification results.

[0123] Distributed database technology is used to store the classification results according to the designed storage structure and selected storage method. Data mining techniques are then used to analyze and mine the stored classification results to provide users with personalized services and suggestions.

[0124] In this embodiment, feature extraction and classification processes are organically integrated. Features extracted from a large model are directly input into the classifier for classification, greatly simplifying the processing flow and improving processing efficiency. Furthermore, the rapid conversion from raw signal to classification result reduces error accumulation in intermediate steps, improving the system's stability and reliability.

[0125] In some embodiments, homomorphic encryption algorithms are used to encrypt any vital sign information to obtain encrypted vital sign information. When any vital sign information is the initial vital sign information, model prediction is performed based on the encrypted vital sign information; when any vital sign information is from the training sample set, model training is performed based on the encrypted vital sign information. Novel storage materials and storage structures are employed to improve storage density and read / write speed. Homomorphic encryption technology is used to perform calculations and processing under encrypted data conditions, protecting user privacy.

[0126] Based on the above embodiments, Figure 6 This application provides a preprocessing procedure for electromyographic signals. For example... Figure 6 As shown, the preprocessing workflow for electromyographic signals based on a large model includes:

[0127] The electromyography (EMG) signal acquisition unit 601 can acquire EMG signals from source data. Optionally, a suitable EMG sensor and acquisition location can be determined based on the user's target application scenario. EMG signals are acquired from the source data based on the appropriate EMG sensor and acquisition location.

[0128] Signal preprocessing 602 can identify active segments in the acquired electromyographic signals and then extract valid data. Signal preprocessing 602 includes hardware noise reduction, bandpass filtering, signal segmentation, data augmentation, and storage.

[0129] Figure 7 This is a schematic diagram of a large-scale electromyography signal processing device provided in an embodiment of this application. Figure 7 As shown, the electromyography signal processing device 700 based on a large model includes: an acquisition module 701, a preprocessing module 702, a feature extraction module 703, and a classification and recognition module 704.

[0130] The acquisition module 701 is used to determine the target application scenario in which the user is located, and to acquire the user's initial vital signs information based on the target application scenario. The initial vital signs information includes at least initial electromyographic signals.

[0131] Preprocessing module 702 is used to preprocess the initial vital signs information to obtain target vital signs information;

[0132] Feature extraction module 703 is used to extract features from the target vital signs information based on the target large model corresponding to the target application scenario, and obtain the feature vector of the initial vital signs information;

[0133] The classification and recognition module 704 is used to classify and recognize the feature vector based on the target classification model to obtain the user's target action label.

[0134] In some embodiments, the preprocessing module 702 is further configured to perform noise reduction, filtering, and segmentation on the initial vital signs information to obtain the target vital signs information.

[0135] In some embodiments, the preprocessing module 702 is further configured to perform hardware noise reduction on the initial vital signs information to obtain noise-reduced vital signs information;

[0136] Based on the frequency characteristics of the initial electromyography signal and the target application scenario, the filtering parameters of the filter are configured, and the noise-reduced vital sign information is filtered based on the configured filter to obtain filtered vital sign information.

[0137] The energy value of the filtered vital signs information is determined, and the filtered vital signs information is segmented based on the energy value to obtain the target vital signs information.

[0138] In some embodiments, the feature extraction module 703 is further configured to train the target large model, the training process including:

[0139] Collect sample vital signs information from different application scenarios. The sample vital signs information includes at least sample electromyography signals. Preprocess the sample vital signs information to obtain the target vital signs information of the sample.

[0140] Obtain action tags that match the target's vital signs information, and construct a training sample set based on the target's vital signs information and action tags;

[0141] For each application scenario, the input dimensions and format of the input and output layers of the pre-trained large model are adjusted according to the application scenario.

[0142] The large model is fine-tuned based on the training sample set to obtain the target large model corresponding to the application scenario.

[0143] In some embodiments, the feature extraction module 703 is further configured to generate multiple preprocessing tasks based on the sample vital signs information and send the preprocessing tasks to the distributed nodes;

[0144] Receive the target vital signs information of the sample from the distributed nodes.

