Martial art training auxiliary system for teaching based on electroencephalogram control

Abstract

The invention discloses a teaching martial art training auxiliary system based on electroencephalogram control, and relates to the martial art training auxiliary field, and the system specifically comprises a collection module used for collecting an electroencephalogram signal and a posture signal of a user in real time; the analysis module is used for analyzing the electroencephalogram signal and the posture signal and generating a discharge control instruction; the execution module is used for converting the discharge control instruction into an electric signal to stimulate target muscles; and the communication module is used for realizing near-field direct connection communication between the acquisition module and the analysis module and remote communication between the analysis module and the execution module. According to the system, through a highly integrated and lightweight acquisition device, a neural network model for precise analysis and multi-task processing and a hybrid communication architecture for efficient communication and real-time feedback, the comfort, comprehensiveness and real-time performance of the martial art training auxiliary system are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the field of martial arts training assistance, and more particularly to a martial arts training assistance system for teaching based on brain electrical control. Background Art

[0002] Martial arts training assistance technology is an important field that combines sports science and human-computer interaction. It aims to improve movement standardization, training efficiency and safety through intelligent means. It is especially urgently needed in the teaching of high-difficulty movements and independent training.

[0003] Traditional martial arts training assistance devices are mostly designed for single movements and lack the ability to dynamically analyze and provide feedback on complex movements. For example, existing devices are unable to adjust assistance strategies in real time based on movement type, resulting in rigid training models that are difficult to adapt to the diverse needs of different martial arts routines. Furthermore, mechanical devices rely on preset programs and are unable to synchronize neural intent with body movements, limiting the coherence of dynamic training.

[0004] Furthermore, current devices primarily rely on physical support or visual cues, which can only correct limb form but fail to capture neurocognitive information such as EEG signals. Even when users imitate difficult movements with correct limb posture, incorrect force sequences can still lead to movement distortion or injury. The lack of real-time analysis of motor imagery signals in the brain makes it difficult to reinforce movement memory through neuromuscular synergy feedback, resulting in inefficient training.

[0005] Traditional EEG acquisition devices (such as wet electrode caps) require application of conductive paste and are bulky, severely limiting training flexibility. Mechanical assistive devices, on the other hand, take up a lot of space and are complex to adjust. Furthermore, existing systems often rely on wired communications or single wireless protocols, which suffer from high latency and weak anti-interference capabilities. This makes it difficult to support multi-terminal collaboration and real-time feedback, further restricting user experience and the expansion of training scenarios.

[0006] Therefore, how to design a martial arts training auxiliary system based on EEG control for teaching, which can effectively integrate multimodal signals, improve motion recognition accuracy, enhance equipment comfort and adaptability, and ensure efficient data processing and low-latency transmission is an urgent problem that technical personnel in this field need to solve. Summary of the Invention

[0007] In view of this, the present invention provides a martial arts training auxiliary system for teaching based on EEG control, which can integrate EEG and multimodal posture signals, combine dedicated neural networks with dynamic scheduling mechanisms, and realize accurate recognition and real-time feedback of movement intentions in martial arts training; and adopt a lightweight wearable design to improve wearing comfort and personalized adaptation capabilities; through Bluetooth 5.3, 5G-MQTT and edge computing terminals, an efficient hybrid communication architecture is constructed to ensure low latency, high real-time performance and security of the system.

[0008] In order to achieve the above object, the present invention adopts the following technical solutions:

[0009] A martial arts training auxiliary system for teaching based on brain electrical control, comprising:

[0010] Acquisition module, used to collect the user's EEG signals and posture signals in real time;

[0011] Analysis module, used to analyze EEG signals and posture signals and generate discharge control instructions;

[0012] an execution module, configured to convert the discharge control instruction into an electrical signal to stimulate the target muscle;

[0013] The communication module is used to realize near-field direct communication between the acquisition module and the analysis module, and remote communication between the analysis module and the execution module.

