An all-weather fully autonomous brain-computer interface training system

Abstract

The application provides a kind of all-weather full autonomous brain-computer interface training system, the system is by distributing the weight of behavior instruction and brain control instruction, namely, completely behavior instruction judges the consistency of behavior action and behavior stimulation, phase transition to judge based on brain control instruction completely, and the realization of full autonomous smooth control with higher degree of automation provides convenience for long-term brain science, neurorehabilitation training, brain-computer interface control algorithm and the like research.

Description

Technical Field

[0001] This invention belongs to the field of brain-computer interfaces, specifically relating to an all-weather, fully autonomous brain-computer interface training system. Background Technology

[0002] Brain-computer interface (BCI) technology aims to establish a direct communication pathway between the brain and external devices, and has significant application prospects in areas such as neural function repair, motor assistance, and cognitive enhancement. One of the core components of a BCI system is training the user (including humans or animals) to control external devices through neural activity.

[0003] Most existing brain-computer interface training systems have the following limitations: (1) Limited training time: Traditional brain-computer interface systems rely on wired connections or battery power, which cannot support continuous training for several days or weeks, making it difficult to study long-term changes in neuroplasticity.

[0004] (2) The training process is not autonomous: The training process usually relies on real-time supervision and frequent intervention by the experimenters, such as adjusting parameters, replacing batteries, restarting tasks, etc. The low degree of automation leads to low training efficiency and poor training consistency, making it difficult to stably reproduce experimental results.

[0005] (3) Abrupt transition in training: Traditional training methods often directly conduct brain control training, which is separated from behavioral training. There is a lack of a smooth, natural, and guiding transition mechanism, resulting in a steep learning curve and low training efficiency.

[0006] (4) Insufficient generalization of training paradigms: Many systems restrict the free movement of experimental animals, and the training state is very different from the natural activity state, which significantly affects the characteristics of neural signals, making it difficult to directly extrapolate the brain control effect of training to the scenario of natural activity state.

[0007] Although patent applications CN113180674A and CN113995424A disclose wireless neural recording or wireless power supply schemes, they all focus on signal recording and power supply itself, without involving a complete and autonomous brain-computer interface training closed-loop system, especially lacking fully autonomous and intelligent training progress control and "behavior-brain control" progressive transition algorithm.

[0008] Therefore, there is an urgent need for a brain-computer interface training system that can support free movement, fully automated operation, and intelligently guide experimental animals to naturally transition from behavioral control to neural control. Summary of the Invention

[0009] This invention provides an all-weather, fully autonomous brain-computer interface training system that enables a smooth, fully autonomous transition from behavioral control to brain control.

[0010] This invention provides an all-weather, fully autonomous brain-computer interface training system, comprising: A neural signal recording device used to collect neural electrical signals and behavioral signals from subjects; An external device control device is used to provide behavioral stimuli to the subject, and then collect the behavioral instructions corresponding to the behavioral actions generated by the subject after receiving the behavioral stimuli. It is also used to provide reward feedback to the subject. A neural signal decoding device, wirelessly connected to a neural signal recording device, is used to decode received neural electrical signals and behavioral signals into brain control commands that can predict the subject's behavioral actions and the intensity of the behavioral actions. The training progress control device is connected to both an external device control device and a neural signal decoding device. It sequentially sets the training process into four stages: a fully behavioral control stage, a behavior-dominated stage, a brain-dominated stage, and a fully brain-controlled stage. In the fully behavioral control stage, it determines whether the subject's behavior matches the behavioral stimulus based on behavioral instructions. In the behavior-dominated stage, it performs a weighted fusion of behavioral instructions and brain-controlled instructions, and determines whether the subject's behavior matches the behavioral stimulus based on the weighted fusion result, with behavioral instructions having a higher weight. In the brain-dominated stage, it performs a weighted fusion of behavioral instructions and brain-controlled instructions, and determines whether the subject's behavior matches the behavioral stimulus based on the weighted fusion result, with brain-controlled instructions having a higher weight. In the fully brain-controlled stage, it determines whether the subject's behavior matches the behavioral stimulus based on brain-controlled instructions. In all four stages—behavior-dominated, brain-dominated, and fully brain-controlled—if the subject's behavior matches the behavioral stimulus, a reward feedback instruction is sent to the external device control device.

