Control method and system for lower limb assistive device based on multi-modal physiological signal feedback

Abstract

The application provides a lower limb auxiliary device control method and system based on multi-modal physiological signal feedback, and belongs to the technical field of rehabilitation engineering. The method synchronously determines muscle-bone movement state indexes, neural cognitive state indexes and psychological emotional state indexes by collecting multi-modal physiological signals such as electroencephalogram, electrocardiogram, electromyogram and near-infrared spectrum of a target object, and realizes three-dimensional integrated perception of neural cognition, psychology and emotion and muscle-bone movement. The three-dimensional indexes are weighted and fused to obtain a global training motivation strength, which is mapped to all hardware parameters such as suspension weight reduction, a treadmill, an exoskeleton, functional electrical stimulation, vagus nerve stimulation and a virtual reality scene, so that closed-loop regulation and control of psychological and movement collaborative rehabilitation is realized. Meanwhile, a multi-index comprehensive risk value is constructed based on electromyographic spasm peak value, heart rate fluctuation and oxygenated hemoglobin concentration gradient, so that the training process is classified and protected.

Description

Technical Field

[0001] This invention relates to the field of assistive device technology, and more specifically, to a control method and system for a lower limb assistive device based on multimodal physiological signal feedback, achieving coordinated psychological and motor rehabilitation and possessing multi-index graded safety protection functions. Background Technology

[0002] Stroke and spinal cord injury patients often have lower limb motor dysfunction, so clinical practice needs to carry out motor rehabilitation, neuromodulation, and psychological intervention simultaneously.

[0003] The existing lower limb assistive movement rehabilitation equipment operates in an isolated open-loop manner, with each functional module belonging to an independent control system. The suspension weight reduction device, treadmill, lower limb exoskeleton, functional electrical stimulation module, vagus nerve electrical stimulation module, and virtual reality scene training module are all independent control systems without a unified collaborative scheduling logic. Furthermore, the training process is prone to problems such as lower limb muscle compensation, joint stress imbalance, and increased psychological stress. Summary of the Invention

[0004] In view of this, the present invention provides a lower limb assistive device control method and system based on a three-dimensional integrated perception and coordinated regulation architecture of neurocognition, psychological emotion and musculoskeletal movement, which realizes psychological and motor coordinated rehabilitation and has multi-index graded safety protection functions.

[0005] One aspect of the present invention provides a lower limb assistive device control method based on multimodal physiological signal feedback, comprising: acquiring multiple physiological characteristic signals of a target object; performing feature extraction and weighted processing on the multiple physiological characteristic signals to simultaneously determine musculoskeletal motion state indicators, neurocognitive state indicators, and psychological and emotional state indicators of the target object; determining multiple device control parameters based on the musculoskeletal motion state indicators, neurocognitive state indicators, and psychological and emotional state indicators, wherein the multiple device control parameters include suspension weight reduction ratio, treadmill speed incline, exoskeleton hip and knee joint assist torque, functional electrical stimulation parameters, vagus nerve stimulation parameters, and virtual reality scene parameters; and controlling the lower limb assistive device based on the multiple device control parameters. The musculoskeletal motion state indicators include at least muscle fatigue, the neurocognitive state indicators include at least motor intention intensity and cognitive load, and the psychological and emotional state indicators include at least a comprehensive depression index.

[0006] According to an embodiment of the present invention, multiple device control parameters are determined based on the musculoskeletal motion state index, the neurocognitive state index, and the psychological and emotional state index, including: performing weighted fusion processing on the musculoskeletal motion state index, the neurocognitive state index, and the psychological and emotional state index to obtain a global training incentive intensity; mapping the global training incentive intensity to at least one of the following: suspension weight reduction ratio, treadmill speed incline, exoskeleton hip and knee joint assist torque, functional electrical stimulation parameters, vagus nerve stimulation parameters, and virtual reality scene parameters to obtain a mapping result; and determining the multiple device control parameters based on the mapping result.

[0007] According to an embodiment of the present invention, the method further includes: when the comprehensive depression index in the psychological and emotional state indicators exceeds a preset emotional threshold, triggering a virtual reality scene switching event and generating a synchronization time window; wherein, the synchronization time window is used to limit the adjustment period of the vagus nerve stimulation parameters, so as to achieve temporal synergy between psychological visual intervention and neural regulation.

[0008] According to an embodiment of the present invention, the above-described feature extraction and weighted processing of the plurality of physiological feature signals to simultaneously determine the musculoskeletal motor state indicators, neurocognitive state indicators, and psychological and emotional state indicators of the target object includes: determining the intensity of the target object's motor intention by analyzing the target object's brain activity and muscle activity based on the plurality of physiological feature signals; determining the target object's cognitive load by analyzing the target object's mental workload and metabolic burden based on the plurality of physiological feature signals; determining the target object's muscle fatigue by analyzing the temporal and frequency domain characteristics of the target object's electromyographic signals based on the plurality of physiological feature signals; determining the target object's comprehensive depression index by analyzing the multiple emotional states corresponding to each of the multiple physiological dimensions of the target object based on the plurality of physiological feature signals; and determining the target object's musculoskeletal motor state indicators, neurocognitive state indicators, and psychological and emotional state indicators based on the intensity of motor intention, the cognitive load, the cognitive load, and the comprehensive depression index.

[0009] According to an embodiment of the present invention, the aforementioned plurality of physiological characteristic signals include electroencephalogram (EEG) signals and electromyogram (EMG) signals. The determination of the target object's motor intention intensity based on these plurality of physiological characteristic signals, by analyzing the brain activity and muscle activity of the target object, includes: estimating the power spectral density of the EEG signals to obtain an EEG power spectral signal; obtaining the total power of the motor brain region based on the sum of the integrated power of the EEG power spectral signal in a first target frequency band and the integrated power of the EEG power spectral signal in a second target frequency band; truncating and calculating the root mean square (RMS) of the EMG signals using a pre-activation window to determine the muscle pre-activation RMS, wherein the window termination time of the pre-activation window is the current sampling time; and weighting the total power of the motor brain region and the muscle pre-activation RMS to obtain the motor intention intensity.

[0010] According to an embodiment of the present invention, the aforementioned plurality of physiological characteristic signals include electroencephalogram (EEG) signals and near-infrared spectral signals. The determination of the cognitive load of the target object based on these plurality of physiological characteristic signals by analyzing the mental and metabolic burden of the target object includes: estimating the power spectral density of the EEG signals to obtain an EEG power spectral signal; obtaining an EEG power ratio based on the ratio of the integrated power of the EEG power spectral signal in a third target frequency band to the integrated power of the EEG power spectral signal in a fourth target frequency band; determining the oxyhemoglobin content based on the near-infrared spectral signal; and weighting the EEG power ratio and the oxyhemoglobin content to obtain the cognitive load.

[0011] According to an embodiment of the present invention, the aforementioned plurality of physiological characteristic signals include electromyography (EMG) signals. The determination of muscle fatigue of the target object based on these plurality of physiological characteristic signals by analyzing the time-domain and frequency-domain characteristics of the EMG signals includes: estimating the power spectral density of the EMG signals to obtain an EMG power spectral signal; determining an EMG frequency domain index based on the median frequency value of the EMG power spectral signal; performing root mean square calculation on the EMG signals using a sliding window to obtain an EMG time domain index; and weighting the EMG frequency domain index and the EMG time domain index to obtain the muscle fatigue degree.

[0012] According to an embodiment of the present invention, the aforementioned multiple physiological characteristic signals include electroencephalogram (EEG) signals, electrocardiogram (ECG) signals, and near-infrared spectral signals. The determination of the comprehensive depression index of the target object based on these multiple physiological characteristic signals, by analyzing multiple emotional states corresponding to multiple physiological dimensions of the target object, includes: determining the frontal lobe alpha wave asymmetry index based on the EEG signals; determining the heart rate variability sample entropy based on the ECG signals; determining the oxyhemoglobin content based on the near-infrared spectral signals; and weighting the frontal lobe alpha wave asymmetry index, the heart rate variability sample entropy, and the oxyhemoglobin content to obtain the comprehensive depression index.

[0013] According to an embodiment of the present invention, the method further includes: determining the peak value of electromyography (EMG), heart rate fluctuation, and oxyhemoglobin concentration decrease gradient based on the electromyography (EMG) signal, electrocardiogram (ECG) signal, and near-infrared spectral signal of the target object, respectively; determining a first risk indicator, a second risk indicator, and a third risk indicator based on the peak value of EMG, the heart rate fluctuation, and the oxyhemoglobin concentration decrease gradient, respectively; weighting the first risk indicator, the second risk indicator, and the third risk indicator to obtain a comprehensive risk value; and performing graded control of the lower limb assistive device according to a plurality of preset safety threshold intervals based on the comprehensive risk value, including: reducing the training intensity when the comprehensive risk value is in the first threshold interval; initiating a neuro-soothing intervention and lowering the training parameters when the comprehensive risk value is in the second threshold interval; and executing a shutdown protection when the comprehensive risk value is in the third threshold interval.