[0014] Furthermore, the acquisition module includes: a headband-type EEG cap body, integrated with impedance dry electrodes, ear clip electrodes and an EEG acquisition board;

[0015] The impedance dry electrodes and ear clip electrodes are used to collect EEG signals;

[0016] The EEG acquisition board has a built-in amplifier and ADC converter. The amplifier is used to amplify weak EEG signals, and the ADC converter is used to convert analog signals into digital signals.

[0017] Furthermore, the acquisition module also includes: a posture sensor, including a nine-axis gyroscope and a gel layer; the nine-axis gyroscope is composed of three magnetometers, three accelerometers and three gyroscopes, and is used to detect limb acceleration, angular velocity and magnetic field data.

[0018] Furthermore, the parsing module includes:

[0019] A model selection unit for scheduling a dedicated neural network model type based on attitude sensor data;

[0020] The dedicated neural network model unit is used to analyze the upper limb movements and lower limb movements respectively through the upper limb movement dedicated neural network model and the lower limb movement dedicated neural network model and generate discharge control instructions.

[0021] Furthermore, the working logic of the model selection unit includes:

[0022] When the posture sensor data deployed on the upper limbs meets the following conditions, the upper limb action-specific neural network model is triggered: Peak acceleration ≥ 5m / s 2, and the average angular velocity change rate within the 0.5 second window is ≥50° / s; the magnetic field direction offset is ≤15°, and the spatial trajectory matches the preset upper limb movement pattern ≥80%;

[0023] When the posture sensor data deployed on the lower limbs meets the following conditions, the lower limb action-specific neural network model is triggered: Peak acceleration ≥ 3m / s 2 , and the angular velocity periodic fluctuation amplitude within a 1-second window is ≥30° / s; the magnetic field direction offset is ≥30°, and the spatial trajectory matches the preset lower limb movement pattern ≥75%;

[0024] When the above conditions are met at the same time, the parallel parsing mode is activated, and the upper limb data stream is assigned to the upper limb movement-specific neural network model according to the sensor ID, and the lower limb data stream is assigned to the lower limb movement-specific neural network model.

[0025] Furthermore, the upper limb movement-specific neural network model includes:

[0026] Temporal feature extraction subunit: Two convolutional layers with a kernel size of 1×3, a stride of 1, and an activation function of exponential linear unit are used, followed by a batch normalization layer and a maximum pooling layer with a pooling window size of 1×2, a stride of 2, and effective padding.

[0027] Spatial feature extraction subunit: Two convolutional layers with a kernel size of 3×3, a stride of 1, and an activation function of exponential linear unit are used, followed by a batch normalization layer and a maximum pooling layer with a pooling window size of 2×2, a stride of 2, and equal padding.

[0028] Classification subunit: After mapping the spatiotemporal fusion features through 128 and 64 neuron fully connected layers, the softmax function is used to output the three-classification results of the forearm, upper arm and shoulder force patterns;

[0029] Feature optimization mechanism: A local attention layer is embedded in the temporal feature extraction subunit to enhance the short-term feature association between the wrist and elbow joints in the EEG signals.

[0030] Furthermore, the lower limb movement-specific neural network model includes:

[0031] Temporal feature extraction subunit: Two convolutional layers with a kernel size of 3×1, a stride of 1, and an activation function of rectified linear unit are used, followed by a batch normalization layer and a maximum pooling layer with a pooling window size of 4×1, a stride of 4, and equal padding.

[0032] Spatial feature extraction subunit: Two convolutional layers with a kernel size of 5×5, a stride of 1, and a rectified linear unit activation function are used, followed by a batch normalization layer and a maximum pooling layer with a pooling window size of 3×3, a stride of 2, and effective padding.

[0033] Classification subunit: After mapping the spatiotemporal fusion features through 64 and 32 neuron fully connected layers, the softmax function is used to output the binary classification results of the knee and hip joint force patterns;

[0034] Feature optimization mechanism: A dynamic pooling layer is introduced in the spatial feature extraction subunit, and its window size is adaptively adjusted according to the gait cycle duration detected by the posture sensor.

