Stroke rehabilitation assisting method and device based on multi-source physiological signal wearable device

Abstract

The application discloses an auxiliary stroke rehabilitation method and device based on a multi-source physiological signal wearable device, and the method comprises the following steps: monitoring the pulse waveform variation rate and the sublingual vein characteristic parameters of a subject through a self-powered five-zang monitoring wristband; dynamically evaluating the zang-fu imbalance index of the subject according to the pulse waveform variation rate and the sublingual vein characteristic parameters; generating individualized rehabilitation adjustment parameters according to the zang-fu imbalance index of the subject; synchronously collecting the electroencephalogram signal and the surface electromyogram signal of a target muscle of the subject, and generating an acupoint activation time sequence atlas corresponding to a preset hand movement; optimizing the acupoint activation time sequence atlas according to the individualized rehabilitation adjustment parameters, so as to adjust the acupoint selection, the stimulation intensity and the stimulation time sequence; and presenting an embedded stomach meridian of foot yangming virtual acupuncture scene to the subject through a VR guide instrument. The application integrates the brain and muscle electrical cooperative decoding, the individualized parameter adjustment, the VR and the zang-fu imbalance traditional Chinese medicine diagnosis method into the rehabilitation scheme, and the rehabilitation effect of stroke rehabilitation is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of stroke rehabilitation assistance, in particular to an auxiliary stroke rehabilitation method and device based on a multi-source physiological signal wearable device and a computing device. BACKGROUND

[0002] Traditional stroke rehabilitation methods mainly rely on physical therapy and placeholder therapy (such as muscle stretching, joint movement and functional training). Due to the lack of personalized and precise rehabilitation methods, the rehabilitation effect is poor. Although some modern rehabilitation devices (such as robot-assisted rehabilitation systems, electrical stimulation devices, etc.) have improved the rehabilitation effect to some extent, they still lack real-time monitoring and dynamic adjustment capabilities for the physiological state of patients, and cannot fully reflect the overall health status of patients. In addition, the application of traditional Chinese medicine theory in the field of rehabilitation mainly focuses on traditional methods such as acupuncture and massage, and lacks the combination of modern medical technology. In particular, the concept of "visceral imbalance" in traditional Chinese medicine theory is of great significance in rehabilitation, but lacks a scientific quantitative evaluation method.

[0003] To solve the above problems, the present application provides an auxiliary stroke rehabilitation method based on a multi-source physiological signal wearable device, which integrates brain-muscle electrical decoding, personalized parameter adjustment, VR, and visceral imbalance diagnosis methods into the rehabilitation scheme to improve the rehabilitation effect of stroke rehabilitation. SUMMARY

[0004] In view of the above problems, the present application provides an auxiliary stroke rehabilitation method and device based on a multi-source physiological signal wearable device and a computing device.

[0005] According to one aspect of the present application, an auxiliary stroke rehabilitation method based on a multi-source physiological signal wearable device is provided, comprising:

[0006] The self-powered five-zang monitoring wristband monitors the pulse waveform variation rate and the sublingual vein characteristic parameters of the subject, and the self-powered five-zang monitoring wristband includes a flexible piezoelectric sensor, a photoplethysmographic pulse wave (PPG) sensor and a tongue image acquisition module;

[0007] According to the pulse waveform variation rate and the sublingual vein characteristic parameters, the visceral imbalance index of the subject is dynamically evaluated; and according to the visceral imbalance index of the subject, personalized rehabilitation adjustment parameters are generated, wherein the personalized rehabilitation adjustment parameters include acupoint selection, stimulation intensity, stimulation timing, and stimulation frequency of the acupoint activation timing map, stimulation intensity, stimulation frequency, action time and VR scene interaction mode in the closed loop of collaborative training and electrical pulse stimulation.

[0008] Synchronously collect an electroencephalogram (EEG) signal of a subject and a surface electromyogram (sEMG) signal of a target muscle, analyze a nonlinear coupling relationship between a motor imagery potential of the EEG signal and the sEMG signal through a brain-muscle electrical synergy decoding module, and generate an acupoint activation timing atlas corresponding to a preset hand movement;

[0009] Optimize the acupoint activation timing atlas according to the personalized rehabilitation adjustment parameters to adjust acupoint selection, stimulation intensity, and stimulation timing, present a built-in stomach meridian virtual acupuncture scene to the subject through a portable VR guide instrument, provide real-time tactile feedback corresponding to the preset hand movement through bone conduction earphones, and guide the subject to perform a preset rehabilitation movement.

[0010] In an optional manner, the presenting of the built-in stomach meridian virtual acupuncture scene to the subject through the portable VR guide instrument further includes:

[0011] When the subject completes the preset hand movement in the VR scene and meets a trigger condition based on heart rate variability (HRV) rhythm, trigger personalized electric pulse stimulation according to the zang-fu imbalance index to stimulate a preset acupoint;

[0012] Meanwhile, a dynamic blood flow path corresponding to the preset acupoint is presented through the portable VR guide instrument.

[0013] In an optional manner, a calculation formula of the zang-fu imbalance index is as follows:

[0014]

[0015] wherein, w i is a weight coefficient of the ith zang-fu; P i is a real-time monitored pulse waveform variation rate; P norm,i is a normal variation rate threshold of the corresponding zang-fu; TongueBaseline i is a tongue feature value of the ith zang-fu; TongueFeature i is a tongue baseline value of the ith zang-fu; P crit,i is a critical variation rate threshold of the ith zang-fu; Θ(.) is a Heaviside step function; t is a current time; and τ is a time integral variable.

[0016] In an optional manner, the VR scene interaction mode is to render light and shadow changes of the blood flow path in real time through a Shader Graph method, wherein a color gradient of the blood flow path is positively correlated with a current heart rate variability frequency energy ratio.

[0017] And output the spatialized audio through the bone conduction earphone, wherein the left / right channel phase difference is linearly related to the EEG mu rhythm energy offset, and when the patient's motor imagination potential is enhanced, the audio sound image is automatically offset to the contralateral side.

[0018] The action trajectory is captured through an inertial measurement unit (IMU), and is mapped in real time into the kinematics parameters of the virtual limb in the VR scene to form a somatosensory-visual closed loop.

[0019] In an optional mode, the state space of the preset rehabilitation action includes PWV, HRV, myoelectric signal entropy value, and action completion degree parameter.

[0020] The action space of the preset rehabilitation action includes acupoint activation parameter, VR scene complexity parameter, and electric stimulation parameter.

[0021] The reward function of the preset rehabilitation action is:

[0022]

[0023] Wherein, ΔFuglMeyerScore is the increment of the score scale; StandardTime is the standard time required for the rehabilitation action; ActualTime is the time spent for actually completing the rehabilitation action; and NumberOfAbnormalAlerts is the number of abnormal alerts.