[0014] Another aspect of the present invention provides a lower limb assistive device control system based on multimodal physiological signal feedback, comprising: a signal acquisition module configured to acquire multiple physiological characteristic signals of a target object; a three-dimensional index determination module configured to perform feature extraction and weighted processing on the multiple physiological characteristic signals, and simultaneously determine the musculoskeletal motion state index, neurocognitive state index, and psychological and emotional state index of the target object, thereby achieving integrated perception of the three dimensions of musculoskeletal motion, neurocognition, and psychological and emotional states; a collaborative regulation parameter determination module configured to determine multiple device control parameters based on the musculoskeletal motion state index, neurocognitive state index, and psychological and emotional state index, so as to achieve closed-loop regulation of psychological and motor collaborative rehabilitation, wherein the multiple device control parameters include suspension weight reduction ratio, treadmill speed incline, exoskeleton hip and knee joint assist torque, functional electrical stimulation parameters, vagus nerve stimulation parameters, and virtual reality scene parameters; and a device control module configured to control the lower limb assistive device based on the multiple device control parameters; wherein the musculoskeletal motion state index includes at least muscle fatigue, the neurocognitive state index includes at least motor intention intensity and cognitive load, and the psychological and emotional state index includes at least a comprehensive depression index.

[0015] According to an embodiment of the present invention, a lower limb assistive device control method based on multimodal physiological signal feedback includes: Step 1: Multimodal physiological signal acquisition—acquiring EEG signals, ECG signals, EMG signals, and near-infrared spectral signals of the target object; Step 2: Synchronous determination of three-dimensional state indicators—based on the multimodal physiological signals, synchronously determining: musculoskeletal motor state indicators: at least including muscle fatigue (calculated based on median frequency and root mean square of EMG); neurocognitive state indicators: at least including motor intention intensity (based on EEG μ / β band power) and cognitive load (based on EEG θ / α band power); psychological and emotional state indicators: at least including a comprehensive depression index (based on EEG frontal lobe α wave asymmetry index, heart rate variability sample entropy, and near-infrared oxygenated hemoglobin). (Protein content); Step 3: Co-regulation parameter generation—The three-dimensional state indicators are weighted and fused to obtain the global training incentive intensity, and mapped to the suspension weight reduction ratio, treadmill speed incline, exoskeleton hip and knee joint assist torque, functional electrical stimulation parameters, vagus nerve stimulation parameters, and virtual reality scene parameters; Step 4: Graded safety protection—Based on the peak value of electromyography spasm, heart rate fluctuation, and the gradient of oxygenated hemoglobin concentration decrease, a multi-indicator comprehensive risk value is constructed, and the training process is graded and controlled according to the preset three-level safety threshold (mild load reduction, early warning intervention, and shutdown protection); Step 5: Closed-loop feedback control—Based on the above parameters, the lower limb assistive device is controlled in a closed loop, and physiological signals are continuously collected during the training process to achieve dynamic adjustment.

[0016] According to embodiments of the present invention, by collecting multiple physiological characteristic signals of the target object, and determining physiological indicators such as musculoskeletal movement state indicators, neurocognitive state indicators, and psychological and emotional state indicators such as movement intention intensity, cognitive load, muscle fatigue, and comprehensive depression index based on the multiple physiological characteristic signals, multiple device control parameters can be determined based on the above physiological indicators, and the lower limb assistive device can be controlled based on the multiple device control parameters. This achieves integrated perception and coordinated regulation of the three dimensions of neurocognition, psychological and emotional, and musculoskeletal movement, breaking through the single mode of traditional rehabilitation that only targets muscle training. Attached Figure Description

[0017] The above and other objects, features and advantages of the present invention will become clearer from the following description of embodiments of the invention with reference to the accompanying drawings.

[0018] Figure 1 A flowchart of a lower limb assistive device control method based on multimodal physiological signal feedback according to an embodiment of the present invention is shown.

[0019] Figure 2 A schematic diagram illustrating the adjustment of VR scene parameters and VNS electrical stimulation parameters according to an embodiment of the present invention is shown.

[0020] Figure 3 A schematic diagram of the complete closed-loop control process according to an embodiment of the present invention is shown.

[0021] Figure 4 An architecture diagram of a lower limb assistive device control system based on multimodal physiological signal feedback according to an embodiment of the present invention is shown.

[0022] Figure 5 A block diagram of a lower limb assistive device control system based on multimodal physiological signal feedback according to an embodiment of the present invention is shown. Detailed Implementation

[0023] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the invention. In the following detailed description, numerous specific details are set forth to provide a thorough understanding of the embodiments of the invention for ease of explanation. However, it will be apparent that one or more embodiments may be practiced without these specific details. Furthermore, descriptions of well-known structures and techniques are omitted in the following description to avoid unnecessarily obscuring the concept of the invention.

[0024] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the invention. The terms "comprising," "including," etc., as used herein indicate the presence of the above-described features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0025] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein are to be interpreted in a manner consistent with the context of this specification, and not in an idealized or overly rigid way.

[0026] When using expressions such as "at least one of A, B and C", they should generally be interpreted in accordance with the meaning that is commonly understood by those skilled in the art (e.g., "a system having at least one of A, B and C" should include, but is not limited to, a system having A alone, a system having B alone, a system having C alone, a system having A and B, a system having A and C, a system having B and C, and / or a system having A, B and C, etc.).

[0027] In the embodiments of this invention, the collection, updating, analysis, processing, use, transmission, provision, disclosure, and storage of data (e.g., including but not limited to user personal information) comply with relevant laws and regulations, are used for legitimate purposes, and do not violate public order and good morals. In particular, necessary measures have been taken to prevent unauthorized access to user personal information data and to maintain the security of user personal information and network security.

[0028] In the embodiments of the present invention, the user's authorization or consent is obtained before acquiring or collecting the user's personal information.

[0029] Current rehabilitation training primarily focuses on restoring motor function, neglecting the impact of the patient's psychological state (such as depression and anxiety) on rehabilitation outcomes. Studies show that higher patient engagement correlates with better training results. However, most existing equipment uses adaptive control algorithms to assist patients, which allows them to reduce their effort, leading to apathy and inertia, thus diminishing rehabilitation effectiveness. Furthermore, the lack of real-time monitoring and quantitative assessment of the patient's psychological state prevents simultaneous psychological intervention during training.

[0030] Due to the lack of multimodal fusion perception and coordinated regulation, the following problems are prone to occur in the current training process: Lower limb muscle compensation: Due to muscle weakness or insufficient control, patients rely excessively on the healthy side or auxiliary muscle groups during training, resulting in a solidified compensation pattern; Joint stress imbalance: The assistance provided by the exoskeleton or weight-reduction system does not match the patient's actual motor ability, which can easily cause abnormal joint stress; Increased psychological stress: When the training intensity or difficulty does not match the patient's psychological state, it may induce negative emotions such as anxiety and depression, affecting rehabilitation compliance and effectiveness.

[0031] Current safety protection devices are mostly triggered by a single indicator (such as heart rate or peak electromyography spasm), lacking multi-indicator joint early warning and graded protection strategies, which can easily lead to false triggers or missed alarms. In particular, there is a lack of safety protection mechanisms targeting emotional states when patients' psychological stress intensifies.

[0032] In summary, existing technologies lack a lower limb assistive device control method and system capable of integrated perception and coordinated regulation of musculoskeletal movement, neurocognition, and psychological emotions. This makes it impossible to achieve simultaneous psychological and motor rehabilitation, nor can it provide multi-indicator graded safety protection. The purpose of this invention is to solve the above-mentioned technical problems and provide a lower limb assistive device control method and system based on multimodal physiological signal feedback, achieving integrated perception and coordinated regulation of neurocognition, psychological emotions, and musculoskeletal movement, and possessing multi-indicator graded safety protection functions.

[0033] Figure 1 A flowchart of a lower limb assistive device control method based on multimodal physiological signal feedback according to an embodiment of the present invention is shown.

[0034] like Figure 1 As shown, the method includes operations S110~S140.

[0035] During operation S110, multiple physiological characteristic signals of the target object are collected.

[0036] In embodiments of the present invention, the target subject may wear electrodes for electroencephalography (EEG), electrocardiography (ECG), electromyography (EMG), cerebral oxygenation, and vagus nerve stimulation in the ear, wear a lower limb exoskeleton, and have a suspension weight-reduction device attached to the torso. This allows for the simultaneous acquisition of multiple physiological characteristic signals through a multimodal acquisition unit. These signals include electroencephalography (EEG), functional near-infrared spectroscopy (fNIRS), electromyography (EMG), and electrocardiography (ECG). Hardware power frequency filtering and software adaptive noise reduction processing can be sequentially applied to these multiple physiological characteristic signals.

[0037] During operation S120, feature extraction and weighted processing are performed on multiple physiological characteristic signals to simultaneously determine the musculoskeletal motion state indicators, neurocognitive state indicators, and psychological and emotional state indicators of the target object.

[0038] In embodiments of the present invention, the musculoskeletal movement state indicators include at least muscle fatigue, the neurocognitive state indicators include at least the intensity of motor intention and cognitive load, and the psychological and emotional state indicators include at least the comprehensive depression index.

