Biofeedback electrical stimulation regulation and control device based on multi-modal physiological signal monitoring and control method

Abstract

The invention provides a biofeedback type electrical stimulation regulation and control device based on multi-mode physiological signal monitoring and a control method, the regulation and control device comprises an electronic component layer, a circuit layer, a man-machine interaction part and an adhesion layer, physiological signal acquisition and electrical stimulation regulation technologies are utilized, and a multi-source biosensor and an intelligent processing module are integrated, so that the regulation and control of the biofeedback type electrical stimulation regulation and control device is realized. Dynamic monitoring and real-time analysis of physiological signals are achieved, self-adaptive adjustment of personalized parameters is conducted based on signal characteristics, and a non-invasive biological feedback regulation and control system is formed. Through integrated multi-mode detection and intelligent regulation and control functions, a non-invasive and convenient regulation and control means is provided for wearers with abnormal nerve or physiological states; a portable and comfortable attached structure is adopted, the device can be worn at the Neiguan point for a long time, physiological signal continuous monitoring and electrical stimulation closed-loop feedback adjustment are achieved, and the personalized health management requirement of a user is met.

Description

Technical Field

[0001] The present invention relates to the technical field of wearable medical electronic equipment, and in particular to a biofeedback electrical stimulation control device and a control method based on multimodal physiological signal monitoring. Background Art

[0002] In daily life, many patients suffer from pain, motion sickness and other problems that affect their normal lives. Acute pain or motion sickness is particularly significant, as the negative impact of symptoms on physical and mental health is particularly pronounced. In recent years, non-invasive methods for regulating abnormal neurological or physiological states have gained attention in the medical field. Transcutaneous electrical nerve stimulation (TENS), as a non-invasive means of regulation, has been shown to have significant therapeutic effects in alleviating abnormal neurological or physiological states. However, traditional TENS devices typically require electrodes to be placed directly on the abnormal area for electrical stimulation. These devices are bulky and difficult to carry, making them inconvenient for patients to wear and adjust independently.

[0003] Although some of the existing electrical stimulation devices on the market have gradually acquired portability and adaptive adjustment functions, they still have limitations in terms of intelligence and personalization. For example, some bracelets or patch-type electrical stimulation devices have a single function, lack real-time monitoring of the patient's physiological state, and only provide electrical stimulation of a fixed frequency or intensity, and cannot be adaptively adjusted according to changes in the patient's symptoms. In addition, although some devices have a physiological signal feedback mechanism, they only adjust the intensity of electrical stimulation based on a single physiological signal (such as skin conductance), and fail to combine multimodal information such as heart rate variability and movement status, thereby reducing the accuracy of the assessment of abnormal neural or physiological states.

[0004] With the development of physiological signal detection technology, more and more studies have shown that physiological signals such as heart rate variability, skin conductance, and blood oxygen concentration have great potential for application in the assessment of abnormal neurological or physiological states. Heart rate variability is a physiological indicator that measures changes in heart rhythm and can reflect pain and anxiety; skin conductance can be used as an effective indicator of sympathetic nerve activity, reflecting the body's physiological response to discomfort such as pain or nausea. However, current physiological signal detection technology still faces problems such as motion artifacts and ambient light interference in practical applications. Especially in patch or wristband devices, the impact of motion on PPG signals is significant, so more complete artifact elimination solutions need to be developed to ensure signal accuracy. In addition, the assessment of abnormal neurological or physiological states is highly subjective, and existing equipment lacks personalized models to address individual differences, resulting in limited feedback and regulation of symptoms.

[0005] Furthermore, existing skin patches often use a single material for their adhesive layer, such as traditional chemical adhesives or conductive hydrogels. While chemical adhesives offer strong adhesion, they suffer from poor breathability and can easily cause skin discomfort and allergic reactions with prolonged use. While conductive hydrogels offer high conductivity, they lack mechanical strength and adhesion, making them prone to falling off during exercise or sweating, limiting their effectiveness and comfort in practical applications. Summary of the Invention

[0006] The technical problem to be solved by the present invention is to provide a biofeedback electrical stimulation control device based on multimodal physiological signal monitoring.

[0007] Another technical problem to be solved by the present invention is to provide a control method for the above-mentioned biofeedback electrical stimulation control device based on multimodal physiological signal monitoring.

[0008] In order to solve the above technical problems, the technical solution of the present invention is:

[0009] A biofeedback electrical stimulation control device based on multimodal physiological signal monitoring includes an electronic component layer 101, a circuit layer 102, a human-computer interaction part, and an adhesive layer 103. The adhesive layer 103 is attached to the surface of the circuit layer 102. The circuit layer includes a neural or physiological state abnormality detection circuit and an electrical stimulation part. The electronic component layer includes a microprocessor and a power supply part for powering it.

[0010] The human-computer interaction part includes a touch screen display, a Bluetooth communication module, a voice recognition module and an external memory. The touch screen display, voice recognition module and external memory are respectively connected to the microprocessor circuit. The Bluetooth communication module is paired with the microprocessor and is used to send the data results calculated by the microprocessor to the display device. The Bluetooth communication module, voice recognition module and external memory are arranged on the electronic component layer;

[0011] The neurological or physiological state abnormality detection circuit includes a PPG sensor, an acceleration sensor, and a skin conductance sensor, wherein each of the sensors and the microprocessor are connected and communicate with each other via an electronic circuit;

[0012] The electrical stimulation part includes a bidirectional pulse circuit, a load detection circuit and a voltage stabilizing filter circuit. The bidirectional pulse circuit shares an electrode with the skin conductivity sensor of the nerve or physiological state abnormality detection circuit and is used to apply symmetrical bidirectional pulse electrical stimulation. The load detection circuit is connected to the bidirectional pulse circuit and is used to detect whether the electrode is close to the skin; each of the circuits is respectively connected to a microcontroller, wherein the microcontroller controls the electrical stimulation intensity of the bidirectional pulse circuit by regulating the PWM wave.