[0024] In an optional mode, the electric pulse stimulation parameter adopts a heart rate variability (HRV) feedback model, wherein the expression of the HRV feedback model is:

[0025]

[0026] Wherein, I(t) is the electric pulse stimulation intensity at time t; Kp is the proportional gain coefficient; is the center frequency; φ HRV (t) is the HRV phase modulation function; HRV LF / HF (t) is the low frequency (LF) / high frequency (HF) ratio of the heart rate variability; PWV(t) is the pulse wave velocity at time t; PWV base is the PWV value of the patient in the normal state; f c is the center frequency.

[0027] In an optional mode, the acupoint activation time sequence map models the phase locking value (PLV) between the mu rhythm and motor imagination potential of the myoelectric signal (sEMG) through a recursive least square (RLS) method to obtain a dynamic gain function; wherein the dynamic gain function is:

[0028]

[0029] wherein, G base is a basic gain coefficient; k is a PLV gain adjustment coefficient; PLV max is a preset maximum phase locking value; HRV LF / HF,base is a heart rate variability baseline value of a patient in a resting state;

[0030] The electric pulse stimulation intensity parameter is adjusted in real time according to the dynamic gain function, and the adjustment formula is:

[0031]

[0032] wherein, t onset is a stimulation starting time; τ adapt is an adaptive adjustment time constant, used for smoothing the instantaneous impact of gain change on stimulation intensity; I(t) is an unadjusted electric pulse stimulation intensity at time t; τ adapt is an adaptive adjustment time constant.

[0033] In an optional mode, the brain-muscle electric cooperative decoding module analyzes the sublingual vein texture features output by the tongue image acquisition module in real time through a convolutional neural network (CNN) to extract a tongue image attention weight graph to obtain a tongue image feature attention vector;

[0034] The attention vector is subjected to a tensor product operation with an EEG motor imagery potential to generate an enhanced motor imagery feature matrix;

[0035] A long short-term memory (LSTM) network is used to perform spatio-temporal sequence modeling on the enhanced feature matrix to output an acupoint activation time sequence atlas corresponding to a preset hand movement.

[0036] According to another aspect of the present application, an auxiliary stroke rehabilitation device based on a multi-source physiological signal wearable device is provided, comprising:

[0037] A physiological signal monitoring module is configured to monitor pulse waveform variability and sublingual vein feature parameters of a subject through a self-powered five-organ monitoring wristband, wherein the self-powered five-organ monitoring wristband comprises a flexible piezoelectric sensor, a photoplethysmogram (PPG) sensor and a tongue image acquisition module.

[0038] A personalized rehabilitation parameter generation module is configured to dynamically evaluate an organ imbalance index of a subject according to the pulse waveform variability and the sublingual vein feature parameters; and generate personalized rehabilitation adjustment parameters according to the organ imbalance index of the subject, wherein the personalized rehabilitation adjustment parameters comprise acupoint selection, stimulation intensity, stimulation timing, electric pulse stimulation intensity, stimulation frequency, action time and VR scene interaction mode of a cooperative training closed loop of an acupoint activation time sequence atlas.

[0039] A brain-muscle electro-coupling decoding module is configured to synchronously collect electroencephalogram (EEG) signals of a subject and surface electromyogram (sEMG) signals of a target muscle, analyze a nonlinear coupling relationship between motor imagery potentials of the EEG signals and the sEMG signals through the brain-muscle electro-coupling decoding module, and generate an acupoint activation timing atlas corresponding to a preset hand movement;

[0040] A VR guiding and acupoint stimulation module is configured to optimize the acupoint activation timing atlas according to the personalized rehabilitation adjustment parameters, to adjust acupoint selection, stimulation intensity and stimulation timing, present a built-in stomach meridian virtual acupuncture scene to the subject through a portable VR guiding instrument, provide real-time tactile feedback corresponding to the preset hand movement through bone conduction earphones, and guide the subject to perform a preset rehabilitation movement.

[0041] According to still another aspect of the present application, a computing device is provided, comprising a processor, a memory, a communication interface and a communication bus, the processor, the memory and the communication interface being in communication with each other through the communication bus;

[0042] The memory is configured to store at least one executable instruction, and the executable instruction is configured to enable the processor to perform operations corresponding to the above-mentioned stroke rehabilitation assisting method based on a multi-source physiological signal wearable device.

[0043] According to the scheme provided by the application, the pulse waveform variation rate and the sublingual vein characteristic parameter of a subject are monitored by a self-powered five-zang monitoring wristband, the self-powered five-zang monitoring wristband comprising a flexible piezoelectric sensor, a photoplethysmographic pulse wave (PPG) sensor and a tongue appearance acquisition module; the visceral imbalance index of the subject is dynamically evaluated according to the pulse waveform variation rate and the sublingual vein characteristic parameter; and the individualized rehabilitation adjustment parameter is generated according to the visceral imbalance index of the subject, wherein the individualized rehabilitation adjustment parameter comprises acupoint selection, stimulation intensity, stimulation timing of an acupoint activation timing map, stimulation intensity, stimulation frequency, action time of an electric pulse stimulation in a closed loop of synergistic training and VR scene interaction mode; the electroencephalogram (EEG) of the subject and the surface electromyogram (sEMG) of a target muscle are synchronously acquired, the motor imagery potential of the EEG and the nonlinear coupling relationship of the sEMG are analyzed by a brain-muscle electrical synergy decoding module, and an acupoint activation timing map corresponding to a preset hand movement is generated; the acupoint activation timing map is optimized according to the individualized rehabilitation adjustment parameter, so as to adjust the acupoint selection, stimulation intensity and stimulation timing; and a built-in stomach meridian virtual acupuncture scene of foot yangming is presented to the subject through a portable VR guide instrument, and real-time tactile feedback corresponding to the preset hand movement is provided through a bone conduction earphone, so as to guide the subject to perform a preset rehabilitation action. The application integrates brain-muscle electrical synergy decoding, individualized parameter adjustment, VR and visceral imbalance diagnosis into the rehabilitation scheme, and improves the rehabilitation effect of stroke rehabilitation. Specifically, the visceral imbalance index is dynamically evaluated according to the pulse waveform variation rate and the sublingual vein characteristic parameter of the subject, and the individualized rehabilitation adjustment parameter is generated, so as to ensure the pertinence of the rehabilitation scheme. The nonlinear coupling relationship of EEG and sEMG is analyzed by the brain-muscle electrical synergy decoding module, the relationship between motor imagery and actual movement is more effectively understood, and thus the acupoint activation timing map corresponding to the preset hand movement is more accurately generated, and the efficiency of rehabilitation training is improved. The acupoint activation timing map is optimized according to the individualized rehabilitation adjustment parameter, the acupoint selection, stimulation intensity and stimulation timing are adjusted, the real-time state of the patient can be dynamically adjusted, and a better treatment effect is achieved. The virtual acupuncture scene is presented through the portable VR guide instrument, and the tactile feedback is provided through the bone conduction earphone, so as to improve the participation and immersion of the patient in the training, enhance the motivation of the patient, and help the patient to better complete the rehabilitation action. The concept of visceral imbalance is introduced, and the traditional Chinese medicine diagnosis methods such as tongue appearance and pulse appearance are integrated into the rehabilitation scheme, which embodies the idea of combining traditional Chinese and Western medicine, and is more in line with the needs of Chinese patients.