[0039] In embodiments of the present invention, features can be extracted from electroencephalogram (EEG) signals, near-infrared spectral signals, electromyography (EMG) signals, and electrocardiogram (ECG) signals respectively to obtain corresponding EEG feature vectors, EMG feature vectors, and ECG feature vectors.

[0040] Furthermore, cross-modal temporal alignment cosine similarity loss function can be used to complete the timestamp registration of multiple heterogeneous signals. The expression for the loss function is as follows:

[0041] (1).

[0042] in, EEG feature vector; This is the electromyography time-series feature vector; This is the electromyography time-series feature vector; It is the vector 2 norm.

[0043] The reason for selecting EEG, EMG, and ECG signals to construct the constraint benchmark is that the sampling frequency of EEG is 1000 Hz, EMG is 2000 Hz, and ECG is 1000 Hz, and the sampling rate of the three is the largest and the timing drift is the most serious.

[0044] The overall timing registration adopts a chained alignment logic: firstly, by minimizing... The temporal similarity between EEG and EMG signals, and between EEG and ECG signals, was completed, and millisecond-level timestamp alignment of the three high-frequency signals was achieved. Using the unified time axis of the aligned ECG signals as a reference, linear interpolation was used to correct the temporal offset of the near-infrared spectral signals. Finally, complete temporal alignment of the four physiological signals—EEG, near-infrared spectral, EMG, and ECG—was achieved, eliminating the feature fusion distortion problem caused by the misalignment of sampling frequencies of different sensors.

[0045] The extraction process for EEG, EMG, and ECG feature vectors can be as follows: acquiring physiological feature signals; segmented sliding window extraction; segment denoising and artifact removal; single-channel time-domain / frequency-domain feature calculation; feature standardization; concatenation to generate a single-channel one-dimensional feature vector; synchronous temporal alignment input loss function, respectively obtaining... , and The detailed steps are as follows:

[0046] First, EEG feature vectors Generation. The EEG signal was segmented into windows: sampling rate 1000 Hz, using a sliding time window. The original EEG waveforms were segmented with a 50% window overlap. Preprocessing was performed to remove artifacts: bandpass filtering (0.5–45Hz) removed power frequency and baseline drift; the 3σ criterion removed artifacts related to electrooculography (EOG) and head movement. Finally, multi-dimensional features were extracted. Time-domain multi-dimensional features included signal mean, variance, zero-crossing rate, and peak factor; frequency-domain multi-dimensional features included relative power in each frequency band of δ (0.5–4Hz), θ (4–8Hz), α (8–13Hz), and β (13–30Hz); key derived feature multi-dimensional features included frontal alpha asymmetry (FAA). All these features were concatenated into a one-dimensional array in a fixed order, and normalized to a unified dimension using maximum and minimum values ​​to obtain a fixed-dimensional temporal feature vector. .

[0047] Second, electromyographic feature vectors Generation. Windowing of the electromyographic signal: sampling rate 2000 Hz, sliding window. 50% overlap, matching the transient changes of rapid muscle spasms. Preprocessing: Bandpass filtering 20–500Hz, and full-wave rectification fusion with moving average envelope extraction to remove cable motion noise. Extraction of multi-dimensional features: Time-domain multi-dimensional features include root mean square (RMS), average rectified value (ARV), waveform kurtosis, and spasm peak EMG. spasmMulti-dimensional features at the frequency domain level include: Median Frequency (MDF) and Mean Power Frequency (MPF). All time-domain and frequency-domain features are sequentially concatenated and normalized to output an electromyography (EMG) temporal feature vector. .

[0048] Third, ECG feature vectors Generation. Windowing of the ECG signal: sampling rate 1000 Hz, sliding window. 50% overlap, covering the complete heartbeat cycle. Preprocessing: 0.5–40Hz bandpass filtering to remove electromyographic interference and baseline drift; adaptive thresholding to identify R-wave peaks and extract RR interval sequences. Multi-dimensional feature extraction: Temporal multi-dimensional features include the mean and standard deviation of the RR intervals, and instantaneous heart rate (HR); non-linear multi-dimensional features include heart rate variability and sample entropy. The multi-dimensional characteristics of fluctuation include the change in heart rate relative to the baseline. By concatenating and normalizing the heartbeat time-domain and nonlinear entropy features, the electrocardiogram time-series feature vector is obtained. .

[0049] Fourth, standardize and unify the processing. , , Each feature dimension is independently normalized to its minimum and maximum values:

[0050] (2).

[0051] Where x is , , Any one of them, This is the minimum value under the baseline condition. This represents the maximum value under the baseline condition.

[0052] This eliminates differences in amplitude and dimensions, ensuring the cosine similarity loss function. The calculation is unaffected by differences in signal amplitude.

[0053] In operation of S130, multiple device control parameters are determined based on the intensity of motor intention, cognitive load, muscle fatigue, and comprehensive depression index.

[0054] Multiple device control parameters may include the suspension weight reduction ratio, treadmill speed and incline, exoskeleton hip and knee joint assist torque, functional electrical stimulation (FES) parameters, vagus nerve stimulation (VNS) parameters, and virtual reality (VR) adjustment parameters.

[0055] In operation S140, the lower limb assistive device is controlled based on multiple device control parameters.

[0056] Lower limb assistive devices may include suspended weight-reduction devices, treadmills, lower limb exoskeleton robots, functional electrical stimulation modules, and vagus nerve electrical stimulation modules, based on which lower limb assistive training can be achieved for the target subjects.

[0057] The suspended weight reduction device has a built-in high-precision tension sensor; it collects real-time weight-bearing force parameters of the target object's lower limbs, receives instructions from the central controller to adaptively adjust the weight reduction ratio, and reduces the load on the lower limbs to adapt to target objects with different muscle strength injuries; the tension data is transmitted back to the controller to help determine the degree of muscle compensation in the lower limbs.

[0058] The treadmill is equipped with a servo speed control module and an electric incline adjustment module; it provides a foundation for walking rehabilitation exercises, dynamically adjusting walking speed and road incline according to control commands, and working with an exoskeleton to simulate rehabilitation walking in multiple scenarios such as flat ground, uphill, and downhill.

[0059] The lower limb exoskeleton robot has built-in torque servo motors in the hip and knee joints; it outputs joint assist torque in real time to compensate for the missing lower limb muscle strength of the target object; the stroke mode reduces the instantaneous assist gradient to inhibit spasticity, and the spinal cord injury mode outputs steady-state continuous torque to ensure the stability of the walking joints.

[0060] The functional electrical stimulation module is equipped with electrodes attached to the surface of the lower limb muscles; it targets and stimulates the quadriceps femoris, tibialis anterior, and gastrocnemius muscles; in stroke mode, it prioritizes stimulation of antagonistic muscles to relieve muscle spasms, while in spinal cord injury mode, it continuously activates disused and atrophied muscle groups, thus compensating for the shortcomings of exoskeleton mechanical assistance in activating neuromuscular pathways.

[0061] The vagus nerve electrical stimulation module is equipped with a special conductive stimulation electrode for the auricle; it reduces sympathetic nerve excitation through low-frequency electrical modulation of the peripheral vagus nerve, thereby alleviating training stress and depressive mood; and it establishes a synchronous time window with the VR scene to achieve time-series synergy between neural modulation and visual psychological intervention.

[0062] According to an embodiment of the present invention, by collecting multiple physiological characteristic signals of the target object, and determining physiological indicators such as the target object's motor intention intensity, cognitive load, muscle fatigue, and comprehensive depression index based on the multiple physiological characteristic signals, multiple device control parameters can be determined based on the physiological indicators, and the lower limb assistive device can be controlled based on the multiple device control parameters. This realizes the integrated perception and coordinated regulation of the three dimensions of neurocognition, psychological emotion, and musculoskeletal movement, breaking through the single mode of traditional rehabilitation that only targets muscle training.

[0063] According to embodiments of the present invention, feature extraction and weighted processing are performed on the multiple physiological characteristic signals to simultaneously determine the musculoskeletal motor state indicators, neurocognitive state indicators, and psychological and emotional state indicators of the target object. This includes: determining the intensity of the target object's motor intention by analyzing the target object's brain activity and muscle activity based on multiple physiological characteristic signals; determining the target object's cognitive load by analyzing the target object's mental and metabolic burden based on multiple physiological characteristic signals; determining the target object's muscle fatigue by analyzing the time-domain and frequency-domain characteristics of the target object's electromyographic signals based on multiple physiological characteristic signals; determining the target object's comprehensive depression index by analyzing multiple emotional states corresponding to multiple physiological dimensions based on multiple physiological characteristic signals; and determining the target object's musculoskeletal motor state indicators, neurocognitive state indicators, and psychological and emotional state indicators based on the intensity of motor intention, the cognitive load, the cognitive load, and the comprehensive depression index.

