A method and system for promoting motor development

Abstract

The present application relates to the technical field of brain-computer interface assisted motor rehabilitation. The present application provides a motor development promotion method and system, wherein the motor development promotion method comprises: playing a target video segment containing a target action recorded at a preset visual angle to a target user; collecting a first electroencephalogram signal of a target cortical area of the target user during the playing; determining an energy change rate based on the first electroencephalogram signal; generating a synchronous assistance instruction when the energy change rate exceeds an activation threshold; and sending the synchronous assistance instruction to a flexible exoskeleton to drive the flexible exoskeleton to assist the target user to perform the target action with the target limb. The present application converts the mirror neuron electroencephalogram features originally used for observation into real-time engineering control instructions, so that the cognitively limited population who cannot actively cooperate can also achieve synchronous activation of brain nerves and muscles, thereby improving the effectiveness of motor function intervention.

Description

Technical Field

[0001] This invention relates to the field of brain-computer interface-assisted motor rehabilitation technology, and in particular to a method and system for promoting motor development. Background Technology

[0002] Brain-computer interface (BCI) is a novel communication and control system that does not rely on the brain's normal output pathways (peripheral nerves and muscles). Its core principle is to directly translate a person's conscious intentions into control commands that can drive external devices by collecting and analyzing brain neural activity signals. In the field of motor rehabilitation, BCI technology based on motor imagery has been widely used in neurofunctional rehabilitation training. Traditional BCI rehabilitation programs are mostly geared towards adult stroke patients or children over 6 years old, requiring users to possess the cognitive ability and willingness to actively engage in motor imagery. However, infants aged 0-6 years with special needs or other cognitively limited individuals (such as patients with impaired consciousness or severe stroke patients) are limited by their cognitive development and language comprehension abilities, making it impossible for them to understand and actively perform motor imagery tasks. This renders BCI programs that rely on conscious intention unsuitable for these special populations.

[0003] Therefore, we hope to propose a method and system for promoting motor development in order to improve the applicability and versatility of rehabilitation training. Summary of the Invention

[0004] The invention includes a method for promoting motor development, the method comprising: playing a target video clip to a target user, the target video clip containing a target action recorded from a preset perspective; during the playback of the target video clip, acquiring a first electroencephalogram (EEG) signal of a target cortical region of the target user; determining the energy change rate of the first EEG signal based on the first EEG signal; generating a synchronization assistance command when the energy change rate exceeds an activation threshold; and sending the synchronization assistance command to a flexible exoskeleton to drive the flexible exoskeleton to assist the target user's target limb in performing the target action.

[0005] In some embodiments, before determining the rate of energy change, the method further includes: acquiring inertial data of the flexible exoskeleton in real time; and reconstructing the first electroencephalogram (EEG) signal based on the inertial data and the first EEG signal.

[0006] In some embodiments, reconstructing the first EEG signal based on the inertial data and the first EEG signal includes: reconstructing the first EEG signal based on a dual-thread architecture, the dual-thread architecture including a background filtering thread and a foreground denoising thread; when the linear acceleration in the inertial data is detected to exceed an acceleration threshold, the background filtering thread updates the spatial denoising matrix based on the first EEG signal and the inertial data; the foreground denoising thread receives the first EEG signal and reconstructs the first EEG signal based on the first EEG signal and the spatial denoising matrix.

[0007] In some embodiments, the method further includes: acquiring a second EEG signal from the frontal pole region; extracting a first energy and a second energy from the second EEG signal in a first monitoring frequency band and a second monitoring frequency band, respectively, based on the second EEG signal, wherein the frequency of the first monitoring frequency band is lower than that of the second monitoring frequency band; determining the instantaneous energy ratio of the first energy and the second energy; determining a compensation ratio coefficient when the slope of the change in the instantaneous energy ratio is positive; and adjusting the activation threshold based on the compensation ratio coefficient.

[0008] In some embodiments, during the process of driving the flexible exoskeleton to assist the target user's target limb in performing the target action, the method further includes: real-time acquisition of the actual motion trajectory of the target limb following the motion based on an inertial measurement unit, wherein the inertial measurement unit is disposed inside the flexible exoskeleton; determining the dynamic deviation value between the actual motion trajectory and the target motion trajectory of the target action; when the dynamic deviation value exceeds a preset resistance threshold, switching the control mode of the flexible exoskeleton to a compliant control mode, and updating the torque output parameters for the next synchronous assistance command execution.

[0009] The invention includes a motor development promotion system, comprising: a video playback module configured to play a target video clip to a target user, the target video clip including a target action recorded from a preset perspective; a data acquisition module configured to acquire a first electroencephalogram (EEG) signal of a target cortical region of the target user during the playback of the target video clip; an energy change rate determination module configured to determine the energy change rate of the first EEG signal based on the first EEG signal; an auxiliary instruction generation module configured to generate a synchronous auxiliary instruction when the energy change rate exceeds an activation threshold; and a driving module configured to send the synchronous auxiliary instruction to a flexible exoskeleton to drive the flexible exoskeleton to assist the target user's target limb in performing the target action.

[0010] In some embodiments, the system further includes a reconstruction module configured to: acquire inertial data of the flexible exoskeleton in real time; and reconstruct the first EEG signal based on the inertial data and the first EEG signal.

[0011] In some embodiments, the reconstruction module is further configured to: reconstruct the first EEG signal based on a dual-thread architecture, the dual-thread architecture including a background filtering thread and a foreground denoising thread; when the linear acceleration in the inertial data exceeds an acceleration threshold, the background filtering thread updates the spatial denoising matrix based on the first EEG signal and the inertial data; the foreground denoising thread receives the first EEG signal and reconstructs the first EEG signal based on the first EEG signal and the spatial denoising matrix.

[0012] In some embodiments, the system further includes a threshold adjustment module, which is configured to: acquire a second EEG signal from the frontal pole region; extract a first energy and a second energy from the second EEG signal in a first monitoring frequency band and a second monitoring frequency band, respectively, based on the second EEG signal, wherein the frequency of the first monitoring frequency band is lower than that of the second monitoring frequency band; determine the instantaneous energy ratio of the first energy and the second energy; determine a compensation ratio coefficient when the slope of the change in the instantaneous energy ratio is positive; and adjust the activation threshold based on the compensation ratio coefficient.

[0013] In some embodiments, the system further includes a control mode switching module. During the process of driving the flexible exoskeleton to assist the target user's target limb in performing the target action, the control mode switching module is configured to: collect the actual motion trajectory of the target limb following the motion in real time based on an inertial measurement unit, wherein the inertial measurement unit is disposed inside the flexible exoskeleton; determine the dynamic deviation value between the actual motion trajectory and the target motion trajectory of the target action; and when the dynamic deviation value exceeds a preset resistance threshold, switch the control mode of the flexible exoskeleton to a compliant control mode and update the torque output parameters for the next synchronous assistance command execution.

[0014] The beneficial effects of the above-mentioned invention include, but are not limited to: utilizing the "observation-based firing" characteristic of mirror neurons, bypassing the active thinking process of cognitively limited individuals (such as infants and toddlers), and achieving unconscious passive induction through first-person perspective video. Combined with dynamically calibrated passive activation thresholds, it accurately captures the faint mirror responses of infants and toddlers, and provides synchronous physical assistance through a flexible exoskeleton, constructing a new pathway for promoting motor development through a "visual induction-neural response-proprioceptive closed loop," greatly improving the effectiveness of interventions for the motor development of cognitively limited individuals. Attached Figure Description

[0015] This specification will be further described by way of exemplary embodiments, which will be described in detail with reference to the accompanying drawings. These embodiments are not limiting; in these embodiments, the same reference numerals denote the same structures, wherein:

[0016] Figure 1 These are schematic diagrams illustrating application scenarios of the motor development promotion system according to some embodiments of this specification; Figure 2 This is an exemplary flowchart of a method for promoting motor development according to some embodiments of this specification; Figure 3 This is an exemplary schematic diagram illustrating the reconstruction of a first electroencephalogram signal according to some embodiments of this specification; Figure 4 This is an exemplary schematic diagram illustrating the adjustment of the activation threshold according to some embodiments of this specification; Figure 5 This is an exemplary flowchart illustrating the control modes for switching flexible exoskeletons according to some embodiments of this specification; Figure 6 This is an exemplary block diagram of a motor development promotion system according to some embodiments of this specification. Detailed Implementation

[0017] To more clearly illustrate the technical solutions of the embodiments in this specification, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are merely some examples or embodiments of this specification. For those skilled in the art, these drawings can be applied to other similar scenarios without creative effort. Unless obvious from the context or otherwise specified, the same reference numerals in the drawings represent the same structures or operations.

[0018] It should be understood that the terms “system,” “device,” “unit,” and / or “module” used herein are one way to distinguish different components, elements, parts, sections, or assemblies at different levels. However, if other terms can achieve the same purpose, they may be replaced by other expressions.

[0019] Unless the context clearly indicates an exception, words such as "a," "an," "a kind," and / or "the" do not specifically refer to the singular and may also include the plural. Generally speaking, the terms "comprising" and "including" only indicate the inclusion of explicitly identified steps and elements, which do not constitute an exclusive list, and the method or apparatus may also include other steps or elements.

[0020] Flowcharts are used in this specification to illustrate the operations performed by the system according to embodiments of this specification. It should be understood that the preceding or following operations are not necessarily performed in exact order. Instead, the steps can be processed in reverse order or simultaneously. Furthermore, other operations can be added to these processes, or one or more steps can be removed from them.

[0021] Traditional brain-computer interface (BCI) motor rehabilitation systems heavily rely on users' active motor imagery, making them difficult to apply effectively to cognitively limited populations. Traditional passive interventions, on the other hand, lack the coordinated participation of brain neurons. This specification provides a method and system for promoting motor function based on passive induction by mirror neurons. In some embodiments, utilizing the characteristic that mirror neurons fire when observing the actions of others, a first-person perspective video of standard limb movements is played to the target user. This passively activates mirror neurons in the corresponding motor cortex during unconscious viewing, without requiring comprehension of instructions or active imagery. Simultaneously, a multi-channel EEG acquisition device records EEG signals from the brain regions corresponding to the target limb in real time. Combined with a shared data bus for motor state and independent component analysis to remove artifacts related to body twisting, electromyography (EMG), and electrooculography (EOG), specific frequency band Mu wave desynchronization features are extracted. Based on a pre-calibrated passive activation threshold dynamically defined using a resting baseline and maximum response segment, the rate of change of Mu wave energy within a sliding window is calculated. When the desynchronization decrease exceeds the threshold, the mirror neurons are determined to be effectively activated, and a synchronization auxiliary command is generated. The exoskeleton then assists the target user's limbs in performing movements consistent with the video. Simultaneously, motion sensors compare the actual trajectory with a standard trajectory in real time. When muscle stiffness or active resistance is detected, the movements are automatically softened or the video playback speed is adjusted. In this embodiment, mirror neuron EEG characteristics, originally used only for observation, are transformed into real-time engineering control commands. This enables cognitively limited individuals who cannot actively cooperate to achieve synchronized activation of brain nerves and muscles, improving the effectiveness of motor function intervention.