[0011] Preferably, in the full behavior control stage, based on the received behavior instructions, the neural electrical signals and behavior signals before and after the behavior action are generated are judged, and the initial decoder is trained based on the neural electrical signals and behavior signals before and after the behavior action is generated to obtain a neural signal decoding device.

[0012] Preferably, during the full behavior control phase, after the intensity of the behavior reaches a set threshold, the external device control device provides reward feedback to the subject.

[0013] Preferably, during the training process, a valve threshold is set, which is a success rate threshold, a decoding confidence threshold, or a training duration threshold. When the value exceeds the set valve threshold, the current stage proceeds to the next stage.

[0014] Preferably, in the behavior-dominated phase and the mind-controlled phase, the weights of the behavior commands and the mind-controlled commands are adjusted in real time, and the method for adjusting the weights includes: The behavior-driven or mind-controlled stage is divided into multiple training blocks, each of which includes multiple trials. When the short-term task success rate of the training block is greater than the set success rate threshold, and the average intensity of the decoded behavior action in each training block is greater than the set intensity threshold, the weight of the mind-controlled instruction is increased in the next training block. If the short-term task success rate of a training block is less than the set lower limit of the success rate threshold, then in the next training block, the weight of the brain control command is reduced or paused.

[0015] Preferably, it also includes a host computer training management platform, which is connected to the neural signal recording device, the external device control device, the neural signal decoding device, and the training progress control device, respectively. The host computer training management platform is used to save and record neural signals and behavioral signals, to update the driver firmware of the external device control device in real time to configure training tasks, to update the adjustment algorithm of the training progress control device in real time to configure the training process, and to display and save behavioral instructions and control instructions at each stage of the training process.

[0016] Preferably, it also includes a wireless power supply system, which includes a radio magnetic field transmitting unit, a radio frequency transmitting coil, multiple LC resonant circuits, a radio frequency receiving device, an AC-DC conversion unit, and a power management unit; The radio magnetic field transmitting unit is used to generate a high-frequency alternating current to drive the radio frequency transmitting coil to excite the source electromagnetic field. A test space is set up in each LC resonant circuit, and the subject moves freely in the test space. Each LC resonant circuit is set vertically on the radio frequency transmitting coil at a set interval to form a high-frequency alternating magnetic field from the source electromagnetic field. A radio frequency receiving device is implanted in the subject to obtain high-frequency alternating current through a high-frequency alternating magnetic field obtained by a corresponding LC resonant circuit. The high-frequency alternating current is then transmitted to an AC-DC conversion unit, which converts the high-frequency alternating current into a stable voltage and supplies it to a power management unit. The power management unit then supplies power to the neural signal recording device.

[0017] Preferably, the training progress control device includes a training phase management module, a neural control ratio adjustment module, and a feedback adjustment module; The training phase management module is used to set the training task structure and parameter settings for different training phases; The neural control ratio adjustment module is used to update the weight ratio of neural control and behavioral control in real time according to the valve threshold. The feedback adjustment module is used to update the current training phase in real time based on the valve threshold.

[0018] Preferably, the neural signal recording device includes a head-mounted wireless neural signal recording subsystem and an external wireless signal relay subsystem; The head-mounted wireless neural signal recording subsystem is used to acquire the subject's neural electrical signals and behavioral signals, and transmit the neural electrical signals and behavioral signals to the wireless signal relay subsystem; The wireless signal relay subsystem is used to transmit received neural electrical signals and behavioral signals to the neural signal decoding device.

[0019] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention utilizes a training progress control device to allocate weights between behavioral instructions and brain control instructions. Specifically, it judges the consistency between behavioral actions and behavioral stimuli based entirely on behavioral instructions, and then transitions in stages to judgment based entirely on brain control instructions. This highly automated, fully autonomous, and smooth control provides convenience for research in long-term neuroscience, neurorehabilitation training, and brain-computer interface control algorithms. Attached Figure Description

[0020] Figure 1 The overall system architecture diagram provided for a specific embodiment of the present invention; Figure 2 A schematic block diagram of a wireless power supply system provided for a specific embodiment of the present invention; Figure 3 This is a diagram showing the recorded neural signals of the present invention; Figure 4 This is a circuit block diagram of the neural signal recording device of the present invention; Figure 5 This is a schematic diagram of the large-area wireless power supply system of the present invention; Figure 6 This is an overall block diagram of the wireless power supply system of the present invention; Figure 7 This is a flowchart of brain-controlled training in a specific embodiment of the present invention. Detailed Implementation

[0021] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be noted that the embodiments described below are intended to facilitate the understanding of the present invention and do not constitute any limitation thereof.