[0035] Furthermore, the analysis module further includes:

[0036] The voice command interaction unit is used to receive and recognize the user's voice signals and control the working status of the analysis module, including starting and stopping the analysis device, clearing historical data, triggering the debugging mode, and completing the recalibration of personalized parameters.

[0037] Furthermore, the execution module includes:

[0038] The discharge control board integrates a DAC converter, a programmable constant current output terminal, and an overcurrent protector, and is used to convert the discharge control instructions sent by the analysis module into adjustable electrical signals;

[0039] The low-frequency therapy electrode sheet includes a flexible conductive layer and a shielded wire interface, and is used to stimulate the user's target muscles according to the electrical signal.

[0040] Furthermore, the communication module includes:

[0041] Bluetooth 5.3 unit, used to transmit EEG signals and posture signals from the acquisition module to the analysis module;

[0042] The 5G-MQTT protocol unit is used to transmit the discharge instructions in the parsing module to the execution module.

[0043] It can be seen from the above technical solution that compared with the prior art, the technical solution of the present invention has the following advantages:

[0044] Beneficial effects:

[0045] 1. The system integrates an EEG signal acquisition module and a nine-axis posture sensor to achieve two-dimensional real-time monitoring of the user's neural activity and limb movements. Combined with dedicated neural network models for the upper and lower limbs, and using a spatiotemporal feature extraction and classification optimization mechanism, it significantly improves the accuracy and response speed of action intention recognition.

[0046] 2. The use of a headband-style EEG cap, a gel-layer attached posture sensor, and flexible low-frequency therapy electrodes significantly reduces the size and weight of the device, minimizing interference with training movements. At the same time, through the voice command interaction unit and debugging mode, it supports rapid calibration of user-customized parameters and flexible control of device status.

[0047] 3. A communication solution combining Bluetooth 5.3 near-field direct connection and 5G-MQTT protocol is adopted, taking into account low power consumption, high real-time performance and wide coverage requirements; EEG-gesture signals are locally analyzed through the edge computing terminal (JETSON XAVIERNX), reducing cloud dependence and ensuring low latency in data processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are merely embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0049] Figure 1 A framework diagram of a martial arts training auxiliary system for teaching based on EEG control provided by an embodiment of the present invention;

[0050] Figure 2 A schematic diagram of the main structure of a headband-type EEG cap provided in an embodiment of the present invention;

[0051] Figure 3 A schematic diagram of the structure of a gesture sensor provided in an embodiment of the present invention;

[0052] Figure 4 A schematic diagram of the execution module structure provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0053] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0054] like Figure 1 As shown, this embodiment provides a martial arts training auxiliary system for teaching based on brain electrical control, including:

[0055] Acquisition module, used to collect the user's EEG signals and posture signals in real time;

[0056] Analysis module, used to analyze EEG signals and posture signals and generate discharge control instructions;

[0057] an execution module, configured to convert the discharge control instruction into an electrical signal to stimulate the target muscle;

[0058] The communication module is used to realize near-field direct communication between the acquisition module and the analysis module, and remote communication between the analysis module and the execution module.

[0059] The system significantly improves the comfort, comprehensiveness and real-time performance of the martial arts training assistance system through highly integrated and lightweight acquisition devices, neural network models with precise analysis and multi-tasking processing, and a hybrid communication architecture with efficient communication and real-time feedback. Users can train in a safer and more efficient environment, accelerate the formation of conditioned reflexes, and thus improve training results.

[0060] The following is a further detailed description of each module in the above system;

[0061] like Figure 2 As shown, the acquisition module specifically includes a headband-type EEG cap body 1, which integrates an impedance dry electrode 1-1, an ear clip electrode 1-2 and an EEG acquisition board 1-3;

[0062] The impedance dry electrode 1-1 and the ear clip electrode 1-2 are used to collect EEG signals. Specifically, the ear clip electrode 1-2 is used to provide a reference potential for EEG signal collection. This reference potential eliminates common-mode interference through differential amplification and collects EEG signals synchronously with the impedance dry electrode 1-1 to improve the signal-to-noise ratio.