[0145] In some embodiments, the feature extraction module 703 is further configured to perform data augmentation on the sample vital sign information based on an adversarial network to obtain enhanced electromyographic signals, and to construct a training sample set based on the sample vital sign information and the enhanced vital sign information.

[0146] In some embodiments, the acquisition module 701 is further configured to acquire the user's initial vital signs information based on the target application scenario, including:

[0147] Based on the target application scenario, determine the set of sensors used to collect vital sign information;

[0148] Based on the target application scenario, determine the sampling area of ​​each type of sensor in the sensor set;

[0149] The initial vital signs information is obtained by receiving the signal collected in the sampling area from the sensor.

[0150] In some embodiments, the preprocessing module 701 is further configured to:

[0151] Encrypting any vital sign information using a homomorphic encryption algorithm yields encrypted vital sign information;

[0152] In response to any of the vital signs being the initial vital signs, model prediction is performed based on the encrypted vital signs.

[0153] In response to any of the vital signs being vital signs in the training sample set, model training is performed based on the encrypted vital signs.

[0154] The electromyography (EMG) signal processing device based on a large model provided in this application utilizes the powerful feature extraction capabilities of the large model to automatically learn deep features in EMG signals, breaking the limitations of relying on manual feature extraction. Compared with other deep learning algorithms, it has significant advantages in feature extraction, technical indicators, and application scenario expansion. For example, in feature extraction, it can extract more representative and discriminative features, resulting in a significant improvement in classification accuracy; in terms of technical indicators, both processing speed and classification accuracy far surpass traditional methods and other deep learning methods; in terms of application scenario expansion, it is not only suitable for the traditional medical rehabilitation field, but can also play an important role in virtual reality interaction, emergency rescue, and other fields.

[0155] Furthermore, the feature extraction and classification processes are organically integrated. Features extracted from a large model are directly input into the classifier for classification, greatly simplifying the processing flow and improving efficiency. Moreover, the rapid conversion from raw signal to classification result reduces error accumulation in intermediate steps, enhancing system stability and reliability.

[0156] To implement the above embodiments, this application also proposes an electronic device, including: a processor and a memory communicatively connected to the processor; the memory stores computer execution instructions; the processor executes the computer execution instructions stored in the memory to implement the method provided in the foregoing embodiments.

[0157] To implement the above embodiments, this application also proposes a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the methods provided in the foregoing embodiments.

[0158] To implement the above embodiments, this application also proposes a computer program product, including a computer program that, when executed by a processor, implements the methods provided in the foregoing embodiments.

[0159] The collection, storage, use, processing, transmission, provision, and application of user personal information involved in this application all comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0160] It should be noted that personal information collected from users should be used for legitimate and reasonable purposes and should not be shared or sold outside of these legitimate uses. Furthermore, such collection / sharing should only be conducted after receiving the user's informed consent, including but not limited to notifying the user to read the user agreement / user notice and sign an agreement / authorization that includes authorization of relevant user information before the user uses the function. In addition, any necessary steps must be taken to protect and safeguard access to such personal information data and ensure that others with access to personal information data comply with their privacy policies and procedures.

[0161] This application is intended to provide an implementation scheme for users to selectively prevent the use or access to their personal information data. Specifically, this application is intended to provide hardware and / or software to prevent or block access to such personal information data. Once personal information data is no longer needed, risks can be minimized by restricting data collection and deleting data. Furthermore, where applicable, such personal information is de-identified to protect user privacy.

[0162] In the foregoing descriptions of the embodiments, the terms "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0163] 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 technical features indicated. 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.

[0164] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing custom logic functions or processes, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as should be understood by those skilled in the art to which embodiments of this application pertain.

[0165] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Alternatively, the computer-readable medium may be paper or other suitable media on which the program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in a computer memory.

[0166] It should be understood that various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0167] Those skilled in the art will understand that all or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

[0168] Furthermore, the functional units in the various embodiments of this application can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.

[0169] The storage medium mentioned above can be a read-only memory, a disk, or an optical disk, etc. Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of this application.