[0022] This invention provides a 24 / 7 fully autonomous brain-computer interface training system, such as... Figure 1 As shown, it includes a neural signal recording device, an external device control device, a neural signal decoding device, and a neural signal decoding device.

[0023] The neural signal recording device provided in the specific embodiments of the present invention is used to collect the neural electrical signals and behavioral signals of the subjects.

[0024] In one specific embodiment, the neural signal recording device provided in this embodiment includes a head-mounted wireless neural signal recording subsystem and an external wireless signal relay subsystem. The head-mounted wireless neural signal recording subsystem is used to acquire the neural electrical signals and behavioral signals of the subject and send the neural electrical signals and behavioral signals to the wireless signal relay subsystem. The wireless signal relay subsystem is used to transmit the received neural electrical signals and behavioral signals to the neural signal decoding device.

[0025] The radio frequency receiving device provided in this embodiment of the invention, namely a subcutaneous implantable wireless power acquisition device and a head-mounted wireless neural signal recording subsystem, is connected by a flexible cable. The flexible cable is a wireless power supply connection with a serpentine routing. The radio frequency receiving device is placed in the subcutaneous cavity above the scapula of a mouse. It uses a flexible circuit board and flexible packaging, which can deform slightly with the natural movement of the mouse.

[0026] In one specific embodiment, the head-mounted wireless neural signal recording subsystem provided in this embodiment is a microPCB. This microPCB is fixed to the mouse skull via a neural interface and a power interface, and is electrically connected to the neural microelectrodes implanted in the mouse's skull and the fully implanted radio frequency receiver, respectively. The overall weight of the microPCB is controlled to be less than 2g, making it suitable for the weight and size of small animals such as mice. The microPCB integrates the micro battery charging device, the neural signal acquisition device, and the wireless communication device. The micro battery charging device includes a 10mAh capacity micro lithium battery fixed above the microPCB, responsible for providing necessary energy buffering.

[0027] Specifically, such as Figure 3 and Figure 4As shown, the neural signal acquisition device provided in this specific embodiment of the invention uses an integrated chip, the Intan RHD2000 series RHd2132 chip, which can record up to 32 channels. It is responsible for amplifying and converting the weak analog bioelectrical signals acquired by the implanted neural microelectrodes into digital signals. A passive low-pass filter is also integrated at the front end of the neural signal acquisition chip to filter out high-frequency noise coupled from the 6.78MHz alternating magnetic field in the front-end analog signal. The neural signal acquisition device also integrates a six-axis inertial sensing unit to acquire motion acceleration signals. The wireless communication device uses a Nordic nRF series low-power Bluetooth chip, which integrates a central processing unit to package and send the signals acquired by the neural signal acquisition device to a host computer using the ESB 2.4GHz protocol. The host computer uses a Nordic nRF series development board as a 2.4GHz data relay station to send data to the PC. This invention uses a multi-mode neural signal acquisition setup to achieve low-power acquisition of neural and behavioral data of various modalities. This includes switching the sampling rate and filtering bandwidth of the neural signal acquisition chip to achieve real-time recording of multi-channel neuronal local field potential signals, raw neural signal recording at 20kHz sampling, and triaxial accelerometer signals. It also enables online processing and compression of the recorded data, and can acquire multi-channel neuronal spike signals and record ESA (Entire Spike Activity) signals. All acquired neural and behavioral signals are displayed in real-time on a graphical user interface (GUI) via a host computer and saved in structured data files for subsequent offline analysis and processing.

[0028] The external device control device provided in a specific embodiment of the present invention is used to provide behavioral stimuli to the subject. The behavioral stimuli include sound and light stimuli under PWM modulation. By setting multiple stimulation sources in spatial locations, they can be combined into arbitrarily complex stimulation patterns. Then, the device collects the behavioral instructions corresponding to the behavioral actions generated by the subject after receiving the behavioral stimuli. The device is also used to provide reward feedback to the subject.