[0013] Preferably, in the above-mentioned biofeedback electrical stimulation control device based on multimodal physiological signal monitoring, the PPG sensor is connected to the microprocessor through the signal preprocessing circuit via the data acquisition board, and the skin conductance sensor is connected to the microprocessor through the filtering circuit via the data acquisition board.

[0014] Preferably, in the above-mentioned biofeedback electrical stimulation control device based on multimodal physiological signal monitoring, the power supply part includes a lithium battery and a power conversion circuit thereof, and the lithium battery is connected to the microcontroller through the power conversion circuit to power it.

[0015] Preferably, in the above-mentioned biofeedback electrical stimulation control device based on multimodal physiological signal monitoring, the adhesion layer is composed of a conductive hydrogel electrode portion and a microgroove array biomimetic adhesion structure portion, wherein the conductive hydrogel electrode portion is a double network structure hydrogel composed of PEDOT:PSS (poly 3,4-ethylenedioxythiophene:polystyrene sulfonate = 1:2.5), poly (N-isopropylacrylamide) (NIPAM) and acrylamide (AAm), the PEDOT:PSS is 20 mg / mL, and the poly (N-isopropylacrylamide) is 150 mg / m L, acrylamide 12 mg / mL; the microgroove array bionic adhesion structure portion is made of polydimethylsiloxane (PDMS) material and is prepared into a microgroove array bionic adhesion microstructure through 3D printing technology; the bonding interface between the conductive hydrogel electrode portion and the microgroove array bionic adhesion structure portion adopts a gradual transition design so that the conductive hydrogel gradually transitions to PDMS, forming a smooth interface, avoiding stress concentration and interface peeling problems caused by sudden changes in material properties; the conductive hydrogel precursor is mixed with the PDMS prepolymer and synchronously polymerized in the interface region to form an interpenetrating network.

[0016] The conductive hydrogel precursor is an uncrosslinked liquid mixture comprising a conductive component and an initiator. The conductive component is 1-3 wt% PEDOT:PSS solution and 0.5-2 wt% carbon nanotubes (CNTs); and the initiator is 0.1-0.5 wt% photoinitiator (Irgacure 2959).

[0017] Preferably, in the above-mentioned biofeedback electrical stimulation control device based on multimodal physiological signal monitoring, the gradual transition is designed by gradually adjusting the mixing ratio of the two materials according to a linear gradient in the interface area (0.1-1mm thickness) through gradient ratio control, 100% hydrogel precursor → 50% hydrogel + 50% PDMS → 100% PDMS, so that the conductive hydrogel gradually transitions to PDMS, forming a smooth interface, avoiding stress concentration and interface peeling problems caused by sudden changes in material properties.

[0018] Preferably, in the above-mentioned biofeedback electrical stimulation control device based on multimodal physiological signal monitoring, the synchronous polymerization process includes:

[0019] (1) Initiator matching:

[0020] Add photoinitiator to the hydrogel precursor and irradiate with UV light at 365nm, 10-30mW / cm 2 , 5-10 minutes;

[0021] The PDMS prepolymer was cured (2-5 hours) by platinum catalyst (Sylgard 184) under heating (40-60°C);

[0022] (2) Synergistic curing: UV-induced rapid gelation of the hydrogel (5 min) is prioritized, followed by slow curing of PDMS at room temperature (24 h) to form an interpenetrating network.

[0023] Preferably, in the above-mentioned biofeedback electrical stimulation control device based on multimodal physiological signal monitoring, a small amount (for example, a volume fraction of 0.05%) of silver nanoparticles (AgNPs) or graphene nanosheets is added to the material of the electrohydrogel electrode part during the polymerization process to further improve the conductivity and stability of the conductive hydrogel.

[0024] Preferably, in the above-mentioned biofeedback electrical stimulation control device based on multimodal physiological signal monitoring, the width of the microgrooves in the microgroove array is 100μm to 200μm, the depth is 50μm to 100μm, and the groove spacing is 200μm to 300μm; each microgroove is arranged in a serpentine shape to increase the contact area and adhesion between the biofeedback electrical stimulation control device and the skin.

[0025] Preferably, in the aforementioned biofeedback-based electrical stimulation control device based on multimodal physiological signal monitoring, the PPG sensor utilizes a dual-wavelength LED light source (red and green) to collect heart rate variability signals. This dual-wavelength design not only improves the accuracy of heart rate variability signal acquisition but also effectively suppresses ambient light interference, enabling stable signal acquisition performance in both static and active environments. The heart rate variability signals obtained by the PPG sensor provide the core data foundation for subsequent assessment of neurological or physiological abnormalities.