[0022] Figure 1 These are schematic diagrams illustrating application scenarios of the motor development promotion system according to some embodiments of this specification. For example... Figure 1 The application scenario 100 of the motor development promotion system shown includes a terminal 110, a processor 120, a network 130, a flexible exoskeleton 140, a storage device 150, and an electrical signal acquisition device 160.

[0023] Terminal 110 refers to a display device used to present visual stimuli to a target user and play target video clips. Terminal 110 is used to display video clips containing target actions to the target user from a preset perspective, inducing passive activation responses in the target user's mirror neurons through visual input. In some embodiments, terminal 110 may include one or more devices with video playback capabilities, such as a tablet computer, display screen, projection device, virtual reality (VR) headset, augmented reality (AR) device, etc. In some embodiments, terminal 110 may be fixedly mounted on a stand in front of the target user or worn on the target user's head. In some embodiments, terminal 110 may include multiple display units, presenting different perspectives to the target user's left and right eyes respectively, to enhance the immersive experience of the first-person perspective.

[0024] The processor 120 is capable of processing information and / or data related to the motor development promotion system to perform one or more functions described in this specification. In some embodiments, the processor 120 can be used to receive a first electroencephalogram (EEG) signal acquired by the electrical signal acquisition device 160, reconstruct and denoise the first EEG signal based on inertial data, calculate the energy change rate of the reconstructed EEG signal, compare the energy change rate with a passive activation threshold to determine the activation state of mirror neurons, generate a synchronous auxiliary command containing the action type, target trajectory, and driving torque parameters when activated, and send the command to the flexible exoskeleton 140. In some embodiments, the processor 120 can also be used to calculate a fatigue state compensation coefficient based on a second EEG signal in the frontal pole region, dynamically adjust the passive activation threshold, calculate a dynamic deviation value based on the actual motion trajectory fed back by the flexible exoskeleton, switch the compliance control mode and update the torque output parameters when the deviation exceeds the threshold. For detailed descriptions of related content, please refer to [link to relevant documentation]. Figures 2 to 5 And its related descriptions.

[0025] In some embodiments, processor 120 may include a central processing unit (CPU), a digital signal processor (DSP), a system-on-a-chip (SoC), a microcontroller unit (MCU), an embedded processor, or any combination thereof. In some embodiments, processor 120 may include a single processing unit or multiple processing units, which may execute different computational tasks in parallel, such as using a dual-thread architecture to perform background filtering and foreground real-time noise reduction respectively. In some embodiments, processor 120 may be local or remote. In some embodiments, processor 120 may be implemented on a cloud platform. By way of example only, cloud platforms may include private clouds, public clouds, hybrid clouds, community clouds, distributed clouds, internal clouds, multi-layer clouds, or any combination thereof. In some embodiments, processor 120 may be integrated into or included in one or more other components (e.g., terminal 110, electrical signal acquisition device 160) of the application scenario 100 of the motor development promotion system.

[0026] Network 130 may include any suitable network that facilitates information and / or data exchange within the motor development facilitator system. In some embodiments, one or more components of the motor development facilitator system (e.g., terminal 110, processor 120, flexible exoskeleton 140, storage device 150, electrical signal acquisition device 160, etc.) may communicate information and / or data with one or more other components within the system via network 130. For example, electrical signal acquisition device 160 may transmit acquired electroencephalogram (EEG) signals to processor 120 via network 130; processor 120 may transmit synchronization assistance commands to flexible exoskeleton 140 via network 130. In some embodiments, network 130 may be a wired or wireless network, such as WiFi, Bluetooth, Zigbee, near field communication (NFC), Ethernet, etc., or any combination thereof.

[0027] The flexible exoskeleton 140 refers to an execution device worn on a target limb of a target user to provide physical assist torque. The flexible exoskeleton 140 receives synchronous assist commands generated by the processor 120, drives the flexible actuator to generate assist torque, and causes the target limb to perform a target action corresponding to the target video clip. In some embodiments, the flexible exoskeleton 140 may include a flexible actuator, a joint fixation structure, straps, and an underlying controller. In some embodiments, the flexible actuator may be a pneumatic bellows, shape memory alloy, artificial muscle, or other flexible actuation methods. In some embodiments, the flexible exoskeleton 140 may contain an inertial measurement unit (IMU) and a pressure sensor to collect real-time data on the actual movement trajectory of the target limb and pressure changes within the pneumatic bellows, and feed this data back to the processor 120.

[0028] Storage device 150 is capable of storing data, instructions, and / or any other information. In some embodiments, storage device 150 may store data acquired from electrical signal acquisition device 160, flexible exoskeleton 140, and / or processor 120, such as electroencephalogram (EEG) signal data, inertial data, target video clips, baseline signal energy, passive activation thresholds, visuomotor mapping databases, standard motion trajectory matrices, target user limb parameters, control command records, etc. In some embodiments, storage device 150 may include mass storage, removable storage, volatile read-write memory, read-only memory (ROM), flash memory, etc., or any combination thereof. In some embodiments, storage device 150 may execute on a cloud platform. In some embodiments, storage device 150 may be connected to network 130 to communicate with one or more other components of the motor development facilitation system (e.g., processor 120, terminal 110, etc.). In some embodiments, storage device 150 may be part of processor 120.

[0029] The electrical signal acquisition device 160 refers to a device used to acquire electroencephalogram (EEG) signals from a target user. For example, the electrical signal acquisition device 160 includes a multi-channel dry electrode headband acquisition device, a wet electrode EEG cap acquisition device, a flexible patch acquisition device, etc. The electrical signal acquisition device 160 is used to acquire, in real time, a first EEG signal from the target cortical region and a second EEG signal from the frontal pole region of the target user, and transmits the acquired EEG data to the processor 120 for subsequent processing. In some embodiments, the electrical signal acquisition device 160 may include multi-channel EEG acquisition electrodes, a preamplifier circuit, an analog-to-digital converter, and a data transmission module, etc. In some embodiments, the EEG acquisition electrodes can be dry electrodes or wet electrodes, and the number of electrode channels can be configured according to actual needs, for example, including channels such as C3, C2, Fp1, and Fp2. In some embodiments, the sampling rate of the electrical signal acquisition device 160 can be from 500Hz to 2000Hz to ensure that the temporal resolution of the EEG signal meets the requirements for real-time processing.

[0030] Figure 2 This is an exemplary flowchart of a method for promoting motor development according to some embodiments of this specification. In some embodiments, process 200 may be executed by a processor (such as processor 120). Figure 2 As shown, process 200 includes the following steps.

[0031] Step 210: Play the target video clip to the target user.

[0032] Target users refer to subjects who passively receive visual guidance and physical assistance. For example, target users include infants and young children who need motor rehabilitation training, patients with impaired consciousness, patients with severe stroke, and other patients with severe cognitive impairment.

[0033] A target video clip refers to video material recorded from a preset perspective that can present a specific standard body movement process and is used to visually guide a target user. In some embodiments, a target video clip includes a target action recorded from a preset perspective.

[0034] Preset perspective refers to using the first-person line of sight of the subject of the action as the perspective of the recorded footage. The subject of the action refers to the individual actually performing the target action in the target video clip. The action performed by the subject can serve as a visual stimulus, guiding the activation of neurons and the imitation of limb movements by the target user.

[0035] A target action refers to a standard physical movement performed by the subject in a target video clip. For example, target actions include standard movements such as grasping and kicking.

[0036] In some embodiments, the processing device control terminal 110 retrieves a target video segment from a storage device (such as storage device 150) and decodes and plays it. The video content may be a standard action video recorded from a first-person perspective, such as a grasping action video or a kicking action video.

[0037] In some embodiments, the processing device may also acquire target video segments from the cloud and cache them locally before playback. The processing device may select and play a matching segment from multiple target video segments based on the target limb, age stage, or training task corresponding to the target user.

[0038] Step 220: During the playback of the target video segment, the first EEG signal of the target user's target cortical region is acquired.

[0039] The target cortical region refers to the motor cortical region in the target user's brain specifically responsible for controlling the movement of the target limb. For example, the target cortical region includes the C3 channel in the left motor cortex that controls the right upper limb, or the Cz channel in the central area that controls the lower limb.

[0040] The first EEG signal refers to the time-series data of weak electrical potential changes in the target cortical region recorded by multi-channel electrodes placed on the target user's scalp. The first EEG signal is used to reflect the neural activity state of the motor cortex of the target user during visual stimulation or physical assistance.

[0041] In some embodiments, the processor controls the EEG acquisition device to continuously sample the target cortical region during the playback of the target video segment. Taking right upper limb training as an example, the processor can acquire EEG signals from channel C3; taking lower limb training as an example, the processor can acquire EEG signals from channel C2. The sampling frequency can be a high-frequency sampling rate, such as 1000Hz, to preserve subtle changes within the target frequency band. It should be noted that the acquisition of the target user's EEG signals (such as the first EEG signal, the second EEG signal, etc.) in the embodiments of this specification is carried out with the permission of the target user or their guardian. Furthermore, the acquired EEG signals are only used for the necessary analysis in this specification.

[0042] Step 230: Determine the energy change rate of the first EEG signal based on the first EEG signal.

[0043] The rate of change of energy refers to the percentage change in the current real-time energy of the first EEG signal relative to the baseline signal energy.

[0044] Current real-time energy refers to the real-time signal power of the first EEG signal extracted within the sliding time window.

[0045] Baseline signal energy refers to the average signal power of the first EEG signal collected when the target user is at rest.