[0044] The above description is only a summary of the technical scheme of the application, in order to more clearly understand the technical means of the application, the content of the specification can be implemented, and in order to make the above and other purposes, characteristics and advantages of the application more obvious and easy to understand, the specific embodiments of the application are as follows. BRIEF DESCRIPTION OF DRAWINGS

[0045] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings:

[0046] Figure 1 A flowchart illustrating an embodiment of the stroke rehabilitation method based on a wearable device for multi-source physiological signals is shown.

[0047] Figure 2 This illustration shows an embodiment of the stroke rehabilitation assisted by the present invention. Figure 1 ;

[0048] Figure 3 This illustration shows an embodiment of the stroke rehabilitation assisted by the present invention. Figure 2 ;

[0049] Figure 4 This diagram illustrates the framework of an assisted stroke rehabilitation device based on a wearable device with multi-source physiological signals, according to an embodiment of the present invention.

[0050] Figure 5 A schematic diagram of the structure of a computing device according to an embodiment of the present invention is shown. Detailed Implementation

[0051] Exemplary embodiments of the invention will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the invention are shown in the drawings, it should be understood that the invention may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this invention will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art.

[0052] Figure 1 This diagram illustrates a flowchart of an assisted stroke rehabilitation method based on a wearable device using multi-source physiological signals, according to an embodiment of the present invention. Specifically, as shown... Figure 1 As shown, it includes the following steps:

[0053] Step S101: Monitor the pulse waveform variation rate and sublingual vein characteristic parameters of the subject using a self-powered five-viscera monitoring wristband. The self-powered five-viscera monitoring wristband includes a flexible piezoelectric sensor, a photoplethysmography (PPG) sensor, and a tongue image acquisition module.

[0054] In this embodiment, the flexible piezoelectric sensor is used to monitor the pulse waveform variability, by capturing the small changes in the blood vessel wall, reflecting the health status of the cardiovascular system. The photoplethysmography (PPG) sensor is used to monitor the blood oxygen saturation and heart rate variability, providing real-time data of blood circulation. The tongue image acquisition module includes a high-resolution camera and a light source, which is used to acquire the sublingual vein image to extract the sublingual vein feature parameters. The flexible piezoelectric sensor, PPG sensor and tongue image acquisition module are integrated into a wearable device through a wristband, for example, the flexible piezoelectric sensor is arranged on the inner side of the wristband, and the radial pulse is obvious, usually on the side close to the thumb of the wrist joint. In this embodiment, the piezoelectric sensor in the self-powered five-zang monitoring wristband uses a PVDF piezoelectric film with a thickness of 50 microns, which is cut into a strip shape of 5mm x 20mm and attached to a flexible polyimide substrate. The PPG sensor selects the SFH 7050 PPG sensor of Osram, which integrates a red light LED (660nm) and a photodiode. The light source of the tongue image acquisition module is a ring-shaped LED lamp, which contains 12 warm white LEDs (color temperature 4000K). The camera and LED lamp are fixed on a rotatable support, and the user can adjust the angle, and the image is transmitted to the mobile phone APP through Bluetooth. A 36AWG silver-plated copper wire is used to connect all the sensors, and epoxy resin is used for insulation packaging. A 5cm x 5cm flexible solar cell is integrated on the surface of the wristband to power the PPG sensor and part of the circuit of the tongue image acquisition module. At the same time, the energy is collected by a micro electromagnetic generator through wrist movement and stored in a super capacitor. The mobile phone APP displays the pulse waveform, heart rate, blood oxygen saturation and tongue image.

[0055] In step S102, the zang-fu imbalance index of the subject is dynamically evaluated according to the pulse waveform variability and the sublingual vein feature parameters; and a personalized rehabilitation adjustment parameter is generated according to the zang-fu imbalance index of the subject, wherein the personalized rehabilitation adjustment parameter includes acupoint selection, stimulation intensity, stimulation timing, stimulation frequency, action time and VR scene interaction mode of the acupoint activation timing map in the closed loop of synergistic training and electric pulse stimulation intensity.