[0026] Preferably, in the aforementioned biofeedback-type electrical stimulation control device based on multimodal physiological signal monitoring, the skin conductance sensor utilizes a differential amplifier circuit to eliminate common-mode interference, thereby ensuring the accuracy of skin conductance data. Skin conductance signals can be used to reflect sympathetic nerve activity and indirectly measure the severity of abnormal neurological or physiological states. The skin conductance sensor shares electrodes with the bidirectional pulse circuit of the electrical stimulation component, reducing circuit space and improving data transmission efficiency.

[0027] Preferably, in the above-mentioned biofeedback electrical stimulation control device based on multimodal physiological signal monitoring, the acceleration sensor is used to detect the wearer's motion state in real time, and the data of the acceleration sensor is used to eliminate and compensate for motion artifacts. Combined with the adaptive filtering algorithm, it can reduce the interference of exercise on heart rate variability (PPG) and skin conductance signals.

[0028] Preferably, in the above-mentioned biofeedback electrical stimulation control device based on multimodal physiological signal monitoring, the bidirectional pulse circuit in the electrical stimulation part adopts a two-stage conversion output circuit to generate a symmetrical bidirectional pulse electrical stimulation signal, thereby achieving efficient and stable electrical stimulation output. The bidirectional pulse design avoids the skin discomfort that may be caused by traditional monopolar electrical stimulation, ensuring comfort and safety during the control process. The electrical stimulation intensity and frequency can be dynamically adjusted according to the evaluation results to accurately match the user's actual control needs.

[0029] Preferably, in the above-mentioned biofeedback electrical stimulation control device based on multimodal physiological signal monitoring, the microprocessor uses a low-power processor, which is responsible for real-time processing and analysis of the data from each sensor and controlling the electrical stimulation output. The microprocessor also integrates an edge computing module to support local operation of machine learning algorithms, providing computing support for personalized neural or physiological state abnormality assessment and adaptive electrical stimulation parameter adjustment.

[0030] The aforementioned edge computing module is a commercially available, specialized computing hardware deployed locally on the device (or close to the data source). It processes and analyzes data directly where it is generated, rather than relying on the cloud or a remote server. The module's core function is to locally run machine learning algorithms, enabling real-time, low-latency, personalized neural / physiological state analysis and electrical stimulation parameter adjustment.

[0031] Preferably, in the above-mentioned biofeedback electrical stimulation control device based on multimodal physiological signal monitoring, the power supply device is a rechargeable lithium battery provided with a charging port, which is convenient for users' daily use and supports long-term patch-type wearing.

[0032] The system control method of the biofeedback electrical stimulation control device based on multimodal physiological signal monitoring comprises the following steps:

[0033] Step 1: Signal acquisition: use a PPG sensor with anti-motion interference to obtain pulse wave signals, a skin conductance sensor to obtain skin electrical signals, and an acceleration sensor to collect motion data;

[0034] Step 2: Signal processing: The microprocessor uses wavelet transform and adaptive filtering to reduce noise on the acquired physiological signals and extract characteristic parameters of heart rate variability, including SDNN (standard deviation of normal RR intervals, reflecting overall HRV), RMSSD (root mean square of the difference between adjacent RR intervals), pNN50 (the proportion of adjacent RR intervals with a difference exceeding 50 milliseconds), and SDANN (standard deviation of the average RR interval every 5 minutes in the full recording).

[0035] Step 3: Symptom assessment: The microprocessor uses the support vector regression (SVR) algorithm to grade the degree of abnormal neurological or physiological status, and classifies symptoms into none, mild, moderate, and severe;

[0036] Step 4: Electrical stimulation regulation: The microprocessor uses a deep Q-network (DQN) algorithm to dynamically adjust electrical stimulation parameters, including frequency and intensity, according to symptom classification to achieve personalized electrical stimulation regulation;

[0037] Step 5: Feedback optimization: monitor physiological signals in real time during electrical stimulation. If the symptoms are relieved, reduce the intensity of electrical stimulation. If the symptoms persist, increase the intensity of electrical stimulation and enter the next stimulation cycle.

[0038] Preferably, in the above-mentioned system control method of the biofeedback electrical stimulation control device based on multimodal physiological signal monitoring, the wavelet transform is used to remove high-frequency noise in the PPG signal and extract the low-frequency features of the heart rate variability signal; the normalized least mean square error (NLMS) adaptive filtering algorithm is used to eliminate artifacts in combination with the motion signal of the acceleration sensor.

[0039] Preferably, in the above-mentioned system control method of the biofeedback electrical stimulation control device based on multimodal physiological signal monitoring, the support vector regression (SVR) algorithm is used to predict the degree of abnormality of the wearer's neural or physiological state based on heart rate variability and skin conductance signals, and store the scoring results in a user-exclusive database to improve the personalization of the model.

[0040] Preferably, in the above-mentioned system control method of the biofeedback electrical stimulation control device based on multimodal physiological signal monitoring, the deep Q network (DQN) algorithm is used to optimize the frequency and intensity of electrical stimulation in real time, gradually update the stimulation parameters through feedback data, and improve adaptability.

[0041] A multimodal physiological signal acquisition system method for resisting motion artifacts using the above-mentioned biofeedback electrical stimulation control device based on multimodal physiological signal monitoring includes the following steps:

[0042] Step 1: The two LED light sources of the PPG sensor emit light simultaneously, and the photoelectric receiving tube receives the light signals of two wavelengths;

[0043] Step 2: Filter and amplify the two received optical signals respectively;

[0044] Step 3: Perform differential mode processing on the filtered dual-wavelength optical signal to eliminate ambient light interference and obtain a high-quality pulse wave signal;

[0045] Step 4: Use motion artifact elimination methods to eliminate the influence of the wearer's movement on the control device during use, ensuring that accurate physiological signals can be obtained under different motion states.