Claims

1. A method for electromyographic signal processing based on a large model, characterized in that, The method includes: The target application scenario in which the user is located is determined, and the initial vital signs information of the user is collected based on the target application scenario. The initial vital signs information includes at least the initial electromyographic signal. The target application scenario includes: patient rehabilitation training scenario in medical rehabilitation, immersive experience scenario in virtual reality interaction, and auxiliary assistance scenario in emergency rescue. The initial vital signs information is preprocessed to obtain the target vital signs information; Based on the target large model corresponding to the target application scenario, feature extraction is performed on the target vital signs information to obtain the feature vector of the target vital signs information; The feature vector is classified and identified based on the target classification model to obtain the user's target action label. The construction of the target classification model includes: using a deep neural network classifier in medical rehabilitation and emergency rescue scenarios, and using a convolutional neural network and support vector machine classifier in virtual scene interaction scenarios. The training process of the target large model includes: Collect sample vital sign information from different application scenarios. The sample vital sign information includes at least sample electromyography signals. Preprocess the sample vital sign information to obtain the sample target vital sign information. The preprocessing of the sample vital sign information to obtain the sample target vital sign information includes: generating multiple preprocessing tasks based on the sample vital sign information and sending the preprocessing tasks to distributed nodes. Receive the target vital sign information of the sample from the distributed nodes; Obtain action tags that match the target's vital signs information, and construct a training sample set based on the target's vital signs information and action tags; For each application scenario, the input dimensions and format of the input and output layers of the pre-trained large model are adjusted according to the application scenario. The large model is fine-tuned based on the training sample set to obtain the target large model corresponding to the application scenario.

2. The method according to claim 1, characterized in that, The preprocessing of the initial vital sign information to obtain the target vital sign information includes: The initial vital signs information is denoised, filtered, and segmented to obtain the target vital signs information.

3. The method according to claim 2, characterized in that, The step of denoising, filtering, and segmenting the initial vital sign information to obtain the target vital sign information includes: Hardware denoising is performed on the initial vital signs information to obtain denoised vital signs information; Based on the frequency characteristics of the initial electromyography signal and the target application scenario, the filtering parameters of the filter are configured, and the noise-reduced vital sign information is filtered based on the configured filter to obtain filtered vital sign information. The energy value of the filtered vital signs information is determined, and the filtered vital signs information is segmented based on the energy value to obtain the target vital signs information.

4. The method according to claim 1, characterized in that, The method further includes: Based on adversarial networks, the sample vital signs information is augmented to obtain enhanced electromyographic signals, and a training sample set is constructed based on the sample vital signs information and the enhanced electromyographic signals.

5. The method according to any one of claims 1-4, characterized in that, The initial vital signs information of the user collected based on the target application scenario includes: Based on the target application scenario, determine the set of sensors used to collect vital sign information; Based on the target application scenario, determine the sampling area of ​​each type of sensor in the sensor set; The initial vital signs information is obtained by receiving the signal collected in the sampling area from the sensor.

6. The method according to any one of claims 1 or 4, characterized in that, The method further includes: Encrypting any vital sign information using a homomorphic encryption algorithm yields encrypted vital sign information; In response to any of the vital signs being the initial vital signs, model prediction is performed based on the encrypted vital signs. In response to any of the vital signs being vital signs in the training sample set, model training is performed based on the encrypted vital signs.

7. A large-scale electromyography (EMG) signal processing device, employing the large-scale EMG signal processing method according to any one of claims 1-6, characterized in that, include: The acquisition module is used to determine the target application scenario in which the user is located, and to acquire the user's initial vital signs information based on the target application scenario. The initial vital signs information includes at least initial electromyographic signals. The preprocessing module is used to preprocess the initial vital sign information to obtain the target vital sign information; The feature extraction module is used to extract features from the target vital signs information based on the target large model corresponding to the target application scenario, and obtain the feature vector of the initial vital signs information; The classification and recognition module is used to classify and recognize the feature vector based on the target classification model to obtain the user's target action label.

8. An electronic device, characterized in that, include: A processor, and a memory communicatively connected to the processor; The memory stores computer-executable instructions; the processor executes the computer-executable instructions stored in the memory to implement the method as described in any one of claims 1-6.

Images