[0046] In some embodiments, the processor performs target-band bandpass filtering on the first EEG signal and establishes a sliding time window. The processor calculates the power spectral density of the signal within the current window to obtain the current real-time energy E. c And further based on the pre-stored baseline signal energy E b Determine the rate of energy change. The target frequency band can be the Mu band, for example, 6-9 Hz. A sliding time window refers to a data segment formed by sequentially shifting a certain step size (e.g., 0.5 seconds) in a continuous EEG signal time sequence according to a preset time period (e.g., 2 seconds). The sliding time window is used to extract and analyze the signal features within the current time period at each moment. The current window refers to the data segment corresponding to the current time interval used for signal analysis and calculation during the sequential movement of the sliding time window. For example, the processor can determine the rate of energy change of the first EEG signal based on the following formula (1): ERD (%) = (E b -E c ) / E b ×100%(1) Wherein, ERD(%) represents the rate of change of energy of the first EEG signal, E c E represents the current real-time energy. b This represents the baseline signal energy.

[0047] Step 240: When the energy change rate exceeds the activation threshold, a synchronization auxiliary command is generated.

[0048] The activation threshold is a lower reference limit for determining whether the mirror neurons of the target user are effectively activated. Synchronization assist instructions are digital control signals used to trigger physical devices (such as flexible exoskeletons) to perform actions. Synchronization assist instructions include the torque and motion trajectory parameters of the physical device. The processor generates synchronization assist instructions when it detects that the rate of energy change is greater than or equal to the activation threshold.

[0049] In some embodiments, the processor can determine personalized activation thresholds based on different target subjects before rehabilitation training. In some embodiments, the processor first plays a video of natural scenery without human movement, extracts the average energy of the target frequency band in the first EEG signal as the baseline signal energy E1 in the resting state; then plays a video of the target movement, extracts the energy of the target frequency band in the first EEG signal within the time window of the most significant energy decrease as the response segment energy E2; calculates the target subject's EEG activation rate α = (E1 - E2) / E1; and then multiplies the target subject's EEG activation rate α by a preset coefficient to obtain the activation threshold.

[0050] In some embodiments, the processor may first determine the action type and target trajectory of the target action in the target video segment, and then determine the driving torque parameters of the flexible exoskeleton based on the target trajectory.

[0051] Action type refers to the identifier of the category to which the target action belongs, such as "grasping" or "kicking". In some embodiments, the action type can be identified by the processor reading the metadata or timestamp of the current video frame in real time while the target video segment is playing, and querying a preset video-action mapping database to identify the specific action tag currently displayed in the video frame (e.g., by parsing a specific time period of the video, the processor obtains the action type corresponding to that time period). For example, the processor can set each target video segment to have metadata information indicating the action category it contains, or register the mapping relationship between action categories and the video timeline during the video production stage. During playback, while synchronously capturing video frame information, the processor automatically calls the database to retrieve the action type identifier corresponding to the current time point, thereby realizing the determination and extraction of the action type.

[0052] A target trajectory refers to a set of three-dimensional spatial trajectory points corresponding to a target limb movement, used to guide that movement. In some embodiments, the target trajectory can be obtained by retrieving a three-dimensional spatial baseline trajectory matrix stored in a preset database, based on the standard movement trajectory corresponding to the target movement. This is combined with scaling and coordinate system transformation processing of the current target user's limb parameters (such as arm length, joint range of motion limits, and other calibration data) to generate a set of actual three-dimensional spatial trajectory points specific to the individual target user and adapted to the target movement. For example, when parsing a target video segment played on the current terminal, the processor can identify the type of movement being performed and then retrieve the standard baseline trajectory corresponding to that movement type. Subsequently, based on pre-collected limb parameter data of the target user, the processor performs kinematic parameter transformations (such as forward / inverse kinematics and spatial scaling) on ​​the standard baseline trajectory, ultimately forming a target trajectory that can be used to drive the flexible exoskeleton and matches the individual anatomical features of the target user.

[0053] Flexible exoskeletons are wearable devices that are worn on the target limb and provide physical traction torque assistance to the target user through flexible actuators such as pneumatic bellows.

[0054] Driving torque parameters refer to control parameters used to characterize the output torque required for a flexible exoskeleton to generate the target movement. Driving torque parameters include the required torque value and control signals for pulling the target object to perform the target movement. For example, driving torque parameters include the target inflation pressure and the opening degree of the air intake valve.

[0055] In some embodiments, the processing device may also generate synchronization assistance instructions using a template lookup method. The processing device determines the action stage based on the video playback progress, directly reads the joint angle sequence and air pressure control curve corresponding to that stage from the action template library, and then performs amplitude scaling based on the target user's limb parameters to obtain the synchronization assistance instructions.

[0056] Step 250: Send the synchronization assistance command to the flexible exoskeleton to drive the flexible exoskeleton to assist the target user's target limb in performing the target action.

[0057] The target limb refers to the specific body part of the target user that corresponds to the target action performed in the target video clip. For example, if the target action is raising an arm, the target limb could be the target user's arm; if the target action is kicking, the target limb could be the target user's leg.

[0058] In some embodiments, the processor sends synchronization assistance commands to the underlying controller of the flexible exoskeleton via wired or wireless means. After parsing the motion trajectory parameters and torque parameters, the underlying controller drives actuators such as pneumatic bellows to output traction torque, assisting the target limb in completing the target movement, and smoothly releasing the torque after the movement ends. Smooth torque release means that after the target limb completes the target movement, the underlying controller controls the actuators to gradually and continuously reduce the traction torque to zero, avoiding discomfort or potential damage to the target limb due to sudden torque changes.

[0059] The underlying controller refers to the control unit integrated inside the flexible exoskeleton, which receives and analyzes the motion trajectory parameters and torque parameters in the synchronous assistance commands, and controls the pneumatic bellows and other actuators to output traction torque in real time to assist the target limb in completing the predetermined action.

[0060] In some embodiments, during the process of driving the flexible exoskeleton to assist the target user's target limb in performing a target action, the processor can also determine whether it is necessary to switch the control mode of the flexible exoskeleton based on the dynamic deviation between the actual movement trajectory of the target limb and the target movement trajectory of the target action. For more information, please refer to... Figure 5 And its related descriptions.

[0061] In some embodiments of this specification, the "observation-based firing" characteristic of mirror neurons is utilized to bypass the active thinking process of cognitively limited individuals (such as infants and toddlers), achieving unconscious passive induction through first-person perspective video. Combined with dynamically calibrated passive activation thresholds, the subtle mirror responses of infants and toddlers are precisely captured, and synchronous physical assistance is provided through a flexible exoskeleton. This constructs a new pathway for promoting motor development through a "visual induction-neural response-proprioceptive closed loop," significantly improving the effectiveness of interventions for the motor development of cognitively limited individuals.

[0062] Figure 3 This is an exemplary schematic diagram illustrating the reconstruction of a first electroencephalogram (EEG) signal according to some embodiments of this specification. In some embodiments, process 300 may be executed by a processor (such as processor 120). Figure 3 As shown, process 300 includes the following steps.

[0063] Step 310: Real-time acquisition of inertial data of the flexible exoskeleton.

[0064] Inertial data refers to data sequences that reflect the spatial motion state of a target limb. Inertial data includes angular velocity sequences and linear acceleration sequences of the limb.

[0065] In some embodiments, the processor reads the X-axis, Y-axis, and Z-axis angular velocity data, as well as the X-axis, Y-axis, and Z-axis linear acceleration data, output by the inertial measurement unit inside the flexible exoskeleton at a preset frequency (e.g., 100 Hz) via a motion state sharing data bus located between the flexible exoskeleton and the EEG acquisition layer. The EEG acquisition layer is an EEG acquisition hardware pathway composed of electrodes, a front-end amplifier circuit, and a sampling interface of the EEG acquisition device. The motion state sharing data bus is used to synchronously provide the inertial data of the flexible exoskeleton to the processor and the EEG processing module that cooperates with the EEG acquisition hardware pathway. The processor adds timestamps to each sampling point and generates a continuous inertial data sequence from the X-axis, Y-axis, and Z-axis angular velocity data and linear acceleration data in chronological order. The X-axis, Y-axis, and Z-axis refer to the orthogonal coordinate axes used by the inertial measurement unit in three-dimensional space to represent the front-back, left-right, and up-down directions, respectively, and are spatial parameters characterizing limb movement. In the embodiments of this specification, the X-axis, Y-axis, and Z-axis can be collectively referred to as the three axes.

[0066] Step 320: Reconstruct the first EEG signal based on inertial data and the first EEG signal.

[0067] In some embodiments, the processor reconstructs the first EEG signal based on inertial data and the first EEG signal by: constructing an artifact prior matrix based on the inertial data; decomposing the first EEG signal using a preset analysis algorithm to obtain multiple independent principal components of the first EEG signal; calculating the correlation degree between each independent principal component and the artifact prior matrix, and adjusting the amplitude of the corresponding independent principal components based on the correlation degree; and reconstructing the first EEG signal based on the adjusted multiple independent principal components.

[0068] In some embodiments, the processor first constructs an artifact prior matrix based on inertial data. For example, the processor receives a three-axis angular velocity sequence and a three-axis linear acceleration sequence, and uses a difference algorithm to calculate the discrete first derivative of the linear acceleration sequence with respect to time to obtain a three-axis jerk sequence. The processor aligns the three-axis angular velocities, three-axis linear accelerations, and three-axis jerk by timestamps and combines them into an N×9 two-dimensional artifact prior matrix. Each row in the artifact prior matrix corresponds to the motion state feature of a sampling point, used to characterize the signal artifact components related to the spatial motion state of the target limb.

[0069] In some embodiments, the artifact prior matrix can also be constructed in other ways. For example, the processor can generate a time-stamped matrix based on the peak positions and durations in the inertial data and use the time-stamped matrix as the artifact prior matrix.

[0070] In some embodiments, the processor performs decomposition based on a first EEG signal to obtain multiple independent principal components.

[0071] In some embodiments, the processor employs a preset analysis algorithm to perform blind source separation on the first EEG signal. The preset analysis algorithm can be a blind source separation algorithm capable of decomposing the multi-channel first EEG signal into multiple independent temporal components, such as the Fast Independent Component Analysis (FIC) algorithm. The processor inputs the multi-channel first EEG signal into the preset analysis algorithm to obtain multiple statistically independent temporal components, and uses each temporal component as an independent principal component.

[0072] In some embodiments, the processor may also employ an information-maximizing independent component analysis algorithm to decompose the first EEG signal. The processor separates multiple independent principal components based on an iteratively updated unmixing matrix and outputs the corresponding mixing or unmixing matrix for subsequent amplitude adjustment and signal reconstruction.