[0029] The neural signal decoding device provided in the specific embodiment of the present invention is wirelessly connected to the neural signal recording device, and is used to decode the received neural electrical signals and behavioral signals into brain control commands that can predict the subject's behavior and the intensity of the behavior.

[0030] The training progress control device provided in this specific embodiment of the invention is connected to an external device control device and a neural signal decoding device, respectively, for setting the training process sequentially into a fully behavioral control stage, a behavior-led stage, a brain-controlled led stage, and a fully brain-controlled stage. In the fully behavioral control stage, the device judges whether the subject's behavior is consistent with the behavioral stimulus based on behavioral instructions. In the behavior-led stage, the behavioral instructions and brain-controlled instructions are weighted and fused, and the consistency between the subject's behavior and the behavioral stimulus is judged based on the weighted fusion result, wherein the behavioral instructions are given a higher weight. In the brain-controlled led stage, the behavioral instructions and brain-controlled instructions are weighted and fused, and the consistency between the subject's behavior and the behavioral stimulus is judged based on the weighted fusion result, wherein the brain-controlled instructions are given a higher weight. In the fully brain-controlled stage, the consistency between the subject's behavior and the behavioral stimulus is judged based on the brain-controlled instructions. In the behavior-led stage, brain-controlled led stage, and fully brain-controlled stage, if the subject's behavior is consistent with the behavioral stimulus, a reward feedback instruction is sent to the external device control device.

[0031] In one specific embodiment, during the full behavior control stage, based on the received behavior instructions, the neural electrical signals and behavior signals before and after the behavior action are determined, and an initial decoder is trained based on the neural electrical signals and behavior signals before and after the behavior action to obtain a neural signal decoding device.

[0032] In one specific embodiment, during the full behavior control phase, once the intensity of the behavior reaches a set threshold, the external device control device provides reward feedback to the subject.

[0033] In one specific embodiment, during the training process, a threshold is set, which may be a success rate threshold, a decoding model confidence threshold, a training duration threshold, etc. When the threshold is exceeded, the current stage proceeds to the next stage.

[0034] In one specific embodiment, during the behavior-dominated phase and the mind-controlled phase, the weights of the behavioral instructions and the mind-controlled instructions are adjusted in real time. The method for adjusting the weights includes: The behavior-driven or mind-controlled stage is divided into multiple training blocks, each of which includes multiple trials. When the short-term task success rate of the training block is greater than the set success rate threshold, and the average intensity of the decoded behavior action in each training block is greater than the set intensity threshold, the weight of the mind-controlled instruction is increased in the next training block. If the short-term task success rate of a training block is less than the set lower limit of the success rate threshold, then in the next training block, the weight of the brain control command is reduced or paused.

[0035] The all-weather, fully autonomous brain-computer interface training system provided in a specific embodiment of the present invention also includes a host computer training management platform. The host computer training management platform is connected to a neural signal recording device, an external device control device, a neural signal decoding device, and a training progress control device, respectively. It is used to save and record neural signals and behavioral signals, to update the driver firmware of the external device control device in real time to configure training tasks, to update the adjustment algorithm of the training progress control device in real time to configure the training process, and to display and save behavioral instructions and control instructions at each stage of the training process.

[0036] The host computer training management platform, training progress control device, and neural signal decoding device provided by this invention are all configured in a PC. They use a universal configurable execution device interface to support various control systems such as feeding and watering devices.

[0037] In one specific embodiment, the all-weather, fully autonomous brain-computer interface training system provided in this embodiment also includes a wireless power supply system. Addressing the issue of insufficient generalization in training paradigms—where many systems restrict the free movement of experimental animals, resulting in significant differences between training and natural activity states, and substantially affecting neural signal characteristics that make it difficult to directly extrapolate the brain-control effects of training to scenarios in natural activity states—the wireless power supply system provided in this specific embodiment can provide subjects with a space for free movement of experimental animals, reducing the difference between training and natural activity states. This is explained in detail below: The wireless power supply system provided in this embodiment includes a radio magnetic field transmitting unit, a radio frequency transmitting coil, multiple LC resonant circuits, a radio frequency receiving device, an AC-DC conversion unit, and a power management unit. The radio magnetic field transmitting unit generates a high-frequency alternating current to drive the radio frequency transmitting coil to excite a source electromagnetic field. A test space is set in each LC resonant circuit, where the subject can move freely. Each LC resonant circuit is vertically arranged on the radio frequency transmitting coil at a predetermined interval to form a high-frequency alternating magnetic field. A radio frequency receiving device is implanted in the subject, which obtains high-frequency alternating current through the corresponding LC resonant circuit's high-frequency alternating magnetic field. This high-frequency alternating current is transmitted to the AC-DC conversion unit, which converts it into a stable voltage and supplies it to the power management unit, which then powers the neural signal recording device. This invention utilizes the above structure to provide the subject with sufficient space for free movement, minimizing the difference between the training state and the natural activity state. The reduced impact on neural signals allows the brain-controlled training effect to be directly extrapolated to a natural activity state.