[0056] In this embodiment, the pulse waveform contains rich cardiovascular physiological information, and its variability reflects the dynamic changes of heart function, vascular elasticity, and neural and humoral regulation. By analyzing the pulse waveform, the functional status of the heart, liver, spleen, lungs, and kidneys can be objectively quantified and evaluated. The morphology, color, thickness, and degree of varicosity of the sublingual vein are closely related to the movement of qi and blood. By extracting the objective characteristic parameters of the sublingual vein through image recognition technology, the pathological state of qi stagnation and blood stasis, and phlegm-dampness resistance can be judged. By continuously monitoring the changes in the pulse waveform and the sublingual vein, the dynamic trends of the functions of the zang-fu organs can be understood, and the evolution of the disease can be more accurately grasped. Traditional acupuncture emphasizes syndrome differentiation and treatment, and the selection of acupoints along the meridians. Combined with the zang-fu imbalance index, a personalized acupoint activation time sequence map is generated, and the selection of acupoints, stimulation intensity, and stimulation time sequence parameters are determined, which improves the effectiveness of acupoint stimulation. The combination of electrical pulse stimulation and VR scene interaction mobilizes the active participation of patients and promotes the recovery of neuromuscular function. The intensity and frequency of electrical pulse stimulation are dynamically adjusted according to the real-time physiological response of the patient, achieving precise intervention. Since stroke patients often have disorders of the heart, liver, spleen, and kidneys (such as palpitations and fatigue due to heart qi deficiency, emotional fluctuations due to liver fire caused by liver stagnation, and heavy limbs due to spleen deficiency and dampness), this embodiment can objectively quantify and evaluate the zang-fu function status of stroke patients, providing a basis for developing personalized rehabilitation programs. For the common problems of motor disorders, language disorders, and cognitive disorders in stroke patients, a personalized acupoint activation time sequence map and collaborative training program are generated by combining the zang-fu imbalance index, further improving the rehabilitation effect. For motor disorders, the following acupoints are selected: shoulder, Quchi, Shousanli, Hegu for the upper limbs, and Huantiao, Fengshi, Zusanli, and Yanglingquan for the lower limbs. For language disorders, the following acupoints are selected: Lianquan, Yumen, and Tongli. For cognitive disorders, the following acupoints are selected: Baihui, Sishencong, and Shenmen. At the same time, according to the zang-fu imbalance index, the corresponding acupoints are selected. For heart qi deficiency, the stimulation of Xinshu and Neiguan is strengthened. For liver yang hyperactivity, the stimulation of Taichong and Xingjian is strengthened. For phlegm-dampness blocking the collaterals, the stimulation of Fenglong and Yiningquan is strengthened. For stimulation intensity, the appropriate intensity of acupoint stimulation is determined according to the patient's age, constitution, and disease condition. Stroke patients may feel sluggish, so the stimulation intensity should not be too high to avoid discomfort. According to the principles of "opening acupoints" and "reinforcing and reducing" in traditional Chinese medicine, a time sequence program for acupoint stimulation is developed. For example, stimulating acupoints in the distal part of the limbs first, and then stimulating acupoints in the proximal part, to promote the circulation of qi and blood. The intensity of electrical pulse stimulation is 1-3 mA (adjusted according to the patient's tolerance, starting from low intensity and gradually increasing), and the frequency is 20-50 Hz (to promote muscle contraction) or 2 Hz (to calm and relieve pain), with an action time of 20-30 minutes. For motor disorders, the VR scene selects the kitchen, living room, and park. For language rehabilitation, it simulates dialogue scenes such as shopping, asking for directions, and seeking medical treatment. For cognitive rehabilitation, it simulates memory games and attention training.The patient completes tasks in the VR scene by controlling the movement of the limbs, such as picking up and putting down objects, walking, going up and down stairs, etc. The patient has a conversation with a virtual character in the VR scene, practicing pronunciation, speech speed, and expression ability. In addition, the patient completes tasks such as memory matching and spatial reasoning in the VR scene. Figure 2 , Figure 3 As shown in FIGS. 1-3, a stroke patient with right hemiplegia and symptoms of heart qi deficiency and blood stasis is selected. The acupoints are selected as Jianyu (right), Quchi (right), Shousanli (right), Hegu (right), Huantiao (right), Zusanli (right), Yanglingquan (right), Xinshu, and Neiguan. The stimulation intensity is 1-2 mA (electrical pulse), and the finger pressure massage is performed until the patient feels sore and swollen. The stimulation timing is to stimulate the acupoints at the distal end of the upper and lower limbs (such as Hegu and Yanglingquan) first, then stimulate the acupoints at the proximal end (such as Jianyu and Huantiao), and finally stimulate Xinshu and Neiguan. The intensity of the electrical pulse stimulation is 1.5 mA, the frequency is 30 Hz, and the action time is 25 minutes. The VR scene simulates a kitchen scene, and the patient needs to use the right hand to complete simple cooking tasks (picking up and putting down bowls and chopsticks, pouring water, and stirring ingredients). The interaction mode is: the patient maps the limb movement to the VR scene by wearing a motion capture device, controls the virtual character to complete the cooking task, records the movement trajectory and completion time of the patient, and provides feedback. The patient follows the above rehabilitation program for acupoint stimulation and VR scene interaction training. The patient's heart rate, blood pressure, and muscle activity are monitored in real time, and the patient's right limb motor function is regularly evaluated (such as using the Fugl-Meyer Assessment Scale). According to the evaluation results, the rehabilitation program is adjusted. If the patient's limb motor function improves, the difficulty of the VR scene can be increased, for example, the complexity of the task can be increased or assistance can be reduced.

[0057] In an alternative way, the formula for calculating the imbalance index of the Zangfu organs is:

[0058]

[0059] where w i is the weight coefficient of the i-th Zangfu organ; P i is the real-time monitored pulse waveform variability; P norm,i is the normal variability threshold value corresponding to the Zangfu organ; TongueBaseline i is the tongue feature value of the i-th Zangfu organ; TongueFeature i is the tongue baseline value of the i-th Zangfu organ; P crit,i is the critical variability threshold value of the i-th Zangfu organ; Θ(.) is the Heaviside step function; t is the current time; τ is the time integral variable.

[0060] In this embodiment, for stroke patients, the weight coefficient can be adjusted, for example, the weight of the heart and liver is increased, because stroke is usually related to cardiovascular and cerebrovascular diseases and liver dysfunction. At the same time, according to the specific situation of the patient, the normal threshold and critical threshold are adjusted.

[0061] In an alternative way, the electric pulse stimulation parameter adopts a heart rate variability HRV feedback model, wherein the expression of the heart rate variability HRV feedback model is:

[0062]

[0063] wherein I(t) is the electric pulse stimulation intensity at time t; Kp is the proportional gain coefficient; is the center frequency; φ HRV (t) is the HRV phase modulation function; HRV LF / HF (t) is the low frequency LF / high frequency HF ratio of the heart rate variability; PWV(t) is the pulse wave velocity at time t; PWV base is the PWV value of the patient in the normal state; f c is the center frequency.

[0064] In this embodiment, the HRV feedback can effectively regulate the autonomic nervous system, balance the activity of sympathetic and parasympathetic nerves, and help to improve the blood circulation of the brain and promote the recovery of neural function. This embodiment constitutes a closed-loop control system, which monitors the physiological indicators of the patient in real time and adjusts the stimulation parameters according to these indicators to ensure that the stimulation effect is always in the best state. By adjusting the proportional gain coefficient, the intensity of the electric pulse stimulation is limited to avoid damage to the patient caused by excessive stimulation.

[0065] In an alternative way, the acupoint activation time sequence atlas obtains a dynamic gain function by recursive least squares RLS real-time modeling of the phase locking value PLV between the mu rhythm and motor imagination potential of the electromyogram sEMG; wherein the dynamic gain function is:

[0066]

[0067] wherein G base is the basic gain coefficient; κ is the PLV gain adjustment coefficient; PLV max is the preset maximum phase locking value; HRV LF / HF,base is the heart rate variability baseline value of the patient in the resting state;

[0068] The electric pulse stimulation intensity parameter is adjusted in real time according to the dynamic gain function, and the adjustment formula is:

[0069]

[0070] wherein t onset is the stimulation start time; τ adapt is the adaptive adjustment time constant for smoothing the instantaneous impact of gain change on stimulation intensity; I(t) is the unadjusted electrical pulse stimulation intensity at time t; τ adapt is the adaptive adjustment time constant.

[0071] In this embodiment, the phase locking value (PLV) between the mu rhythm and motor imagery potential of the electromyographic signal (sEMG) is modeled in real time by recursive least squares (RLS), and the electrical pulse stimulation intensity is dynamically adjusted to more accurately respond to changes in the patient's physiological state. The dynamic gain function takes into account the patient's basic gain coefficient, PLV gain adjustment coefficient, and heart rate variability (HRV) personalized parameters, making the treatment plan more targeted. By introducing an adaptive adjustment time constant, the instantaneous impact of gain change on stimulation intensity is smoothed, avoiding drastic fluctuations in stimulation intensity and improving the safety of treatment.

[0072] In step S103, the electroencephalogram (EEG) of the subject and the surface electromyogram (sEMG) of the target muscle are synchronously collected, the nonlinear coupling relationship between the motor imagery potential of the EEG and the sEMG is analyzed by a brain-muscle electrical cooperative decoding module, and an acupoint activation timing atlas corresponding to a preset hand movement is generated.