[0073] In some embodiments, the processor may also perform the decomposition in other ways. For example, the processor may use a second-order blind identification algorithm to separate the first EEG signal to obtain multiple independent principal components.

[0074] In some embodiments, the processor calculates correlation and magnitude adjustments based on independent principal components (IPCs) and an artifact prior matrix to obtain adjusted IPCs. For example, the processor evaluates the correlation of each IPC to identify artifact components introduced by body motion. Iterating through each IPC, the processor calculates the Pearson correlation coefficient between the component vector and each column vector in the artifact prior matrix, and takes the maximum absolute value of each correlation coefficient as the correlation of that IPC. The correlation characterizes the degree of correlation between the IPC and limb movement state; a higher correlation indicates a greater influence of motion artifacts on the component.

[0075] When the correlation of independent principal components exceeds a preset threshold, it is determined that the independent principal component is mainly composed of volumetric conducted noise caused by body twisting. The processor multiplies the independent principal component sequence by a preset adjustment coefficient to suppress its amplitude. The preset adjustment coefficient is 0 or a positive number less than 1; when the adjustment coefficient is 0, the corresponding independent principal component is completely removed; when the adjustment coefficient is a positive number less than 1, the amplitude of the corresponding independent principal component is partially attenuated to achieve a balance between suppressing artifacts and preserving effective signals.

[0076] In some embodiments, adjusting the amplitude of the corresponding independent principal components based on the correlation degree further includes: when the processor determines that the correlation degree is within a first preset range, retaining the original amplitude of the independent principal components; when the processor determines that the correlation degree is within a second preset range, reducing the amplitude of the independent principal components by a preset weighting function; and when the processor determines that the correlation degree is greater than an extremely high critical threshold, setting the amplitude of the independent principal components to zero.

[0077] Both the first and second preset intervals represent the numerical range of the correlation score. When the correlation score falls within the first preset interval, it indicates that the principal components are less affected by motion interference and contain valid EEG features. When the correlation score falls within the second preset interval, it indicates that valid EEG and artifacts are severely mixed in the principal components, and the independent principal components are significantly affected by motion artifacts. The second preset interval is higher than the first preset interval. For example, the first preset interval could be [0.4, 0.6], and the second preset interval could be [0.6, 0.85]. In some embodiments, the preset weighting function refers to a mapping equation that dynamically outputs a scaling factor between 0 and 1 based on the correlation score value. For example, the preset weighting function can be a sigmoid function or a linear function; when the correlation score is large, the output scaling factor is close to 1, and when the correlation score is small, the output scaling factor is close to 0.

[0078] The extremely high critical threshold refers to the upper limit of the correlation degree at which a principal component is determined to be completely dominated by noise and must be removed. For example, the extremely high critical threshold can be 0.85. In some embodiments, the first preset interval, the second preset interval, and the extremely high critical threshold can be preset based on the statistical distribution of the correlation degree of independent principal components in historical samples, and satisfy that the upper boundary of the first preset interval is not higher than the lower boundary of the second preset interval, and the upper boundary of the second preset interval is not higher than the extremely high critical threshold.

[0079] In some embodiments, the processor can calculate the correlation between the time series of each independent principal component and the artifact prior matrix to obtain the correlation degree of each independent principal component. The correlation degree can be a normalized correlation coefficient or a similarity value. The weight coefficient refers to the proportional coefficient assigned to the principal component according to the correlation degree of the independent principal component, based on a preset rule and weight function, to adjust the magnitude of its sampling point values. The weight coefficient ranges from 0 to 1, where 1 indicates that the principal component value is completely retained, and 0 indicates that the principal component value is set to zero. Different correlation degree intervals correspond to different weight coefficient settings. When the correlation degree is lower than the lower boundary of the first preset interval, the processor determines that the independent principal component has a low correlation with the motion artifact and sets the weight coefficient to 1.0; when the correlation degree is within the first preset interval, the weight coefficient is also set to 1.0, and the original amplitude of the independent principal component at each sampling point is retained. When the correlation degree is within the second preset interval, the processor substitutes the correlation degree into the preset weight function to obtain a weight coefficient between 0 and 1, and multiplies the value of the independent principal component at each sampling point by the weight coefficient. When the correlation degree exceeds the extremely high critical threshold, the processor sets the weight coefficient to 0 and the amplitude of the independent principal component to zero. The original amplitude can be the original value of the independent principal component output from the decomposition stage at each sampling point.

[0080] In some embodiments, the processor can also calculate the correlation degree based on the projection energy ratio or normalized cross-correlation peak between the independent principal components and the artifact prior matrix, and perform amplitude adjustment using an exponential decay function.

[0081] In some embodiments, amplitude adjustment can also be achieved in other ways. For example, the processor can map different correlation intervals to corresponding weight coefficients according to a preset lookup table relationship, and scale the independent principal components point by point accordingly.

[0082] In some embodiments, the processor performs reconstruction of the first EEG signal based on a plurality of adjusted independent principal components to obtain the first EEG signal after artifact removal.

[0083] In some embodiments, when the processor outputs the mixing matrix during the decomposition stage, it can multiply the adjusted multiple independent principal components with the mixing matrix to obtain the reconstructed multi-channel first EEG signal. For the retained independent principal components, their contributions to the reconstructed signal are fully preserved; for the attenuated or zeroed independent principal components, their artifact contributions are suppressed or removed accordingly.

[0084] In some embodiments, when the processor outputs the unmixing matrix during the decomposition phase, it can perform back projection on the adjusted independent principal components based on the inverse or pseudo-inverse matrix of the unmixing matrix to obtain the reconstructed first EEG signal. For example, the processor can recombine all the independent principal components after amplitude adjustment into a new set S_adjusted. This is achieved using the inverse matrix (W...) of the unmixing matrix W.-1 The inverse transformation operation is performed to project the signal back into the original sensor space, resulting in the reconstructed first EEG signal.

[0085] In some embodiments, the reconstruction of the first EEG signal can also be achieved in other ways. For example, the processor can superimpose multiple adjusted independent principal components one by one according to their channel contribution coefficients to obtain the reconstructed first EEG signal.

[0086] Some embodiments in this specification utilize inertial priors and segmented amplitude adjustment mechanisms to suppress motion artifacts and avoid the loss of effective EEG caused by direct removal by hard thresholds, thereby improving the fidelity of the reconstructed signal.

[0087] In some embodiments, the processor reconstructs the first EEG signal based on inertial data and the first EEG signal, including: reconstructing the first EEG signal based on a dual-thread architecture, the dual-thread architecture including a background filtering thread and a foreground denoising thread; when the linear acceleration in the inertial data exceeds an acceleration threshold, the processor updates the spatial denoising matrix based on the first EEG signal and the inertial data through the background filtering thread; the processor receives the first EEG signal through the foreground denoising thread, and reconstructs the first EEG signal based on the first EEG signal and the spatial denoising matrix. The spatial denoising matrix can be stored in the background filtering thread.

[0088] A dual-threaded architecture refers to two execution units in a processor that run in parallel and are assigned different computational priorities.

[0089] Background filtering threads refer to slow processes that run at low priority, perform high-complexity matrix factorization operations, and asynchronously update filtering parameters.

[0090] The foreground denoising thread refers to a high-priority, fast process that handles real-time data streams and performs low-latency matrix multiplication operations.

[0091] In some embodiments, the processor can initialize a background filtering thread and a foreground denoising thread. The processor assigns a low priority to the background filtering thread and a high priority to the foreground denoising thread. The processor allocates a shared memory region for the background filtering thread and the foreground denoising thread to store the spatial denoising matrix and its update timestamp, so that the background filtering thread updates the spatial denoising matrix asynchronously, while the foreground denoising thread reads the spatial denoising matrix in real time. The spatial denoising matrix is ​​a linear transformation matrix used to map multi-channel EEG signals to a shadowless space.

[0092] In some embodiments, the processor can also set up a double-buffered parameter area. After the background filtering thread updates the spatial denoising matrix, the updated spatial denoising matrix is ​​written to the spare buffer. The processor then switches the version number to enable the foreground denoising thread to read the updated spatial denoising matrix, so as to reduce the blocking of real-time denoising during the spatial denoising matrix update process.

[0093] In some embodiments, when the processor detects that the linear acceleration in the inertial data exceeds the acceleration threshold, it updates the spatial denoising matrix based on the first EEG signal and the inertial data through a background filtering thread.

[0094] An acceleration threshold is a physical scalar critical value used to determine the intensity of limb movement and trigger an update of the spatial denoising matrix. In some embodiments, the acceleration threshold can be preset based on linear acceleration baseline statistics, the range of motion intensity corresponding to the target action type, or individual parameters of the target user.

[0095] In some embodiments, the processor continuously monitors the linear acceleration data stream in the inertial data through a background filtering thread. When the vector sum of the three-axis accelerations exceeds the acceleration threshold within N consecutive sampling points, the processor intercepts the first EEG signal and inertial data for the corresponding time period. The processor performs a full blind source separation operation on the data for this time period to obtain the unmixing matrix W and the corresponding independent principal components, and constructs an amplitude adjustment coefficient matrix A based on the artifact component removal results. Here, A is used to characterize the retention, attenuation, or zeroing relationship of each independent principal component. The unmixing matrix refers to the matrix that decomposes the original observed signal into independent principal components by performing a linear transformation on the multi-channel signal in the blind source separation algorithm. The unmixing matrix is ​​used to describe the mapping relationship between the observed signal and the independent principal components. In some embodiments, the unmixing matrix can be obtained by the background filtering thread during the full blind source separation algorithm (such as the independent component analysis algorithm) performed on the intercepted first EEG signal and inertial data. Specifically, when the background filtering thread detects that the linear acceleration vector and N consecutive sampling points exceed the acceleration threshold, it extracts the signal data for the corresponding time period, decomposes the multi-channel observation signal into a set of mutually independent principal components through the blind source separation process, and calculates the unmixing matrix accordingly.

[0096] For example, the processor can update the spatial denoising matrix based on the following formula (2): W clean =W×A×W -1 (2) Among them, W clean Let W represent the spatial denoising matrix, and W represent the demixing matrix. -1 Let W be the inverse of the demixing matrix W, and A be the amplitude adjustment coefficient matrix.