[0038] Specifically, such as Figure 2 , Figure 5 and Figure 6As shown, the wireless power supply system provided in this embodiment adopts a multi-coil 6.78MHz magnetic resonance wireless power supply scheme. The radio frequency transmitting device uses a voltage-regulated Class D power amplifier to amplify the high-frequency AC signal generated by a 6.78MHz crystal oscillator and the DC power supply. The DC power supply supplies power to the power management unit through an adaptive wireless transmission power regulation unit and uses an automatic antenna tuner for impedance matching to generate a high-frequency alternating current to drive the radio frequency transmitting coil and excite the 6.78MHz electromagnetic field.

[0039] The two primary RF replica coils are vertically and uniformly distributed above the primary RF coils through a parallel ultra-low loss air capacitor to form an LC resonant circuit. They exchange energy with the source 6.78MHz electromagnetic field, forming a stable and efficient 3-transmit coil structure and generating a stronger 6.78MHz alternating magnetic field than the source 6.78MHz electromagnetic field around all the primary RF replica coils.

[0040] A mouse cage is placed inside the radio frequency replication primary coil. The experimental mouse moves freely within the cage. The radio frequency receiving device implanted in the mouse picks up energy from a 6.78MHz alternating magnetic field through an LC resonant circuit. The voltage conditioning circuit (AC-DC conversion unit) receives the induced high-frequency AC power, which is rectified by a full-bridge circuit and filtered by a passive low-pass filter before being input to a micro-charging device and finally to a low-dropout linear regulator (LDO). A constant 3.3V voltage is output to supply the back-end digital circuitry, i.e., the power management unit. This power management unit includes a dynamic power supply circuit and a battery protection circuit. The dynamic power supply circuit powers the wireless neural signal recording system. To prevent power instability caused by rapid mouse movement, a 10mAh lithium battery (integrated into the micro-charging device as claimed) is connected in parallel in the circuit. It charges when the wireless power supply is sufficient and instantly discharges to replenish the insufficient power when the power supply is insufficient.

[0041] This invention provides a method for using an all-weather, fully autonomous brain-computer interface training device. The method of using the all-weather, fully autonomous brain-computer interface training device described above includes the following steps: S1: Implant neural electrodes and the subcutaneous implantable wireless power acquisition device into the experimental animal and connect them to the head-mounted wireless neural signal recording subsystem; S2: Place the fully recovered experimental animal into the wireless power supply system to establish an all-weather wireless power supply; S3: Connect the wireless signal receiving relay subsystem and the host computer training management platform to establish a wireless communication link and start the head-mounted wireless neural signal recording subsystem; S4: Activate the external device control device, synchronously record behavioral data and neural signals, and construct the neural signal decoding device by training and optimizing the online decoding module; S5: Gradually reduce the weight of behavioral control and increase the weight of brain control. Through the training progress control device, guide the experimental animals to gradually rely entirely on neural activity for control. S6: When the success rate of brain control tasks continues to reach the set threshold, it enters the full brain control stage and realizes all-weather autonomous brain-computer interface training. In step S5, an adaptive control algorithm strategy based on machine teaching algorithm is adopted to make smooth and personalized adjustments according to the real-time performance of the experimental animals. The system monitors the working status and animal behavior in real time during training. When abnormal events are detected, such as signal loss or changes in the health status of experimental animals, the training is automatically paused and the training difficulty is adjusted to ensure data quality and the welfare of experimental animals. One application example of this invention is a brain-controlled lever task for acquiring water droplets, such as... Figure 7 The diagram illustrates the training algorithm and process.