[0073] In this embodiment, the electroencephalogram (EEG) and the surface electromyogram (sEMG) of the target muscle are synchronously collected, and the nonlinear coupling relationship between the motor imagery potential of the EEG and the sEMG is analyzed by a brain-muscle electrical cooperative decoding module to more accurately identify the patient's motor intention. The acupoint activation timing atlas generated based on the patient's own physiological data through analysis of the brain-muscle electrical cooperative relationship better matches the individual differences of the patient, thereby realizing a personalized treatment plan. Motor imagery combined with acupoint activation stimulation promotes the activation of the motor area of the cerebral cortex and strengthens the connection between the nerves and muscles, which helps to restore motor function. In addition, both EEG and sEMG are non-invasive techniques and do not cause trauma to the patient.

[0074] In an alternative manner, the brain-muscle electrical cooperative decoding module analyzes the sublingual vein texture features output by the tongue image acquisition module in real time through a convolutional neural network (CNN) to extract a tongue image attention weight map and obtain a tongue image feature attention vector;

[0075] The attention vector is subjected to a tensor product operation with the motor imagery potential of the EEG to generate an enhanced motor imagery feature matrix;

[0076] The enhanced feature matrix is subjected to spatio-temporal sequence modeling by a long short-term memory network (LSTM) to output an acupoint activation timing atlas corresponding to a preset hand movement.

[0077] In this embodiment, the tongue image attention vector is multiplied by the EEG motor imagery potential to generate an enhanced motor imagery feature matrix, effectively fusing tongue image information and improving the performance of motor intention decoding. Tongue image acquisition is also a non-invasive method. Combined with non-invasive EEG and EMG, the overall scheme has less burden on patients and high compliance. For example, stroke patients with impaired left limb motor function need to use brain-EMG collaborative decoding combined with tongue image features to assist in activating acupuncture points to promote rehabilitation. Place EEG cap on patient's scalp to collect EEG signals, and place sEMG electrodes on the flexor carpi radialis and flexor carpi ulnaris of the patient's left forearm to collect sEMG signals. Use the tongue image acquisition module to take the patient's tongue image, and instruct the patient to perform a specific motor imagery task (such as imagining left hand clenching or stretching). Use the pre-trained CNN model to analyze the patient's tongue image, extract the sublingual vein texture features, and generate a tongue image attention weight map and a tongue image feature attention vector. Multiply the tongue image feature attention vector by the EEG motor imagery potential to generate an enhanced motor imagery feature matrix. Use the trained LSTM model to model the enhanced feature matrix in time and space, and predict the acupuncture point activation time sequence atlas corresponding to the left hand clenching action. This atlas indicates that the hand Laogong point is stimulated first, then the Hegu point, and specifies the stimulation time and intensity of each point. Use the electric stimulation device to stimulate the Laogong and Hegu points of the patient's left hand according to the generated acupuncture point activation time sequence atlas. During the stimulation process, the patient continues to perform motor imagery and attempts to clench the fist. Monitor the patient's EEG, sEMG signals and tongue image changes in real time, and analyze the stimulation effect through the brain-EMG collaborative decoding module. If the stimulation effect is found to be poor, adjust the acupuncture point activation time sequence atlas, for example, adjust the stimulation intensity or change the stimulation order.

[0078] Step S104, optimizing the acupuncture point activation time sequence atlas according to the personalized rehabilitation adjustment parameters to adjust the acupuncture point selection, stimulation intensity and stimulation time sequence; presenting a built-in stomach meridian of foot yangming virtual acupuncture scene through a portable VR guide instrument, and providing real-time tactile feedback of the preset hand movement through bone conduction earphones to guide the subject to perform the preset rehabilitation movement.

[0079] In this embodiment, a portable VR guide instrument is used to present a built-in stomach meridian of foot yangming virtual acupuncture scene, enhancing the patient's immersion and improving the quality of motor imagery and promoting the activation of the motor area of the cerebral cortex. Real-time tactile feedback of the preset hand movement is provided through bone conduction earphones to provide a real motor experience for the patient and promote the remodeling of the sensorimotor cortex. Combined with vision (VR), hearing (bone conduction earphones) and touch (electric stimulation), a multi-sensory stimulation is formed to more effectively activate the cerebral cortex and improve the rehabilitation effect. VR guidance and tactile feedback enhance the patient's sense of participation and interest, and improve the compliance of rehabilitation training. The portable VR guide instrument and bone conduction earphones make it convenient for patients to perform rehabilitation training at home or in community settings.

[0080] In an alternative way, the presenting the built-in stomach meridian virtual acupuncture scene to the subject through the portable VR guide further comprises:

[0081] When the subject completes the preset hand movement in the VR scene and meets the trigger condition based on heart rate variability HRV rhythm, triggering the personalized electric pulse stimulation according to the zang-fu imbalance index to stimulate the preset acupoint;

[0082] At the same time, the portable VR guide presents a dynamic blood flow path corresponding to the preset acupoint.

[0083] In this embodiment, the heart rate variability (HRV) rhythm and the zang-fu imbalance index are combined to trigger personalized electric pulse stimulation, realizing the transformation from "thousands of people" to "one person", and improving the accuracy of treatment. The hand movement in the VR scene as input, HRV as physiological feedback, electric pulse stimulation as intervention, and dynamic blood flow path as visual feedback form a closed-loop control system, which helps patients to enhance the confidence in treatment.

[0084] In an alternative way, the VR scene interaction method renders the light and shadow changes of the blood flow path in real time through the Shader Graph method, wherein the color gradient of the blood flow path is positively correlated with the current heart rate variability frequency energy ratio;

[0085] And output spatialized audio through bone conduction earphones, wherein the left / right channel phase difference is linearly related to the EEG mu rhythm energy offset amount, and when the patient's motor imagination potential is enhanced, the audio sound image automatically shifts to the opposite side;

[0086] The motion trajectory is captured by the inertial measurement unit (IMU), and the kinematics parameters of the virtual limbs are mapped in real time in the VR scene to form a somatosensory-visual closed loop.