[0046] A method for adaptively adjusting electrical stimulation parameters using the above-mentioned biofeedback electrical stimulation control device based on multimodal physiological signal monitoring comprises the following steps:

[0047] Step 1: Initialize the electrical stimulation intensity based on individual user differences, including age, gender, and other information;

[0048] Step 2: Use the Deep Q-Network (DQN) algorithm to update the electrical stimulation parameters, including frequency and intensity, in real time to ensure stimulation adaptability;

[0049] Step 3: During the electrical stimulation process, based on the feedback of heart rate variability and skin conductance signal changes, the electrical stimulation parameters are optimized to ensure personalized adjustment at all times.

[0050] A biofeedback system for user interaction includes the above-mentioned biofeedback electrical stimulation control device based on multimodal physiological signal monitoring and a corresponding mobile phone application, including the following functions: visual display of abnormal neural or physiological state assessment results, electrical stimulation parameters and other results; and providing voice control function for users to turn on / off the device, adjust the electrical stimulation intensity and activate different stimulation modes.

[0051] The specific algorithm used in the system control method of the biofeedback electrical stimulation control device based on multimodal physiological signal monitoring is as follows:

[0052] (1) Heart rate variability signal processing algorithm and multimodal cross-fusion

[0053] The PPG signal is denoised using a multi-layered signal processing algorithm, including wavelet transform, adaptive filtering, and multimodal cross-fusion, to obtain accurate heart rate variability data in a sports environment.

[0054] (1-1) Wavelet transform noise reduction: The PPG signal is first subjected to multi-scale decomposition through wavelet transform, which decomposes the original signal into sub-signals in multiple frequency bands. By screening out low-frequency components and filtering out high-frequency noise, features related to heart rate variability are extracted. The characteristic parameters of heart rate variability extracted by wavelet transform include SDNN (standard deviation of normal RR intervals, reflecting overall HRV), RMSSD (root mean square of the difference between adjacent RR intervals), pNN50 (the proportion of adjacent RR intervals with a difference of more than 50 milliseconds), and SDANN (standard deviation of the average RR interval every 5 minutes in the full recording). These parameters can comprehensively reflect heart rate variability and provide reliable indicators of abnormal neurological or physiological symptoms.

[0055] (1-2) Adaptive Filtering for Artifact Reduction: To reduce artifacts caused by the wearer's motion, a Normalized Least Mean Square (NLMS) adaptive filter based on the NLMS algorithm is employed. The NLMS adaptive filter uses the motion signal collected by the accelerometer as a reference input and dynamically adjusts the filter tap coefficients to minimize motion artifacts. The NLMS algorithm introduces a normalization factor that effectively controls the step size, ensuring rapid convergence and stable filtering.

[0056] (1-3) Multimodal cross-fusion: Based on the user's current physiological state, a weighted fusion algorithm is used to cross-analyze heart rate variability, skin conductance, and motion data. The device prioritizes heart rate variability when stationary and accelerometers when in motion, achieving efficient signal fusion and dynamic control.

[0057] (2) Motion artifact elimination

[0058] Three motion artifact elimination methods eliminate the effects of wearer movement during use, ensuring accurate physiological signals are acquired regardless of motion. The combined application of these three methods enables the device to maintain accurate physiological signal acquisition under varying motion conditions, providing stable data support for subsequent assessments of neurological or physiological abnormalities.

[0059] (2-1) Accelerometer Artifact Compensation

[0060] An accelerometer is used to collect motion data in real time to compensate for motion artifacts in the PPG signal. The accelerometer output serves as the input to an adaptive filtering algorithm. Combined with the NLMS adaptive filter, the filter parameters can be adjusted in real time, effectively reducing motion interference on the PPG signal.

[0061] (2-2)Optimal structure of optical sensor

[0062] The patch design incorporates a red-light selective, highly reflective film on a dual-wavelength LED light source. Coated on a transparent, flexible substrate, the film utilizes the principle of optical interference to filter out external interference signals and adjusts the optical path structure to isolate motion interference signals. This further eliminates motion artifacts by separating artifact signals from physiological signals.

[0063] (2-3) Advanced signal processing algorithms

[0064] Combined with the Least Mean Square (LMS) signal processing algorithm, the improved LMS algorithm utilizes the quasi-periodic characteristics of the PPG signal to rearrange the pulse wave signal, filter out non-periodic components, and further optimize the signal purity.

[0065] (3) Personalized neurological or physiological state abnormality assessment model

[0066] A personalized neural or physiological abnormality assessment model based on edge computing is used. Using the Support Vector Regression (SVR) algorithm, combined with the multi-dimensional features of the user's heart rate variability, skin conductance, and motion data, a personalized neural or physiological abnormality assessment model is constructed. The system also dynamically adjusts model weights based on real-time contextual data (e.g., motion, resting state), ensuring more accurate and reliable assessment results.

[0067] (3-1) Support Vector Regression (SVR) Algorithm

[0068] The multidimensional features of heart rate variability and skin conductance signals are input into the SVR model. By learning the nonlinear relationship between physiological signals and neurological or physiological abnormality scores, the abnormality scores can be predicted. The scores are based on a subjective scoring system commonly used in hospitals, categorizing neurological or physiological abnormalities into four levels: none, mild, moderate, and severe.