[0063] The EEG acquisition boards 1-3 have built-in amplifiers and ADC converters. The amplifiers are used to amplify weak EEG signals, and the ADC converters are used to convert analog signals into digital signals.

[0064] like Figure 3 As shown, the acquisition module also includes a posture sensor 2, which includes a nine-axis gyroscope 2-1 and a gel layer 2-2; the nine-axis gyroscope is composed of three magnetometers, three accelerometers and three gyroscopes, and is used to detect limb acceleration, angular velocity and magnetic field data;

[0065] Specifically, the gel layer 2-2 is made of biocompatible material to improve the stability of the contact between the posture sensor and the skin and the signal acquisition accuracy;

[0066] This acquisition module achieves reliable dual-modal acquisition of EEG-posture signals through high-precision electrode design, low-noise signal chain and dynamic posture perception, providing high-quality input for subsequent analysis.

[0067] In this embodiment, the parsing module is used to parse EEG signals and posture signals and generate discharge control instructions; specifically, it includes:

[0068] A model selection unit for scheduling a dedicated neural network model type based on attitude sensor data;

[0069] The dedicated neural network model unit is used to analyze the upper limb movements and lower limb movements respectively through the upper limb movement dedicated neural network model and the lower limb movement dedicated neural network model and generate discharge control instructions.

[0070] Furthermore, the working logic of the model selection unit includes:

[0071] When the posture sensor data deployed on the upper limbs meets the following conditions, the upper limb action-specific neural network model is triggered: Peak acceleration ≥ 5m / s 2 , and the average angular velocity change rate within the 0.5 second window is ≥50° / s; the magnetic field direction offset is ≤15°, and the spatial trajectory matches the preset upper limb movement pattern ≥80%;

[0072] When the posture sensor data deployed on the lower limbs meets the following conditions, the lower limb action-specific neural network model is triggered: Peak acceleration ≥ 3m / s 2 , and the angular velocity periodic fluctuation amplitude within a 1-second window is ≥30° / s; the magnetic field direction offset is ≥30°, and the spatial trajectory matches the preset lower limb movement pattern ≥75%;

[0073] When the above conditions are met at the same time, the parallel parsing mode is activated, and the upper limb data stream is assigned to the upper limb movement-specific neural network model according to the sensor ID, and the lower limb data stream is assigned to the lower limb movement-specific neural network model.

[0074] Furthermore, the upper limb movement-specific neural network model includes:

[0075] Temporal feature extraction subunit: Two convolutional layers with a kernel size of 1×3, a stride of 1, and an activation function of exponential linear unit are used, followed by a batch normalization layer and a maximum pooling layer with a pooling window size of 1×2, a stride of 2, and effective padding.

[0076] Spatial feature extraction subunit: Two convolutional layers with a kernel size of 3×3, a stride of 1, and an activation function of exponential linear unit are used, followed by a batch normalization layer and a maximum pooling layer with a pooling window size of 2×2, a stride of 2, and equal padding.

[0077] Classification subunit: After mapping the spatiotemporal fusion features through 128 and 64 neuron fully connected layers, the softmax function is used to output the three-classification results of the forearm, upper arm and shoulder force patterns;

[0078] Feature optimization mechanism: A local attention layer is embedded in the temporal feature extraction subunit to enhance the short-term feature association between the wrist and elbow joints in the EEG signals.

[0079] Furthermore, the lower limb movement-specific neural network model includes:

[0080] Temporal feature extraction subunit: Two convolutional layers with a kernel size of 3×1, a stride of 1, and an activation function of rectified linear unit are used, followed by a batch normalization layer and a maximum pooling layer with a pooling window size of 4×1, a stride of 4, and equal padding.

[0081] Spatial feature extraction subunit: Two convolutional layers with a kernel size of 5×5, a stride of 1, and a rectified linear unit activation function are used, followed by a batch normalization layer and a maximum pooling layer with a pooling window size of 3×3, a stride of 2, and effective padding.