[0087] In this embodiment, Shader Graph allows developers to create complex shaders in a graphical way, and the rendering of the blood flow path can be more realistic and detailed, and can more intuitively show the state of blood flow. The color of the blood flow path is associated with the HRV frequency energy ratio, so that the patient can intuitively understand the state of his own nervous autonomic regulation through visual feedback. For example, when the sympathetic nerve is active, the color is reddish, and when the parasympathetic nerve is active, the color is bluish. The bone conduction earphone directly transmits sound to the inner ear, reducing the interference of environmental noise. The spatialization of audio is associated with the EEG mu rhythm, and the patient's motor imagery is used to control the audio image, which can enhance the patient's active participation. When the motor imagery potential is enhanced, the audio image automatically shifts to the opposite side, encouraging the patient to actively perform motor imagery training. When the patient imagines left limb movement, the audio image will shift to the right side, and vice versa, which helps to promote the neural plasticity of the cerebral cortex. The IMU can accurately capture information such as the patient's motion, position, posture, and speed. The motion captured by the IMU is mapped in real time to the virtual limbs in the VR scene, allowing the patient to feel that his limbs are moving, even if the actual limb movement is limited, promoting the brain's perception of the limbs, and helping to improve motor control and coordination. Among them, the VR headset is a high-refresh-rate, low-latency VR headset, and the hand motion capture device is Leap Motion or a VR handle, which is used to capture hand motion. The heart rate sensor uses a heart rate strap or a finger pulse sensor to collect heart rate data. The electroencephalogram acquisition device (EEG) can be a dry electrode or a wet electrode. The inertial measurement unit (IMU) uses multiple IMU sensors, which are worn on the patient's limbs to capture motion trajectories. The bone conduction earphone is used to output spatialized audio. As Figure 3 As shown in FIG. 8, the patient with post-stroke sequelae has left limb movement disorder. The VR scene simulates a kitchen scene, and the patient needs to complete tasks such as washing vegetables, cutting vegetables, and frying in the VR scene, that is, the patient needs to imagine the left hand to complete the actions such as washing vegetables, cutting vegetables, and frying. The patient tries to complete the corresponding action with the left hand, and the IMU sensor captures the motion trajectory. The patient sees the virtual left hand completing the corresponding action in the VR scene. When the patient imagines the left hand movement, the audio image shifts to the right side, encouraging the patient to continue motor imagery. According to the HRV frequency energy ratio, the color and light of the virtual left hand blood flow path are adjusted in real time. Through the combination of motor imagery, action execution, visual feedback, and auditory feedback, the motor function recovery of the left limb is promoted, and the patient's self-care ability is improved.

[0088] In an optional manner, the state space of the preset rehabilitation action includes PWV, HRV, electromyogram entropy value, and action completion degree parameter;

[0089] The action space of the preset rehabilitation action includes acupoint activation parameter, VR scene complexity parameter, and electrical stimulation parameter;

[0090] The reward function of the preset rehabilitation action is:

[0091]

[0092] wherein, ΔFuglMeyerScore is the increment of the score scale; StandardTime is the standard time required for the rehabilitation action; ActualTime is the time actually spent in completing the rehabilitation action; and NumberOfAbnormalAlerts is the number of abnormal alerts.

[0093] In this embodiment, the parameters of the rehabilitation action are automatically adjusted according to the real-time state and action feedback of the patient. The state space includes PWV (pulse wave velocity, reflecting vascular elasticity), HRV (heart rate variability, reflecting autonomic nervous system function), muscle EMG entropy value (reflecting muscle activity complexity) and movement accuracy parameters, which can comprehensively reflect the physiological state and action quality of the patient. The action space includes acupoint activation parameters (frequency, intensity and duration of electrical stimulation), VR scene complexity parameters (task difficulty and visual stimulation) and electrical stimulation parameters, which provide rich adjustment options. The reward function takes the Fugl-Meyer score increment, action completion efficiency and safety into account, guiding the reinforcement learning algorithm to optimize in the direction of improving rehabilitation effect, improving efficiency and reducing risk. By adjusting the VR scene complexity parameters, the rehabilitation difficulty is adaptively adjusted according to the patient's ability, maintaining the patient's participation and enthusiasm. For example, a patient with post-stroke sequelae has right upper limb movement disorder. State space: PWV = 0.8, HRV_LFHF = 1.5, EMG_Entropy = 0.6, Movement_Accuracy = 0.4; acupoint activation parameters: electrical stimulation frequency = 20 Hz, electrical stimulation intensity = 5 mA, electrical stimulation duration = 0.5 s; VR scene complexity parameters: task difficulty = medium, visual stimulation = high; electrical stimulation parameters: same as acupoint activation parameters; reward function: after one rehabilitation training, the Fugl-Meyer score increment is 2. The standard time required for the rehabilitation action is 10 seconds, and the actual completion time is 12 seconds. No abnormal alert occurs, and the reward value is: R = 2 + (1-10 / 12)-0 = 2.167. The DQN algorithm is used to train the reinforcement learning model, and the patient completes the task of grabbing objects in the VR scene while receiving electrical stimulation. According to the real-time state and action feedback of the patient, the DQN algorithm is used to select the best action parameters and update the model.

[0094] According to the scheme provided in the application, the pulse waveform variation rate and the sublingual vein characteristic parameter of a subject are monitored by a self-powered five-zang monitoring wristband, the self-powered five-zang monitoring wristband comprising a flexible piezoelectric sensor, a photoplethysmographic pulse wave (PPG) sensor and a tongue appearance acquisition module; the visceral imbalance index of the subject is dynamically evaluated according to the pulse waveform variation rate and the sublingual vein characteristic parameter; and the individualized rehabilitation adjustment parameter is generated according to the visceral imbalance index of the subject, wherein the individualized rehabilitation adjustment parameter comprises acupoint selection, stimulation intensity, stimulation timing of an acupoint activation timing map, stimulation intensity, stimulation frequency, action time of an electric pulse stimulation in a closed loop of synergistic training and VR scene interaction mode; the electroencephalogram (EEG) of the subject and the surface electromyogram (sEMG) of a target muscle are synchronously acquired, the motor imagery potential of the EEG and the nonlinear coupling relationship of the sEMG are analyzed by a brain-muscle electrical synergistic decoding module, and an acupoint activation timing map corresponding to a preset hand movement is generated; the acupoint activation timing map is optimized according to the individualized rehabilitation adjustment parameter, so as to adjust the acupoint selection, stimulation intensity and stimulation timing; and a built-in stomach meridian virtual acupuncture scene of foot yangming is presented to the subject through a portable VR guide instrument, and real-time tactile feedback corresponding to the preset hand movement is provided through a bone conduction earphone, so as to guide the subject to perform a preset rehabilitation action. The application integrates brain-muscle electrical synergistic decoding, individualized parameter adjustment, VR and visceral imbalance diagnosis into a rehabilitation scheme, thereby improving the rehabilitation effect of stroke rehabilitation. Specifically, the visceral imbalance index is dynamically evaluated according to the pulse waveform variation rate and the sublingual vein characteristic parameter of the subject, and the individualized rehabilitation adjustment parameter is generated based on the visceral imbalance index, so as to ensure the pertinence of the rehabilitation scheme. The nonlinear coupling relationship of EEG and sEMG is analyzed by the brain-muscle electrical synergistic decoding module, so as to more effectively understand the relationship between motor imagery and actual movement, thereby more accurately generating an acupoint activation timing map corresponding to a preset hand movement, and improving the efficiency of rehabilitation training. The acupoint activation timing map is optimized according to the individualized rehabilitation adjustment parameter, so as to adjust the acupoint selection, stimulation intensity and stimulation timing, and the real-time state of the patient can be dynamically adjusted, so as to achieve a better treatment effect. The virtual acupuncture scene is presented through the portable VR guide instrument, and the tactile feedback is provided through the bone conduction earphone, so as to improve the participation and immersion of the patient in the training, enhance the motivation of the patient, and help the patient to better complete the rehabilitation action. The concept of visceral imbalance is introduced, and the traditional Chinese medicine diagnosis methods such as tongue appearance and pulse appearance are integrated into the rehabilitation scheme, which embodies the idea of combining traditional Chinese and Western medicine, and is more in line with the needs of Chinese patients.