[0069] (3-2) Personalized User Database

[0070] When using the device for the first time, the user enters basic information (such as age, gender, and health status). The system uses this data to initialize the model and set the initial parameters for electrical stimulation. During use, the device continuously updates the user database by collecting real-time heart rate variability data and subjective feedback, adjusting the weighting of various parameters to achieve more accurate and personalized assessments.

[0071] (3-3) Feedback loop mechanism

[0072] When the device detects that the user is in an abnormal neurological or physiological state, the mobile application will prompt the user to score the abnormal state. The subjective scoring data is used to update the SVR model, and the personalized model is optimized in real time through the edge computing module to gradually improve the evaluation accuracy.

[0073] (4) Self-adjustment algorithm for electrical stimulation parameters

[0074] The Deep Q Network (DQN) algorithm is used to achieve adaptive adjustment of electrical stimulation parameters. The reinforcement learning mechanism optimizes the frequency and intensity of electrical stimulation in real time to ensure personalization and scientificity. Through continuous optimization of the DQN algorithm, the electrical stimulation parameters can be adaptively adjusted during use, avoiding the limitations of traditional fixed-parameter electrical stimulation and ensuring the scientificity and effectiveness of the regulation. The self-adjustment process is as follows:

[0075] (4-1) Initialization parameter settings

[0076] When the patch is used for the first time, the system sets the initial electrical stimulation parameters based on the user's age, gender, health status and other information to meet individual differentiated needs.

[0077] (4-2) Real-time feedback regulation and reinforcement learning

[0078] During the stimulation process, the microprocessor dynamically adjusts stimulation parameters using a DQN algorithm, adjusting frequency and intensity in real time based on changes in heart rate variability and skin conductance signals. The DQN model (Deep Q-Network) continuously optimizes the strategy based on feedback from each stimulation cycle to ensure the effectiveness of the stimulation in alleviating symptoms.

[0079] (4-3) Continuous Optimization

[0080] If the user reports that their neurological or physiological abnormalities have eased, the DQN model appropriately reduces stimulation intensity to avoid overstimulation. If symptoms remain severe, the model proceeds to the next stimulation cycle, increasing the stimulation parameters until the optimal control effect is achieved. The feedback data from each stimulation cycle serves as reference input for the next round of stimulation, ensuring optimized and personalized stimulation parameters.

[0081] (5) Intelligent emergency reminder and automatic control mechanism

[0082] A companion mobile app provides visual and user-friendly device operation. The app displays assessment results, electrical stimulation parameters, and other status information. It also provides voice control, allowing users to turn the patch on and off and adjust the electrical stimulation intensity through voice commands, further enhancing the device's usability and user experience. If the system detects heart rate variability and skin conductance exceeding preset safety thresholds, the device sends an alert via the mobile app and, if necessary, initiates emergency electrical stimulation mode for automated control.

[0083] Beneficial effects:

[0084] The above-mentioned biofeedback electrical stimulation control device based on multimodal physiological signal monitoring utilizes physiological signal acquisition and electrical stimulation regulation technology, integrates multi-source biosensors and intelligent processing modules to realize dynamic monitoring and real-time analysis of physiological signals, and adaptively adjusts personalized parameters based on signal characteristics to form a non-invasive biofeedback control system; through integrated multimodal detection and intelligent control functions, it provides non-invasive and convenient control means for wearers with abnormal neurological or physiological conditions; it adopts a lightweight and comfortable adhesive structure and can be worn for a long time at the Neiguan acupoint, realizing continuous monitoring of physiological signals and closed-loop feedback regulation of electrical stimulation to meet the user's personalized health management needs.

[0085] The adhesion layer of the device adopts a composite adhesion layer structure, including a conductive hydrogel electrode part and a microgroove array bionic adhesion structure part, which has excellent electrical and mechanical properties. Among them, the conductive hydrogel electrode part has excellent electrical conductivity and biocompatibility, and can realize efficient electrophysiological signal transmission between the patch and the skin; the microgroove array bionic adhesion structure part imitates the adhesion mechanism of the octopus suction cup, providing strong adhesion and good air permeability, ensuring that the patch is stably attached under different skin conditions.

[0086] The device integrates multimodal sensors such as heart rate variability, skin conductance, and motion acceleration to achieve real-time monitoring, precise assessment, and personalized regulation of abnormal neural or physiological states. It collects the user's multimodal physiological signals, including heart rate variability, skin conductance, motion status, etc., and uses multimodal cross-fusion signal processing and machine learning algorithms to establish a personalized neural or physiological state abnormality assessment model, accurately analyze the wearer's symptom status, and through intelligent algorithms and adaptive adjustment of electrical stimulation parameters, perform personalized and dynamic symptom regulation on patients, providing an intelligent and non-invasive health management method.

[0087] Based on symptom monitoring, the device adopts a deep Q-network (DQN) algorithm and a dual feedback mechanism to automatically adjust the frequency and intensity of electrical stimulation according to real-time symptom assessment results, making the electrical stimulation regulation dynamically adaptable and able to adjust autonomously according to changes in the wearer's physiological state.