[0064] According to an embodiment of the present invention, multiple physiological characteristic signals include electroencephalogram (EEG) signals and electromyogram (EMG) signals. Based on these multiple physiological characteristic signals, the intensity of the target object's motor intention is determined by analyzing the brain activity and muscle activity of the target object. This includes: estimating the power spectral density of the EEG signals to obtain an EEG power spectral signal; obtaining the total power of the motor brain region based on the sum of the integrated power of the EEG power spectral signal in a first target frequency band and the integrated power of the EEG power spectral signal in a second target frequency band; determining the root mean square (RMS) of muscle preactivation by truncating the EMG signals using a preactivation window and calculating the RMS, wherein the window termination time of the preactivation window is the current sampling time; and weighting the total power of the motor brain region and the RMS of muscle preactivation to obtain the intensity of motor intention.

[0065] Intensity of movement intention It can be represented as:

[0066] (3).

[0067] in, This represents the sum of the μ and β band power in the real-time motor brain region. This represents the resting baseline power. The root mean square of muscle pre-activation; and The weights are weights ranging from 0 to 1; the higher the value, the stronger the brain's willingness to actively move. And the weights satisfy... Default value for stroke hemiplegia mode , Using the brain's motor cortex rhythm as the primary diagnostic criterion, and adapting it to the characteristics of brain-muscle pathway disconnection in hemiplegia; spinal cord injury pattern values. , It balances the weights of EEG and EMG, reducing misjudgments caused by spasms.

[0068] The resting state means that the target subject's body is relaxed, without active limb exertion, without imagining lower limb movements, and without deliberately generating the desire to move; no rehabilitation walking tasks are performed, and the exoskeleton and suspension weight-reduction device are only worn for support, without providing assistance or causing the limbs to swing; the subject's eyes are open, and their gaze is fixed on a static cross mark on the screen, minimizing head turning, teeth clenching, and large eye movements; natural and calm breathing is allowed, and high-intensity cognitive mental activities such as mental arithmetic and recall are not performed.

[0069] The quadriceps, tibialis anterior, and gastrocnemius muscles can be selected as target muscle groups for electromyography (EMG) sampling at a rate of 2000 Hz. The pre-activation window is a time window that backs 150 ms from the current moment. This window is in the preparatory stage of movement before significant joint displacement of the limb, excluding EMG signals during large-amplitude limb swinging. Simultaneously, the exoskeleton joint angle signal is used for verification. If significant joint displacement is detected, the pre-activation feature calculation for this window is abandoned to avoid EMG interference caused by passive traction of the limb by the exoskeleton. The raw EMG signal first undergoes 20-500 Hz bandpass filtering, full-wave rectification, and envelope extraction. Samples within the pre-activation window are extracted, and the raw root mean square of the window's EMG signal is calculated.

[0070] Existing methods for motor intention recognition often rely on single EEG or single EMG features, which suffer from pathological adaptation defects: EEG μ / β rhythms alone can only capture subjective brain signals; in stroke-induced hemiplegia, the brain-muscle conduction pathway is impaired, easily resulting in EEG detecting intention but peripheral muscle unresponsiveness, leading to device control mismatch. EMG signals alone can miss subjective brain intentions in severely hemiplegic subjects with weak muscle recruitment, while spastic pseudo-EMG signals from spinal cord injuries are easily misidentified as active motor intentions. Therefore, combining both types of physiological signals can characterize the subjective motor attempts of the motor cortex, allowing for the capture of central motor intentions even in cases of limb muscle paralysis. Furthermore, the root mean square of muscle preactivation characterizes the pre-recruitment state of peripheral lower limb muscles, used to verify EEG output and reflect the actual output capacity of the descending pathway. The motor intention intensity of this invention is not a simple weighted sum of EEG and EMG features, but rather a differentiated weighting based on the injury mechanisms of the two types of diseases: In stroke hemiplegia, the lesion is located in the brain, and the disruption of the brain-muscle pathway is the primary issue; therefore, the weight of EEG features is increased to prioritize the identification of the brain's subjective motor intention, avoiding underestimation of the target subject's active will due to low EMG levels in the affected limb; In spinal cord injury, the target subject's cerebral cortex is often intact, and involuntary muscle spasms are common. Therefore, the weight of EMG pre-activation is appropriately increased to verify peripheral pseudo-signals and suppress misjudgments of intention caused by spasms. This parameter output is a continuous intensity scalar, directly used as the input to the global closed-loop control vector, enabling continuous and smooth adjustment of the entire rehabilitation hardware output based on the target subject's real-time motor intention.

[0071] According to embodiments of the present invention, multiple physiological characteristic signals include electroencephalogram (EEG) signals and near-infrared spectral signals. Based on these multiple physiological characteristic signals, the cognitive load of the target subject is determined by analyzing the mental workload and metabolic burden. This includes: estimating the power spectral density of the EEG signals to obtain an EEG power spectral signal; obtaining an EEG power ratio based on the ratio of the integrated power of the EEG power spectral signal in a third target frequency band to the integrated power of the EEG power spectral signal in a fourth target frequency band; determining the oxyhemoglobin content based on the near-infrared spectral signal; and weighting the EEG power ratio and the oxyhemoglobin content to obtain the cognitive load.

[0072] Cognitive load It can be represented as:

[0073] (4).

[0074] in, and These are the real-time EEG theta and alpha power, respectively. This refers to the content of oxyhemoglobin; The prefrontal cortex is from the same open-eye resting baseline acquisition phase prior to this training session. and The weight coefficients are values ​​ranging from 0 to 1, and the weights satisfy... Default value for stroke hemiplegia mode , Spinal cord injury pattern values , . A high level indicates that the target audience is under too much mental strain and their attention is exhausted. This can automatically reduce the difficulty of virtual reality tasks and slow down the treadmill speed. A lower level indicates a cognitive surplus, and the training complexity can be appropriately increased.

[0075] Most existing lower limb rehabilitation devices only monitor musculoskeletal motor parameters, lacking real-time quantification of cognitive and mental workload during rehabilitation walking. A few solutions only use EEG or near-infrared signals to assess mental load, failing to consider the limitations of each modality, and further neglecting to differentiate them for the pathological characteristics of stroke hemiplegia and spinal cord injury. They also cannot output continuous scalar values ​​for hardware closed-loop regulation. This invention integrates two features from completely different physiological sources: the first... The theta / alpha power ratio in prefrontal cortex EEG is shown to be a factor. Neurophysiology indicates that when cognitive load increases, the power of the theta band, representing mental exertion, rises, while the power of the alpha band, representing the resting inhibitory state, decreases, and the theta / alpha ratio increases accordingly. EEG has high temporal resolution and can capture instantaneous fluctuations in attention during walking, but it is susceptible to interference from motion artifacts and rhythmic distortions caused by brain lesions. The second item... Oxygenated hemoglobin from the prefrontal cortex (fNIRS) is observed as cognitive fatigue accumulates, leading to a decrease in prefrontal metabolic reserves and a reduction in HbO2 compared to the resting baseline. This hemodynamic metabolic characteristic is highly resistant to motion-induced electrical artifacts, but it exhibits an inherent time lag in hemodynamics, thus forming a complementary technique. Therefore, differentiated weights are set based on the pathology of the two types of diseases: In stroke hemiplegia, there are cortical lesions, and cognitive load fluctuates significantly instantaneously; therefore, the weight of the EEG theta / alpha ratio is increased to capture rapid changes in mental capacity. In spinal cord injury, the cerebral cortex is mostly without organic damage, and cognitive fatigue mainly stems from the continuous metabolic consumption of maintaining body balance over long periods. Furthermore, body swaying during training is more likely to introduce EEG artifacts; therefore, the weights of EEG and blood oxygenation characteristics are balanced, relying on fNIRS metabolic signals to suppress assessment distortions caused by artifacts. As one of the four dimensions of mind and body, when A high value indicates cognitive overload. The system can automatically reduce the difficulty of virtual reality tasks, lower the intensity of treadmill training, and appropriately increase exoskeleton assistance to avoid gait compensation and psychological frustration caused by cognitive overload, thus achieving coordinated management of exercise and mental load.

[0076] According to an embodiment of the present invention, multiple physiological characteristic signals include electromyography (EMG) signals. Based on these multiple physiological characteristic signals, the muscle fatigue level of the target object is determined by analyzing the time-domain and frequency-domain characteristics of the EMG signals. This includes: estimating the power spectral density of the EMG signals to obtain an EMG power spectral signal; determining an EMG frequency domain index based on the median frequency value of the EMG power spectral signal; calculating the root mean square of the EMG signals using a sliding window to obtain an EMG time domain index; and weighting the EMG frequency domain index and the EMG time domain index to obtain the muscle fatigue level.

[0077] Muscle fatigue It can be represented as:

[0078] (5).

[0079] in, This is the median frequency of the electromyographic power spectrum signal, i.e., the median frequency of electromyography. The root mean square of the electromyographic signal; The median frequency of the muscle's resting force baseline during the initial training phase; The root mean square of the baseline for muscle resting force during the initial stage of training; and The weight coefficients are values ​​ranging from 0 to 1, and the weights satisfy... ; Values ​​for stroke hemiplegia mode , Spinal cord injury pattern values , In special cases, adaptive adjustment may be implemented, such as when spasms are frequent. , If the high-frequency noise in electromyography is too high, then... , .