[0095] Figure 4 A framework schematic diagram of an auxiliary stroke rehabilitation device based on a multi-source physiological signal wearable device is shown. The auxiliary stroke rehabilitation device based on the multi-source physiological signal wearable device comprises:

[0096] The physiological signal monitoring module 410 is used to monitor the pulse waveform variation rate and sublingual vein characteristic parameters of the subject through a self-powered five-viscera monitoring wristband. The self-powered five-viscera monitoring wristband includes a flexible piezoelectric sensor, a photoplethysmography (PPG) sensor, and a tongue image acquisition module.

[0097] The personalized rehabilitation parameter generation module 420 is used to dynamically assess the subject's organ imbalance index based on the pulse waveform variation rate and sublingual vein characteristic parameters; and to generate personalized rehabilitation adjustment parameters based on the subject's organ imbalance index, wherein the personalized rehabilitation adjustment parameters include acupoint selection, stimulation intensity, stimulation sequence, electrical pulse stimulation intensity, stimulation frequency, action time, and VR scene interaction mode in the acupoint activation timing map.

[0098] The brain-muscle electrocoagulation decoding module 430 is used to simultaneously acquire the subject's electroencephalogram (EEG) signal and the surface electromyography (sEMG) signal of the target muscle. The brain-muscle electrocoagulation decoding module analyzes the nonlinear coupling relationship between the motor imagery potential of the EEG signal and the surface electromyography (sEMG) signal to generate an acupoint activation time sequence map with preset hand movements.

[0099] The VR guidance and acupoint stimulation module 440 is used to optimize the acupoint activation timing map according to the personalized rehabilitation adjustment parameters, so as to adjust the acupoint selection, stimulation intensity and stimulation timing; present the built-in virtual acupuncture scene of the Stomach Meridian of Foot Yangming to the subject through a portable VR guide device, and provide real-time tactile feedback with the preset hand movements through bone conduction headphones to guide the subject to perform preset rehabilitation movements.

[0100] Figure 5 The diagram shows a structural schematic of an embodiment of the computing device of the present invention. The specific embodiments of the present invention do not limit the specific implementation of the computing device.

[0101] like Figure 5 As shown, the computing device may include: a processor 502, a communications interface 504, a memory 506, and a communications bus 508.

[0102] The processor 502, communication interface 504, and memory 506 communicate with each other via communication bus 508. Communication interface 504 is used to communicate with other network elements such as clients or other servers. The processor 502 executes program 510, specifically performing the relevant steps in the above-described embodiment of the stroke rehabilitation method based on a wearable device with multi-source physiological signals.

[0103] In particular, the program 510 can include program code comprising computer operating instructions.

[0104] The processor 502 can be a central processing unit CPU, or an application specific integrated circuit ASIC, or one or more integrated circuits configured to perform the functions of an embodiment of the application. The one or more processors included in the computing device can be of the same type or different types of processors, such as one or more CPUs and one or more ASICs.

[0105] The memory 506 is used to store the program 510. The memory 506 can include a high speed RAM memory and can also include a non-volatile memory, such as at least one disk memory.

[0106] According to the scheme provided by the application, the pulse waveform variation rate and the sublingual vein characteristic parameter of a subject are monitored by a self-powered five-zang monitoring wristband, the self-powered five-zang monitoring wristband comprising a flexible piezoelectric sensor, a photoplethysmogram (PPG) sensor and a tongue appearance acquisition module; the visceral imbalance index of the subject is dynamically evaluated according to the pulse waveform variation rate and the sublingual vein characteristic parameter; and the individualized rehabilitation adjustment parameter is generated according to the visceral imbalance index of the subject, wherein the individualized rehabilitation adjustment parameter comprises acupoint selection, stimulation intensity, stimulation timing of an acupoint activation timing map, stimulation intensity, stimulation frequency, action time of an electric pulse in a closed loop of synergistic training and VR scene interaction mode; the electroencephalogram (EEG) of the subject and the surface electromyogram (sEMG) of a target muscle are synchronously acquired, the motor imagery potential of the EEG and the nonlinear coupling relationship of the sEMG are analyzed by a brain-muscle electrical synergistic decoding module, and an acupoint activation timing map corresponding to a preset hand action is generated; the acupoint activation timing map is optimized according to the individualized rehabilitation adjustment parameter, so as to adjust the acupoint selection, stimulation intensity and stimulation timing; and a built-in stomach meridian virtual acupuncture scene of foot yangming is presented to the subject through a portable VR guide instrument, and real-time tactile feedback corresponding to the preset hand action is provided through a bone conduction earphone, so as to guide the subject to perform a preset rehabilitation action. The application integrates brain-muscle electrical synergistic decoding, individualized parameter adjustment, VR and visceral imbalance diagnosis into a rehabilitation scheme, thereby improving the rehabilitation effect of stroke rehabilitation. Specifically, the visceral imbalance index is dynamically evaluated according to the pulse waveform variation rate and the sublingual vein characteristic parameter of the subject, and the individualized rehabilitation adjustment parameter is generated based on the visceral imbalance index, so as to ensure the pertinence of the rehabilitation scheme. The nonlinear coupling relationship of the EEG and the sEMG is analyzed by the brain-muscle electrical synergistic decoding module, so as to more effectively understand the relationship between motor imagery and actual action, thereby more accurately generating the acupoint activation timing map corresponding to the preset hand action, and improving the efficiency of rehabilitation training. The acupoint activation timing map is optimized according to the individualized rehabilitation adjustment parameter, so as to adjust the acupoint selection, stimulation intensity and stimulation timing, and the real-time state of the patient can be dynamically adjusted to achieve a better treatment effect. The virtual acupuncture scene is presented through the portable VR guide instrument, and the tactile feedback is provided through the bone conduction earphone, so as to improve the participation and immersion of the patient in the training, enhance the motivation of the patient, and help the patient to better complete the rehabilitation action. The concept of visceral imbalance is introduced, and the tongue appearance and pulse appearance are integrated into the rehabilitation scheme, thereby embodying the idea of combining traditional Chinese medicine with Western medicine, and being more in line with the needs of Chinese patients.

[0107] Those skilled in the art will appreciate that the modules in the apparatuses in the embodiments can be adapted and placed in one or more apparatuses other than the embodiments. The modules or units or components in the embodiments can be combined into one module or unit or component, and furthermore can be divided into multiple sub-modules or sub-units or sub-components. Any combination of all the features disclosed in the present specification (including the accompanying claims, abstract and drawings), and any method or apparatus so disclosed, can be taken in any combination, except that at least some of such features and / or processes or units are mutually exclusive, unless explicitly stated otherwise. Each feature disclosed in the present specification (including the accompanying claims, abstract and drawings) can be replaced by alternative features serving the same, equivalent or similar purpose, unless explicitly stated otherwise. Furthermore, the skilled person will appreciate that the combination of features of different embodiments implies that the features of the different embodiments are meant to be combined, unless explicitly stated otherwise. For example, in the claims below, any of the embodiments can be used in any combination. The application can be implemented by means of hardware comprising several distinct elements, and by means of a suitably programmed computer. In the unitary claim, several of the devices mentioned in the embodiments can be implemented by means of one and the same hardware item. The steps of the above-described embodiments, unless explicitly stated otherwise, are not to be understood as having to be carried out in the order in which they are described.