[0088] Specifically, the device is capable of multimodal physiological signal acquisition, signal fusion analysis, personalized neural or physiological state abnormality assessment, and dynamic electrical stimulation regulation:

[0089] 1. Accurately collect and intelligently integrate multimodal physiological signals: By integrating PPG, skin conductance, and accelerometer sensors, it can acquire multimodal physiological signals such as heart rate variability, skin conductance, and movement status. This multidimensional data collection significantly improves the accuracy of neurological or physiological abnormality assessments, avoids the limitations of single-signal feedback, and enables comprehensive symptom analysis. The cross-integration of heart rate variability, skin conductance, and accelerometer signals significantly enhances the accuracy of neurological or physiological abnormality assessments.

[0090] 2. The composite adhesion layer structure has both excellent electrical and mechanical properties, which can achieve efficient electrophysiological signal transmission between the patch and the skin, while providing strong adhesion and good breathability, ensuring stable adhesion and comfortable use of the patch under various skin conditions.

[0091] 3. Advanced motion artifact elimination technology: A multi-level artifact elimination solution is adopted, including acceleration sensor artifact compensation, optical structure optimization and advanced signal processing algorithms (such as the NLMS algorithm), which can effectively reduce the interference of artifacts on the signal under motion conditions, ensuring the stability and accuracy of physiological signal acquisition.

[0092] 4. Build a personalized model for assessing abnormal neurological or physiological states: Using a support vector regression (SVR) algorithm and a user-specific database, a personalized model for assessing abnormal neurological or physiological states is implemented. The system builds the model based on individual information such as the user's age, gender, and health status. Through a feedback loop, the system continuously optimizes the assessment accuracy, making it more tailored to the physiological characteristics of each user, providing precise symptom grading and a scientific basis for electrical stimulation regulation.

[0093] 5. Dynamic Adaptive Adjustment of Electrical Stimulation Parameters: A Deep Q-Network (DQN) algorithm enables real-time adaptive adjustment of electrical stimulation parameters. During stimulation, the system dynamically adjusts the frequency and intensity of electrical stimulation based on symptom status, forming a closed-loop feedback control system. This ensures that electrical stimulation is tightly coupled with changes in the wearer's symptoms, preventing overstimulation and improving the scientific and effective regulation.

[0094] 6. Portability and User-Friendly Interactive Design: The patch's lightweight and ergonomic design makes it suitable for long-term application to the Neiguan acupoint. The accompanying mobile app provides a visual display of physiological signals and monitoring of electrical stimulation status. Users can control the intensity and switching of electrical stimulation modes via voice commands, enhancing ease of use and user experience.

[0095] 7. Personalized edge computing and emergency automatic control: Utilizing a user-specific database and edge computing, we gradually optimize the evaluation model to ensure personalized processing. The system automatically adapts to different scenarios and supports emergency control to ensure user safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0096] Figure 1 Schematic diagram of the structure of the biofeedback electrical stimulation control device based on multimodal physiological signal monitoring according to the present invention;

[0097] Figure 2 This is a schematic diagram of the electronic components and circuit layer of the biofeedback electrical stimulation control device based on multimodal physiological signal monitoring according to the present invention;

[0098] Figure 3 This is a microscopic schematic diagram of the adhesion layer of the biofeedback electrical stimulation control device based on multimodal physiological signal monitoring according to the present invention;

[0099] Figure 4 This is a schematic block diagram of the hardware composition of the biofeedback electrical stimulation control device based on multimodal physiological signal monitoring according to the present invention;

[0100] Figure 5 This is a schematic diagram of the main control flow of the biofeedback electrical stimulation control device based on multimodal physiological signal monitoring according to the present invention;

[0101] Figure 6 This is the principle diagram of the PPG sensor artifact removal algorithm;

[0102] Figure 7 This is the principle diagram of the DQN algorithm.

[0103] In the figure: 1-patch; 2-skin tissue;

[0104] 101 - electronic component layer; 102 - circuit layer; 103 - adhesion layer. DETAILED DESCRIPTION

[0105] The following describes the biofeedback electrical stimulation control device and control method based on multimodal physiological signal monitoring according to the present invention in conjunction with the embodiments and drawings.

[0106] Example 1

[0107] like Figure 1-4 As shown, the biofeedback electrical stimulation control device based on multimodal physiological signal monitoring includes an electronic component layer 101, a circuit layer 102, a human-computer interaction part, an adhesive layer 103 and a shell. The adhesive layer 103 is attached to the surface of the circuit layer 102. From top to bottom, it is a three-layer structure of the electronic component layer 101, the circuit layer 102 and the adhesive layer 103. The electronic component layer and the circuit layer are placed inside the shell, wherein,

[0108] The adhesive layer 103 is embedded with an electrode sheet that can be directly attached to the Neiguan acupoint to ensure accurate conduction of electrical stimulation and facilitate the detection of physiological signals. The adhesive layer is mainly divided into two parts. One is the conductive hydrogel electrode portion: a double-network hydrogel composed of PEDOT:PSS (poly (3,4-ethylenedioxythiophene):polystyrene sulfonate = 1:2.5), poly (N-isopropylacrylamide) (NIPAM), and acrylamide (AAm). The PEDOT:PSS is 20 mg / mL, the poly (N-isopropylacrylamide) is 150 mg / mL, and the acrylamide is 12 mg / mL. This ratio can form a double-network hydrogel with excellent electrical conductivity and mechanical properties. A small amount of silver nanoparticles (AgNPs) or graphene nanosheets can be added during the polymerization process to further improve the conductivity and stability of the conductive hydrogel. The other is the microgroove array biomimetic adhesion structure: This part uses polydimethylsiloxane (PDMS) material and is fabricated using 3D printing technology to create a microgroove array biomimetic adhesion microstructure. The PDMS microgroove structure is produced using 3D printing technology. The microgrooves have a width of 100μm to 200μm, a depth of 50μm to 100μm, and a groove spacing of 200μm to 300μm. The microgrooves are arranged in a serpentine pattern to increase the contact area and adhesion between the patch 1 (i.e., the biofeedback electrical stimulation and control device described in the present invention) and the skin tissue 2. The interface between the conductive hydrogel electrode and the microgroove array biomimetic adhesion structure utilizes a gradual transition design. The conductive hydrogel gradually transitions to the PDMS, forming a smooth interface and avoiding stress concentration and interfacial delamination caused by sudden changes in material properties. The conductive hydrogel precursor is mixed with the PDMS prepolymer and simultaneously polymerized at the interface to form an interpenetrating network.