[0080] The root mean square (RMS) of electromyography (EMG) is a temporal characteristic of surface EMG, representing the overall amplitude level of the EMG signal within a window and reflecting the activation and exertion intensity of lower limb muscles. The real-time calculation process for the RMS value is as follows: Perform 20-500 Hz bandpass filtering and power frequency notch filtering on the raw EMG sampled at 2000 Hz; segment the signal using a 0.5s sliding window with 50% window overlap. This window can use the same segmented data as the median frequency (MDF) of the EMG, facilitating cross-verification between time-domain and frequency-domain features; calculate the original RMS value of the window, averaging the values ​​from multiple lower limb muscle group channels to obtain the real-time RMS value; if large amplitude artifacts or spastic bursts are detected, mark the RMS frame as having low reliability. The root mean square (RMS) baseline was acquired during the baseline calibration phase before the formal start of this rehabilitation training session. It was collected in the same session as the median frequency baseline, motor intention baseline, and cognitive load-related physiological baseline. Subjects performed small-amplitude lower limb pre-exertions, avoiding spastic bursts. Using the same signal processing procedure as real-time calculations, the average of multiple valid segments was taken to obtain an individual-specific baseline, without reusing baselines across training sessions. Relative activation was calculated and fused with the relative decrease in median frequency to construct muscle fatigue.

[0081] Most existing rehabilitation equipment uses preset duration and preset stimulation intensity for training, lacking real-time quantification of muscle fatigue; a few solutions use median frequency (MDF) or root mean square (RMS) electromyography (EMG) to assess fatigue alone: ​​relying solely on MDF is susceptible to interference from EMG noise and pathological motor unit recruitment distortions in hemiplegic subjects; relying solely on RMS cannot distinguish between true metabolic muscle fatigue and EMG amplitude spikes caused by pathological spastic reflexes, and most only provide threshold alarms, failing to output continuous scalar values ​​for multi-hardware closed-loop regulation. This invention integrates two types of EMG characteristics with different physiological meanings: the first item The term representing the relative decrease in median frequency (MDF) in electromyography (EMG) indicates that during metabolic peripheral fatigue of the muscle, the EMG power spectrum shifts to lower frequencies, resulting in a decrease in the MDF. This term is highly sensitive to actual metabolic fatigue, but its spectrum calculation is easily affected by noise. The second term... The RMS (Relative Increase in Root Mean Square) term of electromyography (EMG) is used because muscle fatigue compensation increases EMG amplitude. While time-domain features are simple to calculate and resistant to spectral noise, spastic bursts can also cause a sharp increase in RMS, making them unsuitable as a sole criterion for fatigue. This is compensated for by cross-checking frequency and time-domain features. Simultaneously, weights are dynamically allocated based on pathology and real-time signal conditions: for stroke-induced hemiplegic targets with disordered motor unit discharges, the frequency-domain MDF weight is increased to reduce assessment errors caused by amplitude fluctuations; for spinal cord injury targets susceptible to spectral artifacts, the time-domain RMS weight is appropriately increased; when spasticity is frequently detected, the MDF weight is automatically increased to suppress misjudgments caused by spastic RMS; when EMG spectral noise is excessive, the MDF is discarded, and only the time-domain RMS is considered. As a parameter of four-dimensional physical and mental state, when An increase in the numerical value indicates that the fatigue of the lower limb muscles has accumulated and worsened. The system automatically increases the suspension weight reduction, enhances the exoskeleton joint assistance, reduces the stimulation output, and decreases the treadmill speed and incline to achieve gradual load reduction, while taking into account both the continuity of rehabilitation training and muscle protection.

[0082] According to embodiments of the present invention, multiple physiological characteristic signals include electroencephalogram (EEG) signals, electrocardiogram (ECG) signals, and near-infrared spectral signals. Based on these multiple physiological characteristic signals, a comprehensive depression index for the target object is determined by analyzing multiple emotional states corresponding to multiple physiological dimensions of the target object. This includes: determining the frontal lobe alpha wave asymmetry index based on EEG signals; determining the heart rate variability sample entropy based on ECG signals; determining the oxyhemoglobin content based on near-infrared spectral signals; and weighting the frontal lobe alpha wave asymmetry index, heart rate variability sample entropy, and oxyhemoglobin content to obtain the comprehensive depression index.

[0083] Composite Depression Index It can be represented as:

[0084] (6).

[0085] in, The comprehensive depression index ranges from 0 to 100, with higher values ​​indicating stronger depressive stress. It is the frontal lobe alpha wave asymmetry index, reflecting the brain's frontal lobe emotion regulation function; The entropy of the heart rate variability sample is calculated from a 1000 Hz electrocardiogram signal and represents the complexity of the autonomic nervous rhythm. Emotional depression and sympathetic excitation will cause the entropy value to decrease. HbO is the normalized oxygenated hemoglobin concentration collected by frontal fNIRS. Low mood and brain stress fatigue will cause the frontal blood oxygen to decrease. , , These are the weighting coefficients. This is an offset constant. By fusing physiological signals from three dimensions, this invention can accurately and objectively identify a patient's depressive mood state, providing a foundation for subsequent coordinated psychological and motor regulation.

[0086] Frontal alpha wave asymmetry index It can be represented as:

[0087] (7).

[0088] in, Power of the right frontal F4 electrode in the α band; The power of the alpha band of the F3 electrode in the left frontal lobe is shown. Alpha wave power is negatively correlated with the activation level of the cerebral cortex. The higher the alpha power, the weaker the neural activity and the stronger the inhibition in the corresponding brain region.

[0089] Current clinical assessments of depression and training stress primarily rely on subjective scales, which can only be completed offline before and after training and cannot capture real-time dynamic emotional fluctuations during walking rehabilitation training. Furthermore, scale results can be distorted when the target subject has cognitive impairment or subjectively masks their emotions. (FAA frontal alpha asymmetry index, heart rate variability sample entropy) While the frontal lobe fNIRS and oxygenated hemoglobin (HbO) are known physiological markers in emotion research, they all have significant limitations: FAA is easily affected by stroke lesions and gait artifacts; HRV sample entropy in spinal cord injury subjects is affected by inherent autonomic pathway damage; HbO is constrained by hemodynamic lag and is only good at reflecting cumulative emotional stress; no single indicator can reliably quantify depression and training stress directly in dynamic lower limb rehabilitation scenarios. Furthermore, most existing studies focus on resting-state measurements in general depression populations, without setting differentiated fusion weights for stroke hemiplegia and spinal cord injury, and without directly using the fused emotion indicators for closed-loop control of rehabilitation hardware. This invention selects three features from different physiological dimensions: FAA reflects the activity of the prefrontal cortex emotional circuit; It reflects the peripheral autonomic nervous system stress state; HbO reflects the metabolic consumption level of the prefrontal cortex, the brain region regulating emotion; the three form a mutual verification between the central nervous system, autonomic nervous system, and cerebral blood flow metabolism, without the need for additional sensors such as skin conductance and eye movement, reusing the existing EEG, ECG, and fNIRS multimodal sensing hardware of the system, and according to the difference in the pathological origin of emotional disorders in the two types of diseases, two sets of weights are fitted separately: depression in stroke hemiplegia is mostly due to organic damage to the frontal lobe of the brain, so the weight of the EEG FAA feature is increased; the cerebral cortex of spinal cord injury is mostly intact, and emotional stress is more due to autonomic nervous system dysfunction, so the weight of the FAA feature is increased. The weighting of the comprehensive depression index (Dep) is calculated. The output is a continuous scalar value, used as input for four-dimensional physical and mental state parameters, which can be used to dynamically adjust the training parameters of the entire rehabilitation equipment. Furthermore, a synchronous time window constraint is set for vagal nerve electrical stimulation and virtual reality scene switching. When Dep exceeds the emotional threshold, vagal nerve electrical stimulation parameters are only allowed to be updated within the virtual reality scene switching time window, achieving temporal synergy between visual psychological intervention and peripheral vagal nerve regulation. This enables real-time objective quantification of depression and training stress during rehabilitation walking training, and allows for integrated closed-loop intervention of the entire rehabilitation equipment, encompassing motor, neurological, and psychological aspects.

[0090] According to an embodiment of the present invention, the multiple weighting coefficients of the weighted processing are determined based on the frontal lobe state and spinal cord state of the target object.

[0091] , , and It can be obtained by fitting multiple linear regression of the physiological-depression scale of the target subject, distinguishing between two sets of fixed parameters for stroke hemiplegia and spinal cord injury, and realizing the fusion of multiple physiological signals to quantify real-time depressive stress levels. (Stroke hemiplegia mode) Stroke is mainly characterized by damage to the frontal lobe nerves and changes in the emotional brain regions, as indicated by electroencephalogram (EEG). Highest weight. Spinal cord injury pattern. Spinal cord injury has a high incidence of autonomic nerve abnormalities, and upregulation Weighting enhances heart rate-related emotional representations.

[0092] According to an embodiment of the present invention, multiple device control parameters are determined based on the intensity of exercise intention, cognitive load, muscle fatigue, and comprehensive depression index. This includes: weighted fusion processing of musculoskeletal motion state indicators, neurocognitive state indicators, and psychological and emotional state indicators to obtain a global training incentive intensity; mapping the global training incentive intensity to at least one of the following: suspension weight reduction ratio, treadmill speed incline, exoskeleton hip and knee joint assist torque, functional electrical stimulation parameters, vagus nerve stimulation parameters, and virtual reality scene parameters, to obtain a mapping result; and determining the multiple device control parameters based on the mapping result.