Claims

1. An auxiliary stroke rehabilitation device based on multi-source physiological signal wearable device, characterized in that, The method comprises the following steps: a physiological signal monitoring module for monitoring the pulse waveform variation rate and the sublingual vein characteristic parameters of a subject through a self-powered five-zang monitoring wristband, the self-powered five-zang monitoring wristband comprising a flexible piezoelectric sensor, a photoplethysmographic pulse wave (PPG) sensor, and a tongue image acquisition module; a personalized rehabilitation parameter generation module for dynamically evaluating the zang-organ imbalance index of the subject according to the pulse waveform variation rate and the sublingual vein characteristic parameters; and generating personalized rehabilitation adjustment parameters according to the zang-organ imbalance index of the subject, wherein the personalized rehabilitation adjustment parameters comprise acupoint selection, stimulation intensity, stimulation timing of an acupoint activation timing map, stimulation intensity, stimulation frequency, and action time of an electric pulse stimulation in a closed loop of synergistic training, and VR scene interaction mode; a brain-muscle electrical synergy decoding module for synchronously collecting electroencephalogram (EEG) signals and surface electromyogram (sEMG) signals of a target muscle of the subject, analyzing the nonlinear coupling relationship between the motor imagery potential of the EEG signals and the sEMG signals through the brain-muscle electrical synergy decoding module, and generating an acupoint activation timing map corresponding to a preset hand movement; a VR guide and acupoint stimulation module for optimizing the acupoint activation timing map according to the personalized rehabilitation adjustment parameters to adjust the acupoint selection, stimulation intensity, and stimulation timing; presenting a built-in stomach meridian virtual acupuncture scene of the foot yangming to the subject through a portable VR guide instrument, providing real-time tactile feedback corresponding to the preset hand movement through bone conduction earphones, and guiding the subject to perform a preset rehabilitation action; the portable VR guide instrument further comprises: when the subject completes the preset hand movement in the VR scene and meets the trigger condition based on heart rate variability (HRV) rhythm, triggering personalized electric pulse stimulation to stimulate a preset acupoint according to the zang-organ imbalance index; at the same time, presenting a dynamic blood flow path corresponding to the preset acupoint through the portable VR guide instrument; the electric pulse stimulation intensity adopts a heart rate variability (HRV) feedback model, and the expression of the HRV feedback model is: Wherein, I(t) is the electric pulse stimulation intensity at time t; Kp is the proportional gain coefficient; is the center frequency; φ HRV (t) is the HRV phase modulation function; HRV LF / HF (t) is the low frequency LF / high frequency HF ratio of heart rate variability; PWV(t) is the pulse wave conduction velocity at time t; PWV base is the PWV value of the patient in normal state; f c is the center frequency.

2. The multi-source physiological signal based wearable device assisted stroke rehabilitation device according to claim 1, wherein, the calculation formula of the zang-organ imbalance index is: where w i is the weight coefficient of the ith Zangfu organ; P i is the real-time monitored pulse waveform variability; P norm,i is the normal variability threshold value corresponding to the Zangfu organ; TongueBaseline i is the tongue feature value of the ith Zangfu organ; TongueFeature i is the tongue baseline value of the ith Zangfu organ; P crit,i is the critical variability threshold value of the ith Zangfu organ; Θ(.) is the Heaviside step function; t is the current time; and τ is the time integral variable.

3. The multi-source physiological signal based wearable device assisted stroke rehabilitation device according to claim 1, wherein, the VR scene interaction mode uses a Shader Graph method to render the light and shadow changes of the blood flow path in real time, wherein the color gradient of the blood flow path is positively correlated with the current heart rate variability frequency energy ratio; and outputs spatialized audio through bone conduction earphones, wherein the left / right channel phase difference is linearly related to the EEG mu rhythm energy offset amount, and when the motor imagery potential of the patient is enhanced, the audio sound image automatically shifts to the opposite side; the action trajectory is captured through an inertial measurement unit (IMU), and the kinematics parameters of a virtual limb are mapped in real time in the VR scene to form a proprioception-visual closed loop.

4. The multi-source physiological signal based wearable device assisted stroke rehabilitation device of claim 1, wherein, The state space of the preset rehabilitation action comprises PWV, HRV, electromyogram entropy value, and action completion degree parameter; the action space of the preset rehabilitation action comprises acupoint activation parameter, VR scene complexity parameter, and electric stimulation parameter; the reward function of the preset rehabilitation action is: Wherein, the ΔFuglMeyerScore is the increment of the score scale; the StandardTime is the standard time required for the rehabilitation action; the ActualTime is the time actually spent to complete the rehabilitation action; and the NumberOfAbnormalAlerts is the number of abnormal alerts.

5. The multi-source physiological signal based wearable device assisted stroke rehabilitation device of claim 1, wherein, The acupoint activation time sequence map obtains a dynamic gain function by recursive least squares (RLS) real-time modeling of the phase locking value (PLV) between the mu rhythm and the motor imagery potential of the sEMG; wherein the dynamic gain function is: Wherein, G base is a basic gain coefficient; κ is a PLV gain adjustment coefficient; PLV max is a preset maximum phase locking value; HRV LF / HF,base is a heart rate variability baseline value of the patient in a resting state; The electric pulse stimulation intensity parameter is real-time adjusted according to the dynamic gain function, and the adjustment formula is: where t onset is the time of stimulus onset; τ adapt is an adaptive adjustment time constant used to smooth the transient impact of gain changes on stimulus intensity; I(t) is the unadjusted electrical pulse stimulus intensity at time t; τ adapt is an adaptive adjustment time constant.

6. The multi-source physiological signal based wearable device assisted stroke rehabilitation apparatus according to claim 1, wherein, The brain-muscle electric cooperative decoding module real-time analyzes the sublingual vein texture features output by the tongue image acquisition module by a convolutional neural network (CNN) to extract a tongue image attention weight graph to obtain a tongue image feature attention vector; The attention vector is subjected to a tensor product operation with the EEG motor imagery potential to generate an enhanced motor imagery feature matrix; The enhanced motor imagery feature matrix is subjected to space-time sequence modeling by a long short-term memory (LSTM) network to output an acupoint activation time sequence map corresponding to a preset hand action.

7. A computing device comprising: A processor, a memory, a communication interface and a communication bus, the processor, the memory and the communication interface complete communication with each other through the communication bus; The memory is used for storing at least one executable instruction, and the executable instruction makes the processor execute the operation corresponding to the auxiliary stroke rehabilitation device based on the multi-source physiological signal wearable device in any one of claims 1-6.

Images