[0109] The circuit layer includes a nerve or physiological state abnormality detection circuit and an electrical stimulation part, and the electronic component layer includes a microprocessor and a power supply part for supplying power to it. Figure 4As shown, the shell of patch 1 (i.e., the biofeedback electrical stimulation control device described in the present invention) integrates multiple sensors and processing units. Among them, the PPG sensor (model SON1303) adopts a dual-wavelength red and green LED light source design to collect heart rate variability signals (HRV). The dual-wavelength design can effectively reduce ambient light interference and improve the stability of signal acquisition; the skin conductance sensor (commercially available, including a silver metal sheet and a signal amplification circuit connected to it) eliminates common-mode interference through a differential amplification circuit design to ensure the accuracy of skin conductance (GSR) data; the acceleration sensor (model MPU6050) is used to detect the wearer's motion state and input motion interference information into the artifact elimination algorithm to provide a reference basis for artifact compensation. The bidirectional pulse circuit in the electrical stimulation part adopts a two-stage conversion output circuit to generate symmetrical bidirectional pulses, to achieve stable electrical stimulation to human body, the bidirectional pulse circuit in the electrical stimulation part shares an electrode with the skin conductivity sensor of the nerve or physiological state abnormality detection circuit, for applying symmetrical bidirectional pulse electrical stimulation (the load detection circuit is connected to the bidirectional pulse circuit for detecting whether the electrode is close to the skin), which not only reduces the device volume, but also improves the efficiency of signal detection and electrical stimulation output. The microprocessor selects a low-power processor, which not only extends the device life time, but also supports real-time data processing and algorithm calculation. The composition and connection relationship of the five parts of the microprocessor (microprocessor model is STM32F series, internally containing edge detection module), nerve or physiological state abnormality detection circuit, electrical stimulation part, power supply part and human-computer interaction part are as follows:

[0110] 1. The microprocessor (the microprocessor model is STM32F series, which contains an edge detection module) is paired with the Bluetooth communication module in the human-computer interaction part to send the data results calculated by the microprocessor to the display device; the microprocessor is respectively connected to the touch display screen, voice recognition module and external memory circuit in the human-computer interaction part to communicate with each other.

[0111] 2. The neurological or physiological state abnormality detection circuit includes a PPG sensor (model SON1303), an acceleration sensor (model MPU6050), and a skin conductance sensor (commercially available, including a silver metal sheet and a signal amplification circuit connected thereto) to achieve multimodal physiological signal acquisition. Each of the sensors is connected to the microprocessor through electronic circuits and communicates with it. Among them, the PPG sensor is connected to the microprocessor through a signal preprocessing circuit via a data acquisition board. The silver metal sheet of the skin conductance sensor is connected to a filter circuit via a signal amplification circuit, and then connected to the microprocessor through a data acquisition board via the filter circuit, and then transmitted to the microprocessor through the data acquisition board. The signal is transmitted to the microprocessor through the data acquisition board.

[0112] 3. The electrical stimulation circuit includes a bidirectional pulse circuit, a load detection circuit, and a voltage stabilization filter circuit. The bidirectional pulse circuit shares electrodes with the skin conductance sensor of the neural or physiological abnormality detection circuit, and is used to apply symmetrical bidirectional pulse electrical stimulation. The load detection circuit is connected to the bidirectional pulse circuit to detect whether the electrodes are in close contact with the skin. Each of these circuits is connected to a microcontroller, which controls the electrical stimulation intensity of the bidirectional pulse circuit by regulating the PWM wave. The bidirectional pulse circuit is connected to the electrode sheet in the adhesive layer 103.

[0113] 4. The power supply part includes a lithium battery and its power conversion circuit. The lithium battery is connected to the microcontroller through the power conversion circuit to power it.

[0114] 5. The human-computer interaction part includes a touch display screen, a Bluetooth communication module, a voice recognition module and an external memory. The touch display screen, voice recognition module and external memory are respectively connected to the microprocessor circuit. The Bluetooth communication module is paired with the microprocessor and is used to send the data results calculated by the microprocessor to the display device; the Bluetooth communication module, voice recognition module and external memory are arranged on the electronic component layer, and the touch display screen is embedded in the shell.

[0115] This biofeedback-based electrical stimulation control device, based on multimodal physiological signal monitoring, achieves personalized assessment and real-time control of abnormal neural or physiological states through multimodal physiological signal acquisition, signal processing, intelligent algorithm analysis, and electrical stimulation feedback regulation. While ensuring signal acquisition accuracy, it also offers portability and long battery life, supporting all-day physiological signal monitoring and electrical stimulation feedback. The rational layout of its components results in a highly integrated design, allowing the device to adhere easily to the skin and ensuring excellent biocompatibility and wearer comfort.