[0042] Phase 1: Pure Behavioral Control (Days 1-3). The training progress control device sets the brain control weight to 0%. Mice need to press a physical lever with their forepaws and reach a certain pressure threshold to trigger the water droplet. Simultaneously, the system records neural signals (especially motor cortex signals) before and after lever pressing to train an initial decoder (such as Linear Discriminant Analysis (LDA)). This decoder learns to predict the "pressing intention".

[0043] Phase Two: Hybrid Control - Behavior-Driven (Days 4-7). The system begins to introduce mind control. The training algorithm is weighted at 70% for behavior and 30% for mind control. During this phase, mice still need to perform some pressing actions, but the system simultaneously reads the neural decoding results. If the decoder decodes a signal exceeding the pressing pressure threshold with high confidence, even if the actual pressing action is very slight or incomplete, the system will prematurely trigger water delivery and provide an audio cue as positive feedback for "mind control success." This phase strengthens the association between neural activity and reward.

[0044] Phase 3: Hybrid Control - Brain-Dominated (Days 8-12). Gradually increase the weight of brain control to 70% and decrease the weight of behavior to 30%. At this stage, mice only need to produce a slight intention to press or a related neural pattern to receive a reward. The system reduces the requirements for behavioral actions, encouraging mice to rely on "thinking" rather than large movements.

[0045] Phase Four: Complete Mind Control (Day 13 and beyond). Mind control weight is set to 100%. Physical levers are disabled or removed. Mice acquire water droplets entirely by generating specific neural activity patterns. The training progress control device continuously monitors the task success rate; training is considered complete if the success rate is >85% for multiple consecutive trials. Afterward, the system can record the animal's spontaneous neural activity during free mind-controlled drinking around the clock.

[0046] Another application example of this invention is the implementation of an adaptive transition algorithm. The core of the training progress control device is the implementation of an adaptive scaling module. Its decision variables include the short-term task success rate, which is the average success rate of the past 50 trials, and the neural decoding prediction value, which is the pressing pressure value output by the decoder. The following algorithm example illustrates how these decision variables ensure a smooth and personalized transition in training progress: when the short-term task success rate is greater than a set threshold (e.g., 80%) and the average decoding prediction pressure is greater than a set pressing threshold (e.g., 10N), the brain control weight is increased by 5% in the next training block (e.g., 50 trials). Conversely, if the success rate drops by more than 10%, the increase in brain control weight is paused or reduced by 5%. This adaptive mechanism ensures a smooth and personalized transition.

[0047] Another application example of this invention is parallel training of multiple animals. For example... Figure 5 As shown, a wireless power supply system can be configured with multiple LC resonant circuits, covering multiple experimental animal cages. The communication between the head-mounted devices of the animals in each cage and the single wireless signal relay subsystem does not interfere with each other. The host computer platform can run multiple independent training paradigm instances in parallel, processing data streams from different animals, maintaining different decoding models, and performing independent progress control, thereby achieving high-throughput experimental animal brain-controlled training research.

[0048] The embodiments described above provide a detailed explanation of the technical solutions and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the present invention. Any modifications, additions, and equivalent substitutions made within the scope of the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A 24 / 7 fully autonomous brain-computer interface training system, characterized in that, include: A neural signal recording device used to collect neural electrical signals and behavioral signals from subjects; An external device control device is used to provide behavioral stimuli to the subject, and then collect the behavioral instructions corresponding to the behavioral actions generated by the subject after receiving the behavioral stimuli. It is also used to provide reward feedback to the subject. A neural signal decoding device, wirelessly connected to a neural signal recording device, is used to decode received neural electrical signals and behavioral signals into brain control commands that can predict the subject's behavioral actions and the intensity of the behavioral actions. The training progress control device is connected to both an external device control device and a neural signal decoding device. It sequentially sets the training process into four stages: a fully behavioral control stage, a behavior-dominated stage, a brain-dominated stage, and a fully brain-controlled stage. In the fully behavioral control stage, it determines whether the subject's behavior matches the behavioral stimulus based on behavioral instructions. In the behavior-dominated stage, it performs a weighted fusion of behavioral instructions and brain-controlled instructions, and determines whether the subject's behavior matches the behavioral stimulus based on the weighted fusion result, with behavioral instructions having a higher weight. In the brain-dominated stage, it performs a weighted fusion of behavioral instructions and brain-controlled instructions, and determines whether the subject's behavior matches the behavioral stimulus based on the weighted fusion result, with brain-controlled instructions having a higher weight. In the fully brain-controlled stage, it determines whether the subject's behavior matches the behavioral stimulus based on brain-controlled instructions. In all four stages—behavior-dominated, brain-dominated, and fully brain-controlled—if the subject's behavior matches the behavioral stimulus, a reward feedback instruction is sent to the external device control device.