[0093] Global training incentive intensity It can be represented as:

[0094] (8).

[0095] in, The weights for exercise intention are different for stroke and spinal cord injury patterns, representing the proportion of the influence of exercise intention on training intensity. The weighting of cognitive dimensions determines the degree to which mental state restricts training; The weighting of muscle fatigue is higher, and the weighting of spinal cord injury patterns is higher, prioritizing muscle protection. This is the emotional weight; the emotional sensitivity pattern will amplify this weight; it must meet the following conditions. Stroke Hemiplegia Rehabilitation Model: Spinal cord injury rehabilitation model .

[0096] Global training incentive intensity can be used as a global device control command to adjust the suspension weight reduction ratio, treadmill speed and incline, exoskeleton hip and knee joint assist torque, FES stimulation parameters, VNS stimulation parameters, and VR scene adjustment parameters. A higher global training incentive intensity value indicates a stronger exercise intention, lower cognitive and muscle fatigue, and lower depressive stress levels in the target subject.

[0097] A pre-defined linear mapping relationship can be used to... The parameters are converted into six categories of hardware executable parameters: suspension weight reduction ratio, treadmill speed and incline, exoskeleton hip and knee joint assist torque, FES stimulation parameters, VNS stimulation parameters, and VR scene adjustment parameters. All hardware outputs are clamped at upper and lower limits to ensure that the output does not exceed the physical boundaries of the device.

[0098] Weight reduction ratio of suspension weight reduction device Using the mapping formula:

[0099] (9).

[0100] in, The unit is %, and the output clamping range is 20%~100%. The baseline weight loss percentage; This represents the weight reduction ratio gain coefficient; in stroke hemiplegic mode. , Spinal cord injury patterns , .

[0101] when Increased size leads to a decreased weight loss ratio, with the lower limbs bearing more of the body weight; when Decreasing weight leads to a higher weight reduction ratio, allowing the suspension system to bear more weight and reducing the load on the lower limbs; (The last sentence appears to be incomplete and possibly refers to a different context.) Under certain conditions, the weight loss benchmark for spinal cord injury patterns is higher, achieving joint overload protection.

[0102] Treadmill speed ,slope The mapping relationship is as follows:

[0103] (10).

[0104] in, The velocity unit is m / s, and the clamping range is 0~1.2 m / s. As a baseline treadmill speed, This refers to the speed gain coefficient of the treadmill. Slope unit: °; clamping range: 0~10°. Based on the treadmill incline, This refers to the treadmill incline gain coefficient. Stroke hemiplegia mode: Spinal cord injury patterns .

[0105] As the incline increases, the treadmill speed and incline also increase, thus increasing the training load for walking. When the slope is lowered, the speed decreases and the incline approaches 0. In stroke hemiplegia mode, if the peak value of muscle spasm is detected to exceed the limit, the treadmill speed is forcibly reduced to 0.5 times the original speed to prevent further aggravation of spasm.

[0106] Hip joint assist torque of lower limb exoskeleton robot Knee joint assist torque The mapping relationship is as follows:

[0107] (11).

[0108] in, and torque unit ; As the baseline hip joint assist torque, This refers to the hip joint assist torque gain coefficient; As the reference knee joint assist torque, Knee joint assist torque gain coefficient; stroke hemiplegic mode At the same time, a hard constraint is added on the rate of change of torque. Limiting instantaneous assist gradients to avoid inducing muscle spasms; output clamping. Spinal cord injury model: It outputs a steady-state continuous torque to ensure stable support for the hip and knee joints.

[0109] when An increase indicates a strong willingness of the target to move actively, while the exoskeleton's output of assistance decreases, guiding the target to exert force autonomously; when... Decrease or increase the joint assist torque to compensate for the target object's lack of muscle strength.

[0110] Normalized stimulation intensity of functional electrical stimulation module The mapping formula is:

[0111] (12).

[0112] in, The output normalization range is 0~0.85; As the baseline functional electrical stimulation intensity, Functional electrical stimulation intensity gain coefficient; stroke hemiplegic mode: When muscle spasm is detected, the stimulation intensity mapping output remains unchanged, and the stimulation target electrode is switched to the antagonistic muscle group to achieve spasm inhibition; Spinal cord injury mode: It continuously activates disused and atrophied muscle groups.

[0113] The lower the value, the weaker the target object's ability to activate its own muscles, and the higher the intensity of the FES output stimulus.

[0114] Normalized stimulation intensity of vagal nerve electrical stimulation module The mapping formula is:

[0115] (13).

[0116] in, Output clamping 0~0.75; As a baseline for vagal nerve electrical stimulation intensity, Vagus nerve electrical stimulation intensity gain coefficient; Stroke hemiplegia pattern: Spinal cord injury patterns: 5.

[0117] Only synchronized window when switching VR scenes The internal stimulus parameters are updated, and the previous effective output is maintained at other times; when depressive stress increases, The amplitude or duration of low-frequency stimulation of the VNS is increased within the synchronization window to achieve neuro-soothing intervention.

[0118] Setting VR Comprehensive Normalization State Parameters in Virtual Reality Scene Generation Module The mapping formula is:

[0119] (14).

[0120] in, Based on the baseline VR integrated normalized state parameters, The VR integrated normalized state parameter gain coefficient; the stroke hemiplegia mode and the spinal cord injury mode share the same set of parameters. Output clamp [0,1].

[0121] Depend on Further calculation yields the VR sub-parameter: task difficulty. Color temperature coefficient Unit: K; sound effect rhythm coefficient .

[0122] rise, Increase the size of the screen to increase the difficulty of the walking task, increase the saturation of the image, and speed up the pace of the background music. reduce, The difficulty of the task is reduced, the visuals are softened to a lower color temperature, and the background music becomes more soothing. When the comprehensive depression index (Dep) exceeds a preset threshold, the system actively triggers a VR scene switching event, generating the scene switching moment. By opening the VNS synchronization time window, the timing coordination of psychological visual intervention and vagus nerve electrical stimulation can be achieved.

[0123] Figure 2 A schematic diagram illustrating the adjustment of VR scene parameters and VNS electrical stimulation parameters according to an embodiment of the present invention is shown.

[0124] like Figure 2 As shown, commands can be issued synchronously to coordinate all hardware to complete parameter adaptive adjustment; if the real-time Dep comprehensive depression index exceeds the preset emotional threshold, the VR and VNS time-series synchronization constraint mechanism is activated, adjusting the vagus nerve electrical stimulation parameters only within the scene switching window to achieve synchronous psychological-neurological intervention. The constraint formula is as follows:

[0125] (15).

[0126] in, For VR scene switching moments, To preset the synchronization window width, only the vagus nerve stimulation parameters are adjusted within this interval to ensure temporal coordination.

[0127] Furthermore, physiological signals and device operating parameters can be continuously transmitted back at a period of 50 ms, and the three-layer closed-loop logic of "sensing-evaluation-regulation" can be executed cyclically, dynamically updated based on the latest four-dimensional state feature vector. It continuously fine-tunes all hardware operating parameters and matches the dynamic changes in muscle fatigue, cognitive load, and depressive stress during training in real time, forming a continuous adaptive closed-loop regulation.

[0128] According to an embodiment of the present invention, the method further includes: determining the peak value of electromyography (EMG), heart rate fluctuation, and oxyhemoglobin concentration decrease gradient based on the electromyography (EMG) signal, electrocardiogram (ECG) signal, and near-infrared spectral signal of the target object, respectively; determining a first risk indicator, a second risk indicator, and a third risk indicator based on the peak value of EMG, heart rate fluctuation, and oxyhemoglobin concentration decrease gradient, respectively; weighting the first risk indicator, the second risk indicator, and the third risk indicator to obtain a comprehensive risk value; and performing graded control of the lower limb assistive device according to a plurality of preset safety threshold intervals based on the comprehensive risk value, including: reducing the training intensity when the comprehensive risk value is in the first threshold interval; initiating a neuro-soothing intervention and lowering the training parameters when the comprehensive risk value is in the second threshold interval; and executing a shutdown protection mechanism when the comprehensive risk value is in the third threshold interval.

[0129] Parallel monitoring of electromyographic spasm peak values ​​is possible. Heart rate fluctuations and the gradient of cerebral blood oxygen concentration decrease Three risk indicators were used to calculate the risk level of electromyositis. Heart rate fluctuation risk level Cerebral blood oxygen fluctuation risk level The specific formula is as follows:

[0130] (16).

[0131] (17).

[0132] (18).

[0133] in, , and These are preset level 1, level 2, and level 3 electromyography (EMG) spasm thresholds, used to determine the risk level corresponding to the current peak EMG spasm value; , and These are preset level 1, level 2, and level 3 heart rate fluctuation thresholds, used to determine the risk level corresponding to the current heart rate fluctuation. , and These are preset level 1, level 2, and level 3 blood oxygen fluctuation thresholds, used to determine the risk level corresponding to the current blood oxygen fluctuation.