[0082] Classification subunit: After mapping the spatiotemporal fusion features through 64 and 32 neuron fully connected layers, the softmax function is used to output the binary classification results of the knee and hip joint force patterns;

[0083] Feature optimization mechanism: A dynamic pooling layer is introduced in the spatial feature extraction subunit, and its window size is adaptively adjusted according to the gait cycle duration detected by the posture sensor.

[0084] Specifically, the differentiated designs of the upper limb movement-specific neural network model and the lower limb movement-specific neural network model in this embodiment are shown in Table 1 below:

[0085] Table 1

[0086]

[0087]

[0088] Through the above-mentioned differentiated design, the model can more accurately adapt to the physical characteristics and signal features of upper and lower limb movements in martial arts training, ensuring the accuracy, real-time and security of EEG and posture data analysis.

[0089] Furthermore, the parsing module also includes:

[0090] The voice command interaction unit is used to receive and recognize the user's voice signals and control the working status of the analysis module, including starting and stopping the analysis device, clearing historical data, triggering the debugging mode, and completing the recalibration of personalized parameters.

[0091] The analysis module achieves accurate analysis of action intentions and generation of personalized instructions through multi-model dynamic scheduling, spatiotemporal feature optimization and voice interaction, ensuring the real-time and adaptability of training guidance.

[0092] In this embodiment, the execution module is used to convert the discharge control instruction into an electrical signal to stimulate the target muscle; Figure 4 As shown, specifically including:

[0093] The discharge control board 3-1 integrates a DAC converter, a programmable constant current output terminal, and an overcurrent protector, and is used to convert the discharge control instructions sent by the analysis module into adjustable electrical signals. Specifically, the DAC converter converts digital instructions into analog electrical signals; the programmable constant current output terminal controls the strength and frequency of the electrical signal; and the overcurrent protector monitors and limits abnormal current output in real time.

[0094] The low-frequency therapy electrode sheet 3-2 is used to stimulate the user's target muscles according to the electrical signal; it includes a flexible conductive layer and a shielded wire interface. The flexible conductive layer is used to adhere to the skin and conduct electrical signals, and the detachable shielded wire interface realizes a stable connection between the electrode sheet and the discharge control board.

[0095] The execution module achieves effective stimulation of target muscles through precise electrical signal output and comfortable electrode design, helping users better master the force points and methods.

[0096] In this embodiment, the communication module specifically includes: a Bluetooth 5.3 unit, which is used to transmit the EEG signals and posture signals in the acquisition module to the analysis module, and utilizes its low power consumption and high real-time characteristics to ensure the rapid transmission of EEG signals and posture signals; a 5G-MQTT protocol unit, which is used to transmit the discharge instructions in the analysis module to the execution module, and utilizes the high bandwidth and low latency characteristics of the 5G network to enhance the real-time and scalability of the MQTT protocol to meet the needs of multi-terminal collaborative control.

[0097] Furthermore, this section further explains the working process of the martial arts training assistance system by combining the Tai Chi "Cloud Hands" action training scene:

[0098] 1) Wearing and initializing the device;

[0099] The user wears a headband-style EEG cap and a posture sensor to ensure stable contact between the electrodes and the skin; after the system is started, the acquisition module completes self-test, the communication module establishes a connection between Bluetooth 5.3 and the analysis module, and the 5G-MQTT protocol and the execution module are synchronized and ready.

[0100] 2) Signal acquisition and transmission;

[0101] For EEG signal acquisition, ear clip electrodes provide a reference potential, and impedance dry electrodes synchronously acquire EEG signals. Interference is eliminated through differential amplification, and the signals are amplified and converted into digital signals by the EEG acquisition board. For posture signal acquisition, a nine-axis gyroscope detects upper limb acceleration, angular velocity change rate, and magnetic field offset in real time. The data is optimized by the gel layer and then output. For data transmission, Bluetooth 5.3 transmits EEG and posture signals to the analysis module.

[0102] 3) Action analysis and instruction generation;

[0103] The model selection unit detects that the upper limb movement conditions are met and triggers the upper limb dedicated neural network model.