2. The all-weather, fully autonomous brain-computer interface training system according to claim 1, characterized in that, In the full behavior control stage, based on the received behavior instructions, the neural electrical signals and behavior signals before and after the behavior are generated are judged, and the initial decoder is trained based on the neural electrical signals and behavior signals before and after the behavior is generated to obtain the neural signal decoding device.

3. The all-weather, fully autonomous brain-computer interface training system according to claim 1, characterized in that, During the full behavioral control phase, once the intensity of the behavioral action reaches a set threshold, the external device control unit provides reward feedback to the subject.

4. The all-weather, fully autonomous brain-computer interface training system according to claim 1, characterized in that, During training, a threshold is set, which can be a success rate threshold, a decoding confidence threshold, or a training duration threshold. When the threshold is higher than the set threshold, the current stage is moved to the next stage.

5. The all-weather, fully autonomous brain-computer interface training system according to claim 1, characterized in that, In both the behavior-driven and mind-controlled phases, the weights of behavioral and mind-controlled instructions are adjusted in real time. The methods for adjusting these weights include: The behavior-driven or mind-controlled stage is divided into multiple training blocks, each of which includes multiple trials. When the short-term task success rate of the training block is greater than the set success rate threshold, and the average intensity of the decoded behavior action in each training block is greater than the set intensity threshold, the weight of the mind-controlled instruction is increased in the next training block. If the short-term task success rate of a training block is less than the set lower limit of the success rate threshold, then in the next training block, the weight of the brain control command is reduced or paused.

6. The all-weather, fully autonomous brain-computer interface training system according to claim 1, characterized in that, It also includes a host computer training management platform, which is connected to the neural signal recording device, the external device control device, the neural signal decoding device, and the training progress control device. The host computer training management platform is used to save and record neural signals and behavioral signals, to update the driver firmware of the external device control device in real time to configure training tasks, to update the adjustment algorithm of the training progress control device in real time to configure the training process, and to display and save behavioral instructions and control instructions at each stage of the training process.

7. The all-weather, fully autonomous brain-computer interface training system according to claim 1, characterized in that, It also includes a wireless power supply system, which includes a radio magnetic field transmitting unit, a radio frequency transmitting coil, multiple LC resonant circuits, a radio frequency receiving device, an AC-DC conversion unit, and a power management unit; The radio magnetic field transmitting unit is used to generate a high-frequency alternating current to drive the radio frequency transmitting coil to excite the source electromagnetic field. A test space is set up in each LC resonant circuit, and the subject moves freely in the test space. Each LC resonant circuit is set vertically on the radio frequency transmitting coil at a set interval to form a high-frequency alternating magnetic field from the source electromagnetic field. A radio frequency receiving device is implanted in the subject to obtain high-frequency alternating current through a high-frequency alternating magnetic field obtained by a corresponding LC resonant circuit. The high-frequency alternating current is then transmitted to an AC-DC conversion unit, which converts the high-frequency alternating current into a stable voltage and supplies it to a power management unit. The power management unit then supplies power to the neural signal recording device.

8. The all-weather, fully autonomous brain-computer interface training system according to claim 1, characterized in that, The training progress control device includes a training phase management module, a neural control ratio adjustment module, and a feedback adjustment module. The training phase management module is used to set the training task structure and parameter settings for different training phases; The neural control ratio adjustment module is used to update the weight ratio of neural control and behavioral control in real time according to the valve threshold. The feedback adjustment module is used to update the current training phase in real time based on the valve threshold.

9. The all-weather, fully autonomous brain-computer interface training system according to claim 1, characterized in that, The neural signal recording device includes a head-mounted wireless neural signal recording subsystem and an external wireless signal relay subsystem; The head-mounted wireless neural signal recording subsystem is used to acquire the subject's neural electrical signals and behavioral signals, and transmit the neural electrical signals and behavioral signals to the wireless signal relay subsystem; The wireless signal relay subsystem is used to transmit received neural electrical signals and behavioral signals to the neural signal decoding device.

Images