[0097] In some embodiments, updating the spatial denoising matrix by the processor may further include: the processor calculating the root mean square value or peak count of linear acceleration by a background filtering thread according to a time window; when the motion intensity exceeds the acceleration threshold within the time window, the processor extracts the corresponding EEG segment, calculates the covariance matrix of the first EEG signal of the multi-channel, and constructs a projection matrix that suppresses motion artifact components using principal component decomposition or singular value decomposition, and uses the projection matrix as the updated spatial denoising matrix.

[0098] In some embodiments, the processor receives the first EEG signal through a foreground denoising thread and reconstructs the first EEG signal based on the first EEG signal and a spatial denoising matrix.

[0099] In some embodiments, the processor reconstructs the first EEG signal by: the processor causing the foreground denoising thread to run synchronously according to the EEG sampling frequency; when the foreground denoising thread receives a new frame of the first EEG signal and obtains the column vector X of the new frame of the first EEG signal, the processor reads the latest spatial denoising matrix W from the background filtering thread. clean The processor performs matrix multiplication X. r =W clean ×X, and the resulting X r This serves as the first EEG signal output after reconstruction.

[0100] In some embodiments, the processor reconstructing the first EEG signal may further include: the processor causing the foreground denoising thread to receive the first EEG signal in short data blocks; the processor reading the spatial denoising matrix once at the beginning of each data block and performing matrix multiplication on the data block to obtain the reconstructed EEG data block, and then outputting it in chronological order.

[0101] In some embodiments, the reconstruction of the first EEG signal can also be achieved in other ways. For example, the processor performs a weighted summation of the first EEG signal according to the channel weights of the spatial denoising matrix using a foreground denoising thread to obtain the reconstructed first EEG signal.

[0102] Some embodiments in this specification update the spatial denoising matrix asynchronously in the background and perform matrix multiplication in real time in the foreground, so that the denoising model is adjusted with the motion state, while taking into account both the accuracy of EEG reconstruction and the real-time performance of closed-loop control.

[0103] In some embodiments, the foreground denoising thread is further configured to: when the signal-to-noise ratio of the first EEG signal is detected to be lower than a preset signal-to-noise ratio threshold for a period of time exceeding a first preset time period, or when no synchronous auxiliary instruction is generated within a second preset time period, control the transcranial stimulation device to output a rhythmic current consistent with the first EEG signal to the target cortical region; monitor the energy change rate of the first EEG signal in real time, and when the energy change rate reaches an activation threshold, control the transcranial stimulation device to stop outputting the rhythmic current.

[0104] In some embodiments, when the foreground denoising thread detects that the signal-to-noise ratio of the first EEG signal is lower than a preset signal-to-noise ratio threshold for more than a first preset time period, or fails to generate a synchronous auxiliary instruction within a second preset time period, it controls the transcranial stimulation device to output a rhythmic current consistent with the first EEG signal to the target cortical region.

[0105] Signal-to-noise ratio (SNR) is the ratio of the energy of the target signal frequency band to the energy of the background noise frequency band. It can be calculated based on the power spectral density of the target frequency band and the power spectral density of the background noise frequency band within the current time window. The lower the value, the easier it is for the effective EEG features to be submerged by background EEG activity or residual noise.

[0106] A preset signal-to-noise ratio (SNR) threshold is a quality lower limit standard for determining whether effective features in the reconstructed signal are severely submerged. In some embodiments, the preset SNR threshold can be preset based on the target frequency band SNR distribution of the reconstructed first EEG signal, combined with the balance requirement of false trigger rate and missed trigger rate.

[0107] The first preset time period refers to the length of time used to determine whether the first EEG signal remaining in a low signal-to-noise ratio state exceeds the tolerance range. In some embodiments, the first preset time period can be preset. For example, the first preset time period can be a time window.

[0108] The second preset time period refers to the longest waiting time during which no effective neural response is generated. In some embodiments, the second preset time period can be preset. For example, the second preset time period can be 1 second, 2 seconds, 3 seconds, etc.

[0109] Transcranial stimulation (TCS) devices are neuromodulation hardware that non-invasively applies weak alternating current to specific areas of the cerebral cortex via scalp electrodes, such as tACS devices.

[0110] A rhythmic current refers to an alternating current stimulation signal with a fixed frequency at a specific hertz and exhibiting a sinusoidal waveform. In some embodiments, the frequency of the rhythmic current is 6-9 Hz. In this embodiment, the frequency of the rhythmic current can be consistent with the frequency of the target frequency band of the first EEG signal.

[0111] In some embodiments, the foreground denoising thread controlling the output rhythmic current of the transcranial stimulation device includes: the foreground denoising thread performing sliding window processing on the reconstructed first EEG signal; the foreground denoising thread performing frequency domain transformation on the data within each time window to obtain the power spectral density; the foreground denoising thread calculating the signal energy of the target frequency band and the noise energy of the adjacent reference frequency band, and determining the signal-to-noise ratio (SNR) of the current window based on their ratio. When the SNR of the current window is continuously lower than a preset SNR threshold and accumulates to exceed a first preset time period, or when the timing since the last generation of the synchronization assistance command reaches a second preset time period, the foreground denoising thread sends a start command to the transcranial stimulation device through the control interface; if the current training round has not yet generated a synchronization assistance command, the timing start point of the second preset time period can be the start time of video playback or the start time of EEG monitoring in that training round. After receiving the start command, the transcranial stimulation device outputs a rhythmic current to the corresponding electrode of the target cortical region. The frequency of the rhythmic current is consistent with the target frequency band of the first EEG signal; for example, when the main frequency of the first EEG signal is 8Hz, the transcranial stimulation device outputs an 8Hz sinusoidal alternating current.

[0112] In some embodiments, the foreground denoising thread monitors the energy change rate of the first EEG signal in real time, and when the energy change rate reaches the activation threshold, it controls the transcranial stimulation device to stop outputting rhythmic current.

[0113] In some embodiments, the foreground denoising thread monitoring the energy change rate includes: during the output of rhythmic current by the transcranial stimulation device, the foreground denoising thread continuously monitors the energy change rate of the first EEG signal. When the energy change rate reaches an activation threshold, indicating that the intrinsic rhythm of the target frequency band has been effectively activated, the foreground denoising thread sends a stop command to the transcranial stimulation device, or sets the stimulation amplitude to zero to stop the rhythmic current output. Further explanation regarding the determination of the energy change rate of the first EEG signal can be found in [link to relevant documentation]. Figure 2 .

[0114] In some embodiments, the monitoring of energy change rate by the foreground denoising thread may further include: smoothing the target frequency band energy sequence or the corresponding energy change rate sequence for multiple consecutive time windows; when the foreground denoising thread detects that the smoothed energy change rate reaches the activation threshold and continues for a preset window number, controlling the transcranial stimulation device to stop outputting rhythmic current. To reduce discomfort caused by sudden stoppage, the foreground denoising thread may also control the transcranial stimulation device to gradually reduce the stimulation amplitude to zero within a preset ramp time.

[0115] Some embodiments in this specification trigger synchronous stimulation when the signal-to-noise ratio is low or the response is absent, and close the loop to stop stimulation when the energy change rate reaches the threshold. This can actively induce neural resonance and exit the stimulation in a timely manner after the endogenous signal is restored, thereby improving the continuity and safety of the intervention.

[0116] Some embodiments in this specification introduce flexible exoskeleton inertial data as a priori motion artifact and reconstruct the first EEG signal, which can suppress irregular twisting artifacts, improve the EEG signal-to-noise ratio, and reduce false positives and false negatives.

[0117] Figure 4 This is an exemplary schematic diagram illustrating the adjustment of the activation threshold according to some embodiments of this specification. In some embodiments, process 400 may be executed by a processor (such as processor 120). Figure 4 As shown, process 400 includes the following steps.

[0118] Step 410: Obtain the second EEG signal from the frontal pole region.

[0119] The frontal pole is the anterior part of the cerebral cortex, a brain region that reflects cognitive arousal and fatigue. The frontal pole corresponds to the Fp1 / Fp2 channels of the EEG electrodes.

[0120] The second EEG signal refers to the EEG signal sequence that is independent of the target cortical region and is acquired through electrodes in the frontal pole area.

[0121] In some embodiments, the processor activates the frontal pole monitoring circuit while acquiring the first EEG signal, and reads EEG data from the Fp1 or Fp2 channel of the EEG acquisition device as the second EEG signal. The processor can extract a signal window at a preset period, for example, 5 seconds.

[0122] In some embodiments, the processor synchronously reads data from channels Fp1 and Fp2, performs impedance verification, artifact screening, or signal-to-noise ratio evaluation on the two channels, and then selects the channel with higher quality as the second EEG signal, or performs weighted fusion of the two channels to form the second EEG signal.

[0123] Step 420: Based on the second EEG signal, extract the first energy of the second EEG signal in the first monitoring frequency band and the second energy in the second monitoring frequency band.

[0124] The first monitoring frequency band refers to the low-frequency EEG band that reflects the brain's sleepiness state. For example, the first monitoring frequency band could be theta waves with a frequency of 4-8 Hz.

[0125] The first energy refers to the energy value of the second EEG signal within the same time window in the first monitoring frequency band. For example, if the first monitoring frequency band is 4-8 Hz (Theta wave), the sum of the energies of the second EEG signal in the 4-8 Hz frequency band within a certain time window is the first energy corresponding to that time window.

[0126] The second monitoring frequency band refers to the high-frequency EEG band that reflects the brain's awake state. For example, the first monitoring frequency band could be alpha waves with a frequency of 8-13 Hz.

[0127] The second energy refers to the energy value of the second EEG signal within the same time window in the second monitoring frequency band. For example, when the second monitoring frequency band is 8-13Hz (Alpha band), the sum of the energy of the second EEG signal in the 8-13Hz frequency band within a certain time window is the first energy corresponding to that time window.

[0128] In some embodiments, the processor inputs a second EEG signal window into a preset transformation algorithm, such as a dual-tree complex wavelet transform. The processor decomposes the signal into different frequency bands using a multi-level filter bank, reconstructs the detail coefficients in the 4-8Hz band and calculates the sum of squares to obtain the first energy; it also reconstructs the detail coefficients in the 8-13Hz band and calculates the sum of squares to obtain the second energy. Detail coefficients refer to the components reflecting the high-frequency or local variation characteristics of the signal obtained within each specific frequency band after multi-level decomposition of the EEG signal using the preset transformation algorithm. Detail coefficients are used to represent the subtle structure or variation information of the signal within that frequency band.

[0129] Step 430: Determine the instantaneous energy ratio of the first energy to the second energy.