[0116] Example 2

[0117] like Figure 5 As shown, the main control process of the biofeedback electrical stimulation control device based on multimodal physiological signal monitoring described in Example 1 mainly includes four key steps: S1 is the signal acquisition stage, S2 is the data processing and symptom evaluation stage, S3 is the electrical stimulation signal output stage, and S4 is the real-time feedback and adjustment stage.

[0118] The specific steps are as follows:

[0119] During the S1 signal acquisition phase, the PPG sensor and skin conductance sensor collect the wearer's heart rate and skin conductance data, respectively. The PPG sensor generates a pulse wave signal using a red and green dual-wavelength LED light source, further extracting relevant data such as heart rate variability (HRV). The skin conductance sensor measures changes in skin conductance, reflecting the degree of sympathetic nerve activation. The accelerometer monitors the wearer's motion status in real time and transmits motion signals to the microprocessor, providing artifact reference for subsequent data processing.

[0120] During the S2 data processing and symptom assessment phase, the microprocessor uses a preset algorithm to analyze and process the collected physiological signals, extracting heart rate variability and skin conductance features. This feature extraction utilizes wavelet transform and adaptive filtering to obtain multiple characteristic parameters, including SDNN (standard deviation of normal RR intervals, reflecting overall HRV), RMSSD (root mean square of differences between adjacent RR intervals), pNN50 (the proportion of adjacent RR intervals with differences exceeding 50 milliseconds), and SDANN (standard deviation of the average RR intervals per 5-minute period during the entire recording), accurately reflecting heart rate variability. Skin conductance data is also used to assess sympathetic nerve activity. Combining heart rate variability and skin conductance signals, a built-in algorithm analyzes the severity of the wearer's neurological or physiological abnormality and categorizes it as mild, moderate, or severe, providing a basis for selecting the intensity of subsequent electrical stimulation.

[0121] During the S3 electrical stimulation signal output phase, the microprocessor automatically adjusts the frequency and intensity of electrical stimulation based on the symptom level assessed in S2. A pulsed current is applied to the wearer's Neiguan acupoint via the electrodes, stimulating the sympathetic nervous system to alleviate neurological or physiological abnormalities. The electrical stimulation module utilizes a symmetrical bidirectional pulse design to avoid skin discomfort associated with monopolar stimulation, ensuring comfort and safety for the wearer during the stimulation process.

[0122] During the S4 real-time feedback and adjustment phase, the system continuously monitors the wearer's physiological parameters, such as changes in heart rate variability and skin conductance, assesses stimulation effectiveness, and adjusts stimulation parameters in real time. If the system detects a reduction in the wearer's neurological or physiological state, the microprocessor reduces the intensity of stimulation. If symptoms remain severe, the microprocessor initiates a new stimulation cycle, adjusting stimulation parameters until symptoms are alleviated. This closed-loop feedback control process enables intelligent symptom monitoring and adaptive regulation, providing users with a personalized control experience.

[0123] The above-mentioned biofeedback electrical stimulation control device based on multimodal physiological signal monitoring adopts the following algorithm:

[0124] 1. PPG signal processing algorithm

[0125] In the S1 signal acquisition step of the patch control process, PPG signal processing uses a multi-level denoising method based on wavelet transform and normalized least mean square (NLMS) algorithm to achieve accurate extraction of heart rate variability signals, such as Figure 6 shown.

[0126] First, the collected PPG signal undergoes multi-scale decomposition using a wavelet transform. This transform decomposes the original signal into components of different frequency bands, filtering out low-frequency components associated with heart rate variability and removing high-frequency noise. Heart rate variability characteristic parameters extracted using wavelet analysis include SDNN, RMSSD, pNN50, and SDANN, which comprehensively reflect the complexity and variability of heart rate.

[0127] Wavelet is a function φ(t) that satisfies certain conditions, and is a family of functions φ generated by scaling and translation. a，b (t)

[0128]

[0129] Where a is the scale factor and b is the translation factor.

[0130] So the wavelet transform of function f(t) with respect to wavelet function φ(t) is:

[0131]

[0132] The signal denoising process is divided into three steps: wavelet decomposition of the signal, threshold quantization of high-frequency coefficients of wavelet decomposition, and wavelet reconstruction.

[0133] Wavelet transform is essentially a similarity operation. When selecting wavelet basis functions, try to choose wavelets that are similar to the pulse wave to achieve the best denoising effect.

[0134] Secondly, to eliminate artifacts caused by the wearer's motion, the NLMS adaptive filtering algorithm is used to further process the signal. The NLMS algorithm uses the motion signal collected by the accelerometer as a reference input and dynamically adjusts the filter's tap coefficients to maximize motion artifact removal. By introducing a normalization factor to optimize the filter step size, the NLMS algorithm achieves rapid convergence while ensuring filter stability.

[0135] The NLMS algorithm is an adaptive filtering algorithm widely used in fields such as noise cancellation, echo cancellation, and system identification. The NLMS algorithm is based on the least mean square error (LMS) algorithm and improves the step size control by introducing a normalization factor, thereby improving the convergence speed and stability of the algorithm. The output signal is expressed as:

[0136]

[0137] Where x(n) is the input signal, that is Figure 4 The preprocessed PPG signal is shown in Figure 1; W(n) is the filter tap