[0104] Feature extraction: temporal feature extraction, 1×3 convolution kernel extracts short-term features of the wrist and elbow joints, and the local attention layer strengthens the key signal association; spatial feature extraction, 3×3 convolution kernel analyzes the coordination mode of the shoulder and small muscle groups, and maximum pooling compresses redundant information; classification output, the spatiotemporal fusion features are mapped through the fully connected layer, and Softmax outputs the "forearm-dominated force" classification result to generate discharge control instructions.

[0105] 4) Electrical signal execution and feedback;

[0106] The discharge command is sent to the execution module via the 5G-MQTT protocol. Furthermore, the DAC converter on the discharge control board converts the command into an analog signal, and the constant current output is adjusted to a safe intensity. Low-frequency therapy electrodes are then used to stimulate the target forearm muscles, guiding the correct force application.

[0107] Real-time monitoring: the overcurrent protector continuously monitors the current and immediately cuts off the output when an abnormality occurs to ensure user safety.

[0108] Step 5: Interaction and optimization;

[0109] Users control the status of the analysis module or trigger parameter calibration through voice commands; in addition, the system dynamically optimizes model weights based on training results to improve the accuracy of subsequent action analysis.

[0110] Through the above process, users can accurately perceive the forearm force pattern during Tai Chi "Cloud Hands" movement training, accelerate the formation of conditioned reflexes, and at the same time the system ensures the safety and scientific nature of the training process.

[0111] This embodiment builds a precise, real-time martial arts training assistance system by integrating dual-modal acquisition of EEG signals and posture data, differentiated neural network analysis, and efficient communication architecture. With the help of high-precision sensors, dynamic model scheduling, and safe electrical stimulation technology, the system can accurately identify the movement characteristics of the upper and lower limbs and guide the target muscles to exert force in real time. It can provide scientific and personalized technical solutions for intelligent martial arts teaching, and has significant practical value and promotion potential.

[0112] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. References to the same or similar parts between the various embodiments are sufficient. For the systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple. For relevant parts, refer to the method description.

[0113] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not limited to the embodiments shown herein but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A martial arts training auxiliary system for teaching based on brain electrical control, characterized in that: include: Acquisition module, used to collect the user's EEG signals and posture signals in real time; Analysis module, used to analyze EEG signals and posture signals and generate discharge control instructions; an execution module, configured to convert the discharge control instruction into an electrical signal to stimulate the target muscle; The communication module is used to realize near-field direct communication between the acquisition module and the analysis module, and remote communication between the analysis module and the execution module.

2. The brain-electric control-based martial arts training auxiliary system for teaching according to claim 1 is characterized in that: The acquisition module includes: a headband-type EEG cap body, integrated with impedance dry electrodes, ear clip electrodes and an EEG acquisition board; The impedance dry electrodes and ear clip electrodes are used to collect EEG signals; The EEG acquisition board has a built-in amplifier and ADC converter. The amplifier is used to amplify weak EEG signals, and the ADC converter is used to convert analog signals into digital signals.

3. The brain-electric control-based martial arts training auxiliary system for teaching according to claim 1 is characterized in that: The acquisition module also includes: a posture sensor, which includes a nine-axis gyroscope and a gel layer; the nine-axis gyroscope is composed of three magnetometers, three accelerometers and three gyroscopes, and is used to detect limb acceleration, angular velocity and magnetic field data.

4. The brain-electric control-based martial arts training auxiliary system for teaching according to claim 1 is characterized in that: The parsing module includes: A model selection unit for scheduling a dedicated neural network model type based on attitude sensor data; The dedicated neural network model unit is used to analyze the upper limb movements and lower limb movements respectively through the upper limb movement dedicated neural network model and the lower limb movement dedicated neural network model and generate discharge control instructions.