[0134] Subsequently, the weighted fusion formula was used to calculate the overall risk value:

[0135] (19).

[0136] in, , , For normalized weight coefficients, satisfying Stroke model with fixed weights , , Spinal cord injury adaptive adjustment to , , .

[0137] Three levels of security thresholds can be set to comprehensively assess risks. The core basis for judgment: This is a Level 1 safety zone; rehabilitation can proceed normally. The alert level is set at level two; training intensity is reduced and neurological calming interventions are initiated. In the Level 3 high-risk zone, the equipment shuts down and enters relaxed protection mode. For stroke and spinal cord injury, the threshold can be slightly adjusted to suit the specific symptoms. Simultaneously, [the system is set to...]. , , Individual thresholds provide a safety net, preventing severe abnormalities in a single indicator from being masked by weighted averaging, ensuring timely protection against any exceedance of spasms, heart rate, or mood. Risk assessment incorporates multi-window anti-shake features to prevent noise-induced accidental start / stop.

[0138] Figure 3 A schematic diagram of the complete closed-loop control process according to an embodiment of the present invention is shown.

[0139] like Figure 3 As shown, after the task is completed, the controller can store the entire four-dimensional state curve, equipment control parameters, and early warning records, and automatically generate an evaluation report to provide data support for the next training parameter adjustment.

[0140] Figure 4 An architecture diagram of a lower limb assistive device control system based on multimodal physiological signal feedback according to an embodiment of the present invention is shown.

[0141] like Figure 4 As shown, the lower limb assistive device control system based on multimodal physiological signal feedback can include a central control decision layer, a mechanical motion execution layer, a VR psychological interaction layer, a neuro-electrical stimulation intervention layer, and a multimodal physiological perception layer.

[0142] Figure 5 A block diagram of a lower limb assistive device control system based on multimodal physiological signal feedback according to an embodiment of the present invention is shown.

[0143] like Figure 5 As shown, the lower limb assistive device control system based on multimodal physiological signal feedback includes a signal acquisition module 510, a three-dimensional index determination module 520, a collaborative regulation parameter determination module 530, and a device control module 540.

[0144] The signal acquisition module 510 is configured to acquire multiple physiological characteristic signals of the target object.

[0145] The three-dimensional indicator determination module 520 is configured to perform feature extraction and weighted processing on the multiple physiological feature signals, and simultaneously determine the musculoskeletal motion state indicators, neurocognitive state indicators and psychological and emotional state indicators of the target object, so as to realize the integrated perception of the three dimensions of musculoskeletal motion, neurocognition and psychological and emotional.

[0146] The collaborative regulation parameter determination module 530 is configured to determine multiple device control parameters based on bone movement state indicators, neurocognitive state indicators, and psychological and emotional state indicators, so as to achieve closed-loop regulation of psychological and motor collaborative rehabilitation.

[0147] The device control module 540 is configured to control the lower limb assistive device based on multiple device control parameters.

[0148] Among them, musculoskeletal movement status indicators include at least muscle fatigue, neurocognitive status indicators include at least intensity of motor intention and cognitive load, and psychological and emotional status indicators include at least the comprehensive depression index.

[0149] Any one or more of the modules, submodules, units, and subunits according to embodiments of the present invention, or at least part of the functions of any one or more of them, can be implemented in a single module. Any one or more of the modules, submodules, units, and subunits according to embodiments of the present invention can be implemented by being divided into multiple modules. Any one or more of the modules, submodules, units, and subunits according to embodiments of the present invention can be at least partially implemented as hardware circuits, such as field-programmable gate arrays (FPGAs), programmable logic arrays (PLAs), systems-on-a-chip, systems-on-a-substrate, systems-on-package, application-specific integrated circuits (ASICs), or implemented in hardware or firmware by any other reasonable means of integrating or packaging circuits, or implemented in software, hardware, and firmware, or in any suitable combination of any of these three implementation methods. Alternatively, one or more of the modules, submodules, units, and subunits according to embodiments of the present invention can be at least partially implemented as computer program modules, which, when run, can perform corresponding functions.

[0150] For example, any plurality of the signal acquisition module 510, the three-dimensional index determination module 520, the collaborative control parameter determination module 530, and the device control module 540 can be combined into one module / unit / subunit, or any one of these modules / units / subunits can be split into multiple modules / units / subunits. Alternatively, at least part of the functionality of one or more of these modules / units / subunits can be combined with at least part of the functionality of other modules / units / subunits and implemented in one module / unit / subunit. According to embodiments of the present invention, at least one of the signal acquisition module 510, the three-dimensional index determination module 520, the collaborative control parameter determination module 530, and the device control module 540 can be at least partially implemented as hardware circuitry, such as a field-programmable gate array (FPGA), a programmable logic array (PLA), a system-on-a-chip, a system-on-a-substrate, a system-on-package, an application-specific integrated circuit (ASIC), or any other reasonable means of integrating or packaging the circuitry, or implemented in software, hardware, or firmware, or in any suitable combination of any of these three implementation methods. Alternatively, at least one of the signal acquisition module 510, the three-dimensional index determination module 520, the coordinated control parameter determination module 530, and the device control module 540 can be at least partially implemented as a computer program module, which can perform corresponding functions when the computer program module is run.

[0151] The following example uses a stroke patient with hemiplegia and moderate depression to demonstrate the lower limb walking rehabilitation training conducted using the lower limb assistive device control system of this invention. The complete implementation process is as follows:

[0152] First, the device is worn and the hardware is integrated. After the system is powered on and self-tested, the target subject wears the forehead cerebral oxygenation probe, scalp EEG electrodes, chest ECG electrodes, lower limb quadriceps / tibialis anterior muscle electromyography electrodes, and auricular VNS stimulation electrodes; wears a lower limb exoskeleton with a torque motor, and a suspension weight reduction device is attached to the torso; the VR headset is turned on, and the central controller establishes a communication channel with all hardware.

[0153] Second, physiological signal preprocessing. The multimodal acquisition unit simultaneously acquires raw EEG, fNIRS, EMG, and ECG signals. After hardware filtering to remove power frequency noise, a cross-modal time-series alignment loss function is used to complete the unified registration of the timestamps of the four signals, eliminating the misalignment error of the sampling frequencies of different sensors.

[0154] Third, four-dimensional quantification of physical and mental state. The controller extracts temporal features and calculates the intensity of brain motor intention, prefrontal cognitive fatigue score, lower limb muscle activation and spasm peak; substitutes them into the comprehensive depression quantification model to obtain a real-time depression index Dep=68, indicating the presence of moderate training stress and depressive mood.

[0155] Fourth, the stroke hemiplegia mode is coordinated with parameter adjustment. The system automatically switches to stroke hemiplegia mode, increases the weight of electromyography spasm recognition, reduces the instantaneous assistance gradient of the exoskeleton, and prioritizes FES stimulation of the tibialis anterior antagonist muscle; simultaneously, it lowers the treadmill base speed and increases the suspension weight reduction ratio; due to the high depression index, a 2-second VR scene switching synchronization window is set, and the duration of VNS low-frequency soothing stimulation is extended at the scene switching node, the VR screen is switched to a low color temperature green plant trail, and the background music volume is reduced.

[0156] Fifth, real-time closed-loop iterative control. During training, physiological and equipment data are continuously transmitted back at 50 ms intervals. The controller cyclically updates the four-dimensional status parameters and dynamically fine-tunes the exoskeleton assistance, electrical stimulation current, and treadmill incline. When the peak value of electromyography spasm slightly exceeds the first-level threshold at the 8th minute of training, the system automatically reduces the FES stimulation power and slows down the treadmill speed, without interruption, and continuously maintains low-load rehabilitation training.

[0157] Sixth, risk response and training completion. Throughout the entire session, heart rate and cerebral oxygenation levels remained within safe ranges, and no level two or three warnings were triggered. After 30 minutes of rehabilitation training, the system stopped all electrical stimulation, the treadmill was reset to zero, and the suspension device released its tension. The controller stored the entire physiological curve, equipment adjustment records, and depression index change curve, and automatically generated a rehabilitation assessment report, suggesting that the basic exercise load be reduced and the duration of VNS neural modulation be extended for the next training session.

[0158] When the target subject is changed to a spinal cord injury subject, the system automatically switches to the spinal cord injury mode: increases the weight of joint steady-state torque control, widens the range of continuous exoskeleton assistance, prolongs the overall stimulation duration of VNS, and weakens the weight of spasticity recognition, while the rest of the collection, assessment, closed-loop control, and safety warning processes remain unchanged.

[0159] In summary, this invention achieves integrated perception and coordinated regulation across three dimensions: neurocognition, psychological emotion, and musculoskeletal movement, breaking through the traditional single-mode rehabilitation that only targets muscle training; unified closed-loop linkage of multiple hardware components eliminates clinical problems such as temporal misalignment, lower limb compensation, and mismatch between training intensity and physical and mental state; adaptive control logic based on disease type accurately adapts to various subjects, including stroke hemiplegia and spinal cord injury combined with emotional disorders; multi-indicator joint early warning plus tiered emergency protection significantly reduces the safety risks of severe rehabilitation training; modular architecture is compatible with whole-machine sales and old equipment retrofitting, with rich industrial application scenarios and strong clinical adaptability.