[0130] The instantaneous energy ratio refers to the ratio of the energy in the first monitoring frequency band to the energy in the second monitoring frequency band within the same time window. The instantaneous energy ratio of the first energy to the second energy is used to quantify the level of whole-brain arousal.

[0131] Step 440: When the slope of the instantaneous energy ratio change is positive, determine the compensation ratio coefficient.

[0132] The slope of change refers to the sign of the first derivative of the fitted curve of the instantaneous energy ratio over multiple consecutive time periods. A positive slope of the instantaneous energy ratio indicates that the instantaneous energy ratio is increasing over time, suggesting that the target user is transitioning from a state of alertness to a state of drowsiness or fatigue.

[0133] The compensation scaling factor is a multiplier factor used to dynamically scale the original passive activation threshold. The value of the compensation scaling factor ranges between 0 and 1.

[0134] In some embodiments, the processor reads the instantaneous energy ratios of multiple consecutive cycles and performs linear fitting using the least squares method to obtain the slope of change. When the slope of change is greater than 0, the processor performs a preset mapping based on the slope of change and the current instantaneous energy ratio to determine the compensation ratio coefficient; the larger the slope of change and the higher the instantaneous energy ratio, the smaller the compensation ratio coefficient can be.

[0135] In some embodiments, the processor calculates a first-order difference sequence for the instantaneous energy ratios of multiple consecutive cycles, and uses the difference mean or difference median as an approximation of the slope of change. When this approximation is positive, the processor outputs the compensation ratio coefficient using a piecewise lookup table method, and restricts the compensation ratio coefficient to between 0 and 1.

[0136] Step 450: Adjust the activation threshold based on the compensation ratio coefficient.

[0137] In some embodiments, the processor multiplies the original activation threshold by a compensation scaling factor to obtain an updated activation threshold. For example, if the compensation scaling factor is 0.6, the processor scales the original activation threshold by a factor of 0.6 to relax the triggering conditions. The processor can also generate synchronization assistance instructions based on the updated activation threshold, giving the system the ability to adapt to fatigue states and significantly reducing the missed detection rate during long-term rehabilitation training.

[0138] The static passive activation threshold is a fixed value determined once during the calibration phase, and its benchmark is anchored to the EEG response level of the target user in a conscious state. When the target user enters a fatigue state, the excitability of the cerebral cortex decreases, and the EEG desynchronization amplitude generated by the activation of mirror neurons under the same visual stimulus is significantly weakened. This causes the originally effective neural response to be missed by the system because it is below the static threshold, and the triggering condition is decoupled from the user's actual state. In the embodiments of this specification, the user's awake fatigue state is monitored in real time by the EEG signal of the frontal pole. When a trend of deepening fatigue is detected, the passive activation threshold is dynamically lowered using a compensation proportional coefficient. This allows the triggering threshold to adaptively adjust with the drift of the user's EEG baseline, ensuring that the generation process of the synchronization auxiliary command matches the user's current neural response level. This ensures that the effective mirror activation response generated in a fatigue state can still be accurately identified, effectively reducing the missed detection rate in long-term intervention processes and improving the robustness of neural response detection and the continuity of the overall intervention.

[0139] In some embodiments, when the slope of the instantaneous energy ratio change is positive, the processor determines the compensation ratio coefficient, including: obtaining the difference between the instantaneous energy ratio and a preset resting baseline value; when the difference is within the normal fluctuation range, determining the compensation ratio coefficient as a first constant; when the difference is within the moderate fatigue range, dynamically adjusting the compensation ratio coefficient according to a preset function curve; and when the difference exceeds the extreme fatigue critical point, determining the compensation ratio coefficient as a minimum lower limit constant.

[0140] In some embodiments, the processor performs a difference calculation based on the current instantaneous energy ratio and a preset resting baseline value to obtain the difference.

[0141] The preset resting baseline value is the benchmark value for the energy ratio of the target user in a conscious state, recorded during the calibration phase. The calibration phase refers to the process by which the system initially records and collects relevant parameters such as the energy ratio while the user is in a conscious and resting state.

[0142] In some embodiments, the processor obtains the difference by: reading a preset resting reference value stored in the calibration phase from a storage device; reading the instantaneous energy ratio at the current moment; and subtracting the preset resting reference value from the current instantaneous energy ratio to obtain the difference.

[0143] In some embodiments, the processor performs constant assignment to obtain a compensation ratio coefficient based on the difference being within the normal fluctuation range.

[0144] The normal fluctuation range refers to the range within which the difference is considered to fall within a reasonable physiological fluctuation range. In some embodiments, the normal fluctuation range can be determined based on the statistical results of the instantaneous energy ratio of multiple resting windows during the calibration phase, for example, using the natural fluctuation range near the preset resting baseline value TAR0 as the normal fluctuation range.

[0145] The first constant refers to the neutral multiplier that represents the condition for not performing threshold attenuation compensation. For example, the first constant could be 1.0.

[0146] In some embodiments, the processor compares the difference with the upper and lower boundaries of the normal fluctuation range; when the difference is between the upper and lower boundaries, the processor directly assigns the compensation ratio coefficient to the first constant 1.0.

[0147] In some embodiments, the processor performs curve mapping to dynamically adjust the compensation ratio coefficient based on the difference being in the moderate fatigue range.

[0148] The moderate fatigue range determines whether a user has entered the range of mild to moderate fatigue. In some embodiments, the lower boundary of the moderate fatigue range is higher than the upper boundary of the normal fluctuation range, and the upper boundary of the moderate fatigue range is not higher than the extreme fatigue threshold. It can be preset based on the difference distribution of fatigue rise phases during historical interventions.

[0149] The preset function curve can be a nonlinear decay mathematical model with smooth transition characteristics. For example, the preset function curve can be a Sigmoid function curve.

[0150] In some embodiments, the processor substitutes the difference into a preset Sigmoid function equation during continuous control cycles to obtain a compensation ratio coefficient that decreases smoothly as the difference increases.

[0151] In some embodiments, the processor dynamically adjusting the compensation ratio coefficient may further include: pre-generating a lookup table corresponding to a preset function curve; when the difference falls into the moderate fatigue range, reading the corresponding coefficient based on the difference hit or interpolation, so as to reduce the computational overhead caused by real-time exponential calculation.

[0152] In some embodiments, the processor performs a lower limit protection assignment to obtain a compensation ratio coefficient based on the difference exceeding the extreme fatigue threshold.

[0153] The extreme fatigue threshold refers to the upper limit of the safe judgment for determining whether a user has entered a state of extreme drowsiness. In some embodiments, the extreme fatigue threshold can be preset based on the statistical results of the difference between the calibration phase and the historical monitoring phase, and is higher than the upper boundary of the moderate fatigue range.

[0154] The minimum lower limit constant refers to the limit of safety assurance value that allows the threshold to be relaxed. For example, the minimum lower limit constant can be 0.5.

[0155] In some embodiments, when the difference exceeds the extreme fatigue threshold, the processor will forcibly assign the compensation ratio coefficient to a minimum lower limit constant to limit the threshold relaxation range.

[0156] In some embodiments, the processor may first calculate the compensation ratio coefficient according to a preset function curve, and then perform lower limit clipping; when the difference exceeds the extreme fatigue critical point and the calculation result is less than the minimum lower limit constant, the processor outputs the minimum lower limit constant as the final compensation ratio coefficient.

[0157] In some embodiments, the processor can determine the compensation ratio coefficient in other ways. For example, the processor can map the difference using a piecewise monotonically decreasing function, where a first constant is output in the normal fluctuation range, a decreasing value is output in the moderate fatigue range, and a minimum lower limit constant is output beyond the extreme fatigue critical point.

[0158] Some embodiments in this specification achieve both fatigue adaptability and noise resistance safety by partitioning the difference and using smooth attenuation in the moderate fatigue range, continuously adjusting the compensation ratio coefficient, and limiting the relaxation range of the activation threshold with the help of a minimum lower limit constant.

[0159] In some embodiments, during the process of driving the flexible exoskeleton to assist the target user's target limb in performing the target action, the processor can also determine whether to switch the control mode of the flexible exoskeleton to the compliant control mode based on the dynamic deviation between the actual motion trajectory of the target limb following the motion and the target motion trajectory of the target action. Figure 5 This is an exemplary flowchart illustrating the control modes of switching a flexible exoskeleton according to some embodiments of this specification. In some embodiments, process 500 may be executed by a processor (such as processor 120). Figure 5 As shown, process 500 includes the following steps.

[0160] Step 510: Based on the inertial measurement unit, the actual motion trajectory of the target limb following the motion is collected in real time.

[0161] An inertial measurement unit (IMU) is a sensor unit installed inside a flexible exoskeleton to output information on acceleration, angular velocity, and attitude changes during the movement of a target limb. For example, an IMU can be a six-axis IMU or a nine-axis IMU.

[0162] The actual motion trajectory refers to the curve of the actual position and angle change of the target limb in three-dimensional space under the traction of the exoskeleton torque.

[0163] In some embodiments, after the synchronous auxiliary instruction begins execution, the processor reads the acceleration and angular velocity data output by the six-axis inertial measurement unit set inside the flexible exoskeleton, and obtains the three-dimensional coordinates and Euler angle sequence of the target limb joint in real time through high-frequency integration and attitude calculation, and assembles the three-dimensional coordinates and Euler angles into the actual motion trajectory in time sequence.

[0164] In some embodiments, the processor can also perform sensor fusion based on data collected by multiple inertial measurement units (IMUs) and combine it with joint length constraints or structural parameters of a flexible exoskeleton to reconstruct the spatial pose of the target limb at each sampling time, thereby obtaining the actual motion trajectory. Multiple IMUs can be respectively positioned at corresponding locations of adjacent limb segments to improve the stability of trajectory calculation.

[0165] Step 520: Determine the dynamic deviation between the actual motion trajectory and the target motion trajectory of the target action.

[0166] The target motion trajectory refers to the curve of the ideal position and angle change that the target action in the target video segment should achieve under the same time sequence.

[0167] Dynamic deviation value refers to an indicator that quantifies the degree to which the actual trajectory deviates from or lags behind the target trajectory.

[0168] In some embodiments, the processor retrieves the target trajectory point at the same moment from the target motion trajectory of the target action according to the timestamp corresponding to the current synchronization assistance instruction, calculates the Euclidean distance between the actual trajectory point and the target trajectory point, and uses the Euclidean distance as the dynamic deviation value.