5. The brain-electric control-based martial arts training auxiliary system for teaching according to claim 1 is characterized in that: The working logic of the model selection unit includes: When the posture sensor data deployed on the upper limbs meets the following conditions, the upper limb action-specific neural network model is triggered: Peak acceleration ≥ 5m / s 2 , and the average angular velocity change rate within the 0.5 second window is ≥50° / s; the magnetic field direction offset is ≤15°, and the spatial trajectory matches the preset upper limb movement pattern ≥80%; When the posture sensor data deployed on the lower limbs meets the following conditions, the lower limb action-specific neural network model is triggered: Peak acceleration ≥ 3m / s 2 , and the angular velocity periodic fluctuation amplitude within a 1-second window is ≥30° / s; the magnetic field direction offset is ≥30°, and the spatial trajectory matches the preset lower limb movement pattern ≥75%; When the above conditions are met at the same time, the parallel parsing mode is activated, and the upper limb data stream is assigned to the upper limb movement-specific neural network model according to the sensor ID, and the lower limb data stream is assigned to the lower limb movement-specific neural network model.

6. The brain-electric control-based martial arts training auxiliary system for teaching according to claim 1 is characterized in that: The upper limb movement-specific neural network model includes: Temporal feature extraction subunit: Two convolutional layers with a kernel size of 1×3, a stride of 1, and an activation function of exponential linear unit are used, followed by a batch normalization layer and a maximum pooling layer with a pooling window size of 1×2, a stride of 2, and effective padding. Spatial feature extraction subunit: Two convolutional layers with a kernel size of 3×3, a stride of 1, and an activation function of exponential linear unit are used, followed by a batch normalization layer and a maximum pooling layer with a pooling window size of 2×2, a stride of 2, and equal padding. Classification subunit: After mapping the spatiotemporal fusion features through 128 and 64 neuron fully connected layers, the softmax function is used to output the three-classification results of the forearm, upper arm and shoulder force patterns; Feature optimization mechanism: A local attention layer is embedded in the temporal feature extraction subunit to enhance the short-term feature association between the wrist and elbow joints in the EEG signals.

7. The brain-electric control-based martial arts training auxiliary system for teaching according to claim 1 is characterized in that: The lower limb movement-specific neural network model includes: Temporal feature extraction subunit: Two convolutional layers with a kernel size of 3×1, a stride of 1, and an activation function of rectified linear unit are used, followed by a batch normalization layer and a maximum pooling layer with a pooling window size of 4×1, a stride of 4, and equal padding. Spatial feature extraction subunit: Two convolutional layers with a kernel size of 5×5, a stride of 1, and a rectified linear unit activation function are used, followed by a batch normalization layer and a maximum pooling layer with a pooling window size of 3×3, a stride of 2, and effective padding. Classification subunit: After mapping the spatiotemporal fusion features through 64 and 32 neuron fully connected layers, the softmax function is used to output the binary classification results of the knee and hip joint force patterns; Feature optimization mechanism: A dynamic pooling layer is introduced in the spatial feature extraction subunit, and its window size is adaptively adjusted according to the gait cycle duration detected by the posture sensor.

8. The brain-electric control-based martial arts training auxiliary system for teaching according to claim 1 is characterized in that: The parsing module also includes: The voice command interaction unit is used to receive and recognize the user's voice signals and control the working status of the analysis module, including starting and stopping the analysis device, clearing historical data, triggering the debugging mode, and completing the recalibration of personalized parameters.

9. The brain-electric control-based martial arts training auxiliary system for teaching according to claim 1 is characterized in that: The execution module includes: The discharge control board integrates a DAC converter, a programmable constant current output terminal and an overcurrent protector, and is used to convert the discharge control instructions sent by the analysis module into adjustable electrical signals; The low-frequency therapy electrode sheet includes a flexible conductive layer and a shielded wire interface, and is used to stimulate the user's target muscles according to the electrical signal.

10. The brain-electric control-based martial arts training auxiliary system for teaching according to claim 1 is characterized in that: The communication module includes: Bluetooth 5.3 unit, used to transmit EEG signals and posture signals from the acquisition module to the analysis module; The 5G-MQTT protocol unit is used to transmit the discharge instructions in the parsing module to the execution module.

Images