[0160] Those skilled in the art will understand that the features described in the various embodiments of the present invention can be combined and / or combined in various ways, even if such combinations or combinations are not explicitly described in the present invention. In particular, the features described in the various embodiments of the present invention can be combined and / or combined in various ways without departing from the spirit and teachings of the present invention. All such combinations and / or combinations fall within the scope of the present invention.

[0161] The embodiments of the present invention have been described above. However, these embodiments are merely illustrative and not intended to limit the scope of the invention. Although various embodiments have been described above, this does not mean that the measures in the various embodiments cannot be used advantageously in combination. Various substitutions and modifications can be made by those skilled in the art without departing from the scope of the invention, and all such substitutions and modifications should fall within the scope of the invention.

Claims

1. A control method for a lower limb assistive device based on multimodal physiological signal feedback, characterized in that, include: Collect multiple physiological characteristic signals of the target object; Feature extraction and weighted processing are performed on the multiple physiological feature signals to simultaneously determine the musculoskeletal motor state indicators, neurocognitive state indicators, and psychological and emotional state indicators of the target object. Based on the musculoskeletal motion state indicators, neurocognitive state indicators, and psychological and emotional state indicators, multiple device control parameters are determined, including the suspension weight reduction ratio, treadmill speed and incline, exoskeleton hip and knee joint assist torque, functional electrical stimulation parameters, vagus nerve stimulation parameters, and virtual reality scene parameters. The lower limb assistive device is controlled based on the control parameters of the aforementioned multiple devices; The musculoskeletal movement state indicators include at least muscle fatigue, the neurocognitive state indicators include at least the intensity of motor intention and cognitive load, and the psychological and emotional state indicators include at least the comprehensive depression index.

2. The method according to claim 1, characterized in that, Based on the musculoskeletal motion state indicators, neurocognitive state indicators, and psychological and emotional state indicators, multiple device control parameters are determined, including: The musculoskeletal motor state index, the neurocognitive state index, and the psychological and emotional state index are weighted and fused to obtain the global training incentive intensity. The global training incentive intensity is mapped to at least one of the following: suspension weight reduction ratio, treadmill speed and incline, exoskeleton hip and knee joint assist torque, functional electrical stimulation parameters, vagus nerve stimulation parameters, and virtual reality scene parameters, to obtain the mapping result; Based on the mapping results, the control parameters of the plurality of devices are determined.

3. The method according to claim 2, characterized in that, The method further includes: When the comprehensive depression index in the psychological and emotional state indicators exceeds the preset emotional threshold, a virtual reality scene switching event is triggered, and a synchronization time window is generated. The synchronization time window is used to limit the adjustment period of the vagus nerve stimulation parameters in order to achieve temporal synergy between psychological visual intervention and neural regulation.

4. The method according to claim 1, characterized in that, The process of extracting and weighting the multiple physiological feature signals to simultaneously determine the musculoskeletal motor state indicators, neurocognitive state indicators, and psychological and emotional state indicators of the target object includes: Based on the aforementioned multiple physiological characteristic signals, the intensity of the target object's movement intention is determined by analyzing the target object's brain activity level and muscle activity level; Based on the aforementioned multiple physiological characteristic signals, the cognitive load of the target object is determined by analyzing its mental workload and metabolic burden. Based on the aforementioned multiple physiological characteristic signals, the muscle fatigue level of the target object is determined by analyzing the time-domain and frequency-domain characteristics of the electromyographic signals of the target object. Based on the aforementioned multiple physiological characteristic signals, the comprehensive depression index of the target object is determined by analyzing the multiple emotional states corresponding to each of the multiple physiological dimensions of the target object. Based on the intensity of the motor intention, the cognitive load, and the comprehensive depression index, the musculoskeletal motor state indicators, neurocognitive state indicators, and psychological and emotional state indicators of the target object are determined.

5. The method according to claim 4, characterized in that, The multiple physiological characteristic signals include electroencephalogram (EEG) signals and electromyogram (EMG) signals. The determination of the target object's motor intention intensity based on these multiple physiological characteristic signals, by analyzing the target object's brain activity and muscle activity, includes: The power spectral density of the EEG signal is estimated to obtain the EEG power spectral signal; The total power of the motor brain region is obtained based on the sum of the integrated power of the EEG power spectrum signal in the first target frequency band and the integrated power of the EEG power spectrum signal in the second target frequency band. The electromyographic signal is truncated and the root mean square is calculated using a pre-activation window to determine the muscle pre-activation root mean square, wherein the window termination time of the pre-activation window is the current sampling time; The total power of the motor brain region and the root mean square of the muscle preactivation are weighted to obtain the intensity of motor intention.

6. The method according to claim 4, characterized in that, The multiple physiological characteristic signals include electroencephalogram (EEG) signals and near-infrared spectral signals. The determination of the cognitive load of the target subject based on these multiple physiological characteristic signals, through analysis of the target subject's mental workload and metabolic burden, includes: The power spectral density of the EEG signal is estimated to obtain the EEG power spectral signal; The EEG power ratio is obtained based on the ratio of the integrated power of the EEG power spectrum signal in the third target frequency band to the integrated power of the EEG power spectrum signal in the fourth target frequency band. The oxyhemoglobin content was determined based on the near-infrared spectral signal. The cognitive load is obtained by weighting the EEG power ratio and the oxyhemoglobin content.

7. The method according to claim 4, characterized in that, The plurality of physiological characteristic signals include electromyography (EMG) signals. The determination of the muscle fatigue level of the target object based on the plurality of physiological characteristic signals by analyzing the time-domain and frequency-domain characteristics of the EMG signals of the target object includes: The electromyographic signal is subjected to power spectral density estimation to obtain the electromyographic power spectral signal; Based on the median frequency value of the electromyographic power spectrum signal, the electromyographic frequency domain index is determined; The root mean square of the electromyographic signal is calculated using a sliding window to obtain the electromyographic time-domain index. The muscle fatigue level is obtained by weighting the electromyographic frequency domain index and the electromyographic time domain index.

8. The method according to claim 4, characterized in that, The multiple physiological characteristic signals include electroencephalogram (EEG) signals, electrocardiogram (ECG) signals, and near-infrared spectral signals. Based on these multiple physiological characteristic signals, the comprehensive depression index of the target object is determined by analyzing multiple emotional states corresponding to multiple physiological dimensions of the target object, including: Based on the aforementioned EEG signals, the frontal lobe alpha wave asymmetry index was determined; Based on the electrocardiogram signal, determine the sample entropy of heart rate variability; The oxyhemoglobin content was determined based on the near-infrared spectral signal. The comprehensive depression index is obtained by weighting the frontal alpha wave asymmetry index, the heart rate variability sample entropy, and the oxyhemoglobin content.

9. The method according to claim 1, characterized in that, The method further includes: Based on the electromyography (EMG) signals, electrocardiogram (ECG) signals, and near-infrared spectral signals of the target object, the peak value of EMG spasm, heart rate fluctuation, and the gradient of oxyhemoglobin concentration decrease were determined, respectively. Based on the peak value of electromyography spasm, the heart rate fluctuation, and the gradient of oxyhemoglobin concentration decrease, a first risk indicator, a second risk indicator, and a third risk indicator are determined respectively. The first risk indicator, the second risk indicator, and the third risk indicator are weighted to obtain a comprehensive risk value. Based on the comprehensive risk value, the lower limb assistive device is subjected to graded control according to multiple preset safety threshold ranges, including: When the overall risk value is within the first threshold range, reduce the training intensity; When the overall risk value is within the second threshold range, initiate neuro-soothing intervention and lower the training parameters; When the overall risk value is within the third threshold range, shutdown protection is implemented.

10. A control system for a lower limb assistive device based on multimodal physiological signal feedback, characterized in that, include: The signal acquisition module is configured to acquire multiple physiological characteristic signals of the target object; The three-dimensional indicator determination module is configured to perform feature extraction and weighted processing on the multiple physiological feature signals, and simultaneously determine the musculoskeletal motion state indicators, neurocognitive state indicators and psychological and emotional state indicators of the target object, so as to realize the integrated perception of the three dimensions of musculoskeletal motion, neurocognition and psychological and emotional. The collaborative regulation parameter determination module is configured to determine multiple device control parameters based on bone movement state indicators, neurocognitive state indicators, and psychological and emotional state indicators, so as to achieve closed-loop regulation of psychological and motor collaborative rehabilitation. The multiple device control parameters include suspension weight reduction ratio, treadmill speed and incline, exoskeleton hip and knee joint assist torque, functional electrical stimulation parameters, vagus nerve stimulation parameters, and virtual reality scene parameters. The device control module is configured to control the lower limb assistive device based on the plurality of device control parameters; The musculoskeletal movement state indicators include at least muscle fatigue, the neurocognitive state indicators include at least the intensity of motor intention and cognitive load, and the psychological and emotional state indicators include at least the comprehensive depression index.

Images