[0169] In some embodiments, the processor can also calculate the position deviation, attitude angle deviation, and timing lag separately, and perform weighted fusion of the position deviation, attitude angle deviation, and timing lag to obtain a dynamic deviation value. The timing lag is used to characterize the degree of lag in the target limb's following of the target movement rhythm.

[0170] In some embodiments, the dynamic deviation value can also be determined in other ways. For example, the processor performs dynamic time warping calculations on the actual motion trajectory and the target motion trajectory within a sliding time window, and uses the warped distance as the dynamic deviation value.

[0171] Step 530: When the dynamic deviation value exceeds the preset resistance threshold, switch the control mode of the flexible exoskeleton to the compliant control mode and update the torque output parameters for the next synchronous auxiliary command execution.

[0172] A preset resistance threshold refers to the limit of tolerance deviation that a user can tolerate due to physical resistance caused by discomfort. In some embodiments, the preset resistance threshold can be pre-set based on the target user's comfort calibration results, historical resistance event statistics, and the safe traction range corresponding to the target movement, and can be configured separately for different target limbs or training movements. For example, the target user can be organized to complete training movements of different intensities and amplitudes. By collecting the user's surface electromyography signals, movement acceleration fluctuations, and subjective discomfort scores during each movement, the dynamic deviation value distribution when the user exhibits physical resistance behavior can be statistically analyzed. Combined with the maximum safe range of motion of the target training movement, distribution feature analysis and safety boundary setting algorithms are used to finally determine the preset resistance threshold for subsequent control processes.

[0173] Control mode refers to the control strategy state adopted by the flexible exoskeleton when outputting driving torque. Control mode is used to limit the mechanical response of the actuators during the traction, following, or release phases. For example, in traction control mode, the actuators actively output driving force to assist the wearer in lifting their legs; in following control mode, the actuators adjust the output in real time according to the wearer's movements to follow the user's steps; in release control mode, the actuators do not actively output torque, allowing the exoskeleton to move freely.

[0174] Adaptive control mode refers to the exoskeleton abandoning forced traction and switching to a retreating state that follows the limb's active intentions or releases torque.

[0175] Torque output parameters refer to the physical control variables used to set the peak torque and release time of the driver when generating control commands.

[0176] In some embodiments, the processor continuously compares the dynamic deviation value with a preset resistance threshold. When the dynamic deviation value exceeds the preset resistance threshold, the processor interrupts the current forced traction control and sends a mode switching command to the flexible exoskeleton, causing the flexible exoskeleton to enter compliant control mode. Simultaneously, the processor reduces the peak torque at the next execution of the synchronization assist command and extends the torque settling time to update the torque output parameters.

[0177] In some embodiments, the processor can perform mode switching when the dynamic deviation value exceeds a preset resistance threshold for multiple consecutive sampling periods to avoid accidental switching triggered by a single noise event. After entering the compliant control mode, if the flexible exoskeleton is pneumatically driven, the processor controls the actuator to vent the bellows air pressure; if the flexible exoskeleton is motor-driven, the processor lowers the target current or target torque to the follow-up output range. The torque output parameters of the next synchronization auxiliary command can also be updated in stages according to the threshold exceedance range; the larger the threshold exceedance range, the greater the decrease in peak torque and the longer the release time.

[0178] Some embodiments in this specification form a real-time compliant control mechanism through actual trajectory closed-loop detection, deviation threshold avoidance, and next torque degradation, which reduces traction injury and improves the safety and adaptability of exoskeleton assistance.

[0179] In some embodiments, the processor performs the following process: the dynamic deviation value includes the instantaneous mechanical impedance index; determining the dynamic deviation value between the actual motion trajectory and the target motion trajectory of the target action, including: acquiring the rate of change of angular acceleration output by the inertial measurement unit; acquiring the gradient of the change of backflow air pressure in the pneumatic bellows of the flexible exoskeleton; and performing a weighted fusion calculation of the instantaneous rate of change of angular acceleration and the real-time gradient of the change of backflow air pressure to obtain the instantaneous mechanical impedance index.

[0180] In some embodiments, the processor directly uses the instantaneous mechanical impedance index as the dynamic deviation value at the current moment.

[0181] The instantaneous mechanical impedance index is a quantitative indicator that comprehensively evaluates the intensity of physical confrontation between humans and machines.

[0182] In some embodiments, after obtaining the instantaneous mechanical resistance index, the processor directly uses the index as the dynamic deviation value at the current moment for subsequent resistance threshold comparison.

[0183] In some embodiments, the processor may also output the trajectory position deviation or attitude deviation as auxiliary determination information along with the instantaneous mechanical resistance index, wherein the dynamic deviation value used for comparison with the set resistance threshold is still the instantaneous mechanical resistance index.

[0184] In some embodiments, the processor, based on the output of the inertial measurement unit, performs an operation to acquire the rate of change of angular acceleration to obtain kinematic quantities characterizing motion jerks.

[0185] The rate of change of angular acceleration is the time first derivative (Jerk) of the joint angular acceleration in the actual trajectory, which characterizes the abruptness of limb movement.

[0186] In some embodiments, the processor continuously reads the joint angular velocity data output by the inertial measurement unit, performs a discrete difference on adjacent sampling points to obtain angular acceleration, and then performs a discrete difference on the angular acceleration sequence to obtain the angular acceleration rate of change data sequence.

[0187] In some embodiments, the processor can also directly read the angular acceleration channel data output by the inertial measurement unit and perform time difference on the continuous sampling points to obtain the rate of change of angular acceleration; before or after the difference, the processor can perform low-pass filtering or moving average processing on the data to suppress high-frequency noise.

[0188] In some embodiments, the rate of change of angular acceleration can also be obtained in other ways. For example, the processor first calculates the angular velocity and angular acceleration based on the attitude angle sequence output by the inertial measurement unit, and then obtains the rate of change of angular acceleration through differential calculation.

[0189] In some embodiments, the processor performs an operation to obtain the gradient of the return air pressure change based on the air pressure sampling results within the pneumatic bellows, in order to obtain a quantity characterizing the dynamic rebound.

[0190] Pneumatic bellows are inflatable and deflated actuation chambers in flexible exoskeletons that can apply auxiliary force or resistance to limbs through changes in internal air pressure and generate a monitorable air pressure response.

[0191] The backflow pressure gradient is the instantaneous rate of change of the air pressure value monitored by the sensor inside the pneumatic bellows over time, characterizing the dynamic characteristics of the sudden increase in air chamber pressure caused by active obstruction by the limbs.

[0192] In some embodiments, the processor acquires the absolute air pressure value of the air chamber in real time through a high-frequency air pressure sensor installed on the air intake and exhaust path of the pneumatic bellows, and performs a discrete difference on the air pressure values ​​at adjacent times to obtain the gradient of the return air pressure change.

[0193] In some embodiments, the processor can also perform local linear fitting on multiple consecutive pressure sampling points and use the fitting slope as the gradient of the return pressure change to reduce the impact of single-point fluctuations on the gradient calculation results.

[0194] In some embodiments, the return pressure gradient can also be obtained in other ways. For example, the processor calculates the return pressure gradient based on the ratio of the pressure increment to the time interval within a fixed time window.

[0195] In some embodiments, the processor performs a weighted fusion calculation based on the instantaneous rate of change of angular acceleration and the real-time backflow pressure gradient to obtain the instantaneous mechanical impedance index. In some embodiments, the weights corresponding to the instantaneous rate of change of angular acceleration and the real-time backflow pressure gradient can be preset based on resistance samples and normal following samples collected during the calibration phase, or based on empirical parameter tables for different motion phases.

[0196] In some embodiments, the processor can also adaptively adjust the weighting coefficients according to the motion phase or auxiliary mode; for example, increasing the weight corresponding to the rate of change of angular acceleration during the rapid oscillation phase, increasing the weight corresponding to the gradient of change of return air pressure during the high-resistance return phase, and then outputting the instantaneous mechanical impedance index based on the adjusted weights.

[0197] In some embodiments, the instantaneous mechanical impedance index can also be obtained in other ways. For example, the processor maps the rate of change of angular acceleration and the gradient of change of backflow pressure to the instantaneous mechanical impedance index based on a two-dimensional lookup table established during the calibration phase.

[0198] Some embodiments in this specification integrate motion jerking and air pressure rebound into an instantaneous mechanical impedance index, thereby improving the accuracy of active resistance recognition and the sensitivity of safe retreat triggering by simultaneously characterizing kinematic and dynamic deviations.

[0199] In some embodiments, when the processor triggers the compliant control mode, it sends a visual backtracking synchronization signal to the display terminal; the visual backtracking synchronization signal controls the display terminal to reduce the playback speed of the target video segment and backtrack the playback frames so that the target motion trajectory remains synchronized with the actual motion trajectory.

[0200] In some embodiments, when the processor triggers the compliant control mode, it sends a visual back-off synchronization signal to the display terminal.

[0201] Visual back synchronization signal refers to the control command sent to the player to change the current video timeline or playback frame rate.

[0202] In some embodiments, the processor sending the visual backtracking synchronization signal includes: real-time monitoring of the control mode flag of the flexible exoskeleton; when the processor detects that the control mode has switched to compliant control mode, reading the physical coordinate data corresponding to the current dwell position and recording the dwell time stamp; the processor encapsulates the physical coordinate data, the dwell time stamp, and the target video segment identifier into a control data packet and then sends it to the display terminal. The display terminal may be, for example, a tablet computer, a training display screen, or a head-mounted display device.

[0203] In some embodiments, the processor sending the visual backtracking synchronization signal may further include: determining whether to enter the compliance control mode based on the interaction resistance and the motion speed; when the interaction resistance is greater than a preset threshold and the motion speed is less than a speed threshold, the processor maps the current position to the corresponding frame number or time axis position in the target video segment, generates a visual backtracking synchronization signal containing the backtracking offset and the target playback magnification, and sends it to the display terminal.

[0204] In some embodiments, the visual back-off synchronization signal can also be generated and sent in other ways. For example, the processor directly determines the back-off frame number and deceleration parameters based on a pre-established position and frame index lookup table, and sends it to the display terminal in the form of a control message.

[0205] In some embodiments, the processor controls the display terminal to reduce the playback speed of the target video segment and rewind the playback frames based on the visual backtracking synchronization signal, so as to keep the target motion trajectory synchronized with the actual motion trajectory.

[0206] In some embodiments, the processor controls the video player in the display terminal to interrupt the current normal playback progress after receiving the visual backtracking synchronization signal, backtracks the playback position to the moment corresponding to the physical stop point according to the pause timestamp, and reduces the playback magnification to a preset low speed magnification. The processor then continuously acquires the actual motion trajectory data of the exoskeleton and controls the display terminal to restore the corresponding video frames according to the trajectory advancement amount, so that the target motion trajectory in the target video is synchronized with the actual traction trajectory of the exoskeleton.

[0207] In some embodiments, the processor controlling the display terminal to reduce playback speed and rewind playback frames may further include: first performing rewind positioning based on the mapping relationship between physical coordinates and keyframes of the video trajectory, and then switching to frame-by-frame release or low frame rate playback mode. Each time the processor receives a new position increment or angle increment, it controls the display terminal to advance the corresponding number of playback frames, thereby driving the target video progress according to physical progress.

[0208] In some embodiments, playback synchronization can also be achieved in other ways. For example, the processor controls the display terminal to pause continuous playback and advance the video frame by frame according to the sampling increment of the actual motion trajectory using a lookup table method.

[0209] In some embodiments of this specification, after triggering the compliant control mode, a visual backtracking synchronization signal is sent, and the screen is backtracked and playback is slowed down accordingly. By locking the synchronization between the video progress and the actual traction progress, the visual-motor perception misalignment can be reduced and the consistency of the training loop can be improved.

[0210] Figure 6 This is an exemplary block diagram of a motor development promotion system according to some embodiments of this specification. Figure 6As shown, the motor development promotion system 600 includes a video playback module 610, a data acquisition module 620, an energy change rate determination module 630, an auxiliary instruction generation module 640, and a drive module 650.

[0211] In some embodiments, the video playback module 610 is configured to play a target video clip to a target user, the target video clip containing a target action recorded from a preset perspective.

[0212] The acquisition module 620 is configured to acquire a first electroencephalogram (EEG) signal of a target cortical region of the target user during the playback of the target video segment.

[0213] The energy change rate determination module 630 is configured to determine the energy change rate of the first EEG signal based on the first EEG signal.

[0214] The auxiliary instruction generation module 640 is configured to generate a synchronization auxiliary instruction when the energy change rate exceeds the activation threshold.

[0215] The drive module 650 is configured to send the synchronous assistance command to the flexible exoskeleton to drive the flexible exoskeleton to assist the target user's target limb in performing the target action.

[0216] In some embodiments, the system further includes a reconstruction module 660. The reconstruction module 660 is configured to acquire inertial data of the flexible exoskeleton in real time; and reconstruct the first EEG signal based on the inertial data and the first EEG signal.

[0217] In some embodiments, the reconstruction module 660 is further configured to: reconstruct the first EEG signal based on a dual-thread architecture, the dual-thread architecture including a background filtering thread and a foreground denoising thread; when the linear acceleration in the inertial data exceeds an acceleration threshold, the background filtering thread updates the spatial denoising matrix based on the first EEG signal and the inertial data; the foreground denoising thread receives the first EEG signal and reconstructs the first EEG signal based on the first EEG signal and the spatial denoising matrix.

[0218] In some embodiments, the system further includes a threshold adjustment module 670. The threshold adjustment module 670 is configured to: acquire a second EEG signal from the frontal pole region; based on the second EEG signal, extract a first energy from the second EEG signal in a first monitoring frequency band and a second energy from a second monitoring frequency band, wherein the frequency of the first monitoring frequency band is lower than that of the second monitoring frequency band; determine the instantaneous energy ratio of the first energy to the second energy; when the slope of the change in the instantaneous energy ratio is positive, determine a compensation ratio coefficient; and adjust the activation threshold based on the compensation ratio coefficient.

[0219] In some embodiments, the system further includes a control mode switching module 680. During the process of driving the flexible exoskeleton to assist the target user's target limb in performing the target action, the control mode switching module 680 is configured to: based on an inertial measurement unit (IMU), collect in real time the actual motion trajectory of the target limb following the motion, wherein the IMU is disposed inside the flexible exoskeleton; determine the dynamic deviation value between the actual motion trajectory and the target motion trajectory of the target action; when the dynamic deviation value exceeds a preset resistance threshold, switch the control mode of the flexible exoskeleton to a compliant control mode and update the torque output parameters for the next synchronous assistance command execution.

[0220] This specification also provides a motor development promotion device, the device including at least one storage medium and at least one processor; the at least one storage medium is used to store computer instructions; the at least one processor is used to execute the computer instructions to implement the motor development promotion method as shown in any of the above embodiments.

[0221] The basic concepts have been described above. Obviously, for those skilled in the art, the detailed disclosure above is merely illustrative and does not constitute a limitation of this specification. Although not explicitly stated herein, those skilled in the art may make various modifications, improvements, and corrections to this specification. Such modifications, improvements, and corrections are suggested in this specification and therefore remain within the spirit and scope of the exemplary embodiments described herein.

[0222] Finally, it should be understood that the embodiments described in this specification are merely illustrative of the principles of the embodiments described herein. Other variations may also fall within the scope of this specification. Therefore, alternative configurations of the embodiments described herein are intended to be illustrative rather than limiting, and should be considered consistent with the teachings of this specification. Accordingly, the embodiments described herein are not limited to those explicitly introduced and described herein.

Claims

1. A method for promoting motor development, characterized in that, The method includes: Play a target video clip to a target user, the target video clip containing target actions recorded from a preset perspective; During the playback of the target video segment, the first electroencephalogram (EEG) signal of the target user's target cortical region is acquired; Based on the first EEG signal, determine the energy change rate of the first EEG signal; When the energy change rate exceeds the activation threshold, a synchronization assistance command is generated; The synchronous assistance command is sent to the flexible exoskeleton to drive the flexible exoskeleton to assist the target user's target limb in performing the target action.

2. The method for promoting motor development as described in claim 1, characterized in that, Before determining the rate of energy change, the method further includes: Real-time acquisition of inertial data from the flexible exoskeleton; Based on the inertial data and the first EEG signal, the first EEG signal is reconstructed.

3. The method for promoting motor development as described in claim 2, characterized in that, The reconstructing of the first EEG signal based on the inertial data and the first EEG signal includes: The first EEG signal is reconstructed based on a dual-thread architecture, which includes a background filtering thread and a foreground denoising thread. When the linear acceleration in the inertial data exceeds the acceleration threshold, the background filtering thread updates the spatial denoising matrix based on the first EEG signal and the inertial data. The foreground denoising thread receives the first EEG signal and reconstructs the first EEG signal based on the first EEG signal and the spatial denoising matrix.

4. The method for promoting motor development as described in claim 1, characterized in that, The method further includes: Acquire the second EEG signal from the frontal pole region; Based on the second EEG signal, the first energy of the second EEG signal in the first monitoring frequency band and the second energy in the second monitoring frequency band are extracted respectively, wherein the frequency of the first monitoring frequency band is lower than that of the second monitoring frequency band. Determine the instantaneous energy ratio between the first energy and the second energy; When the slope of the change in the instantaneous energy ratio is positive, the compensation ratio coefficient is determined; The activation threshold is adjusted based on the compensation ratio coefficient.

5. The method for promoting motor development as described in claim 1, characterized in that, In the process of driving the flexible exoskeleton to assist the target user's target limb in performing the target action, the method further includes: Based on the inertial measurement unit, the actual motion trajectory of the target limb following the movement is collected in real time. The inertial measurement unit is set inside the flexible exoskeleton. Determine the dynamic deviation value between the actual motion trajectory and the target motion trajectory of the target action; When the dynamic deviation value exceeds the preset resistance threshold, the control mode of the flexible exoskeleton is switched to the compliant control mode, and the torque output parameters are updated for the next synchronous auxiliary command execution.

6. A motor development promotion system, characterized in that, The system includes: The video playback module is configured to play a target video clip to a target user, the target video clip containing a target action recorded from a preset perspective; The acquisition module is configured to acquire first electroencephalogram (EEG) signals of the target cortical region of the target user during the playback of the target video segment. The energy change rate determination module is configured to determine the energy change rate of the first EEG signal based on the first EEG signal; The auxiliary instruction generation module is configured to generate a synchronization auxiliary instruction when the energy change rate exceeds an activation threshold. The drive module is configured to send the synchronous assistance command to the flexible exoskeleton to drive the flexible exoskeleton to assist the target user's target limb in performing the target action.

7. The motor development promotion system as described in claim 6, characterized in that, The system also includes a reconfiguration module, which is configured to: Real-time acquisition of inertial data from the flexible exoskeleton; Based on the inertial data and the first EEG signal, the first EEG signal is reconstructed.

8. The motor development promotion system as described in claim 7, characterized in that, The reconstruction module is further configured as follows: The first EEG signal is reconstructed based on a dual-thread architecture, which includes a background filtering thread and a foreground denoising thread. When the linear acceleration in the inertial data exceeds the acceleration threshold, the background filtering thread updates the spatial denoising matrix based on the first EEG signal and the inertial data. The foreground denoising thread receives the first EEG signal and reconstructs the first EEG signal based on the first EEG signal and the spatial denoising matrix.

9. The motor development promotion system as described in claim 6, characterized in that, The system further includes a threshold adjustment module, which is configured to: Acquire the second EEG signal from the frontal pole region; Based on the second EEG signal, the first energy of the second EEG signal in the first monitoring frequency band and the second energy in the second monitoring frequency band are extracted respectively, wherein the frequency of the first monitoring frequency band is lower than that of the second monitoring frequency band. Determine the instantaneous energy ratio between the first energy and the second energy; When the slope of the change in the instantaneous energy ratio is positive, the compensation ratio coefficient is determined; The activation threshold is adjusted based on the compensation ratio coefficient.

10. The motor development promotion system as described in claim 6, characterized in that, The system also includes a control mode switching module, which is configured to: During the process of driving the flexible exoskeleton to assist the target user's target limb in performing the target action, the control mode switching module is configured to: Based on the inertial measurement unit, the actual motion trajectory of the target limb following the movement is collected in real time. The inertial measurement unit is set inside the flexible exoskeleton. Determine the dynamic deviation value between the actual motion trajectory and the target motion trajectory of the target action; When the dynamic deviation value exceeds the preset resistance threshold, the control mode of the flexible exoskeleton is switched to the compliant control mode, and the torque output parameters are updated for the next synchronous auxiliary command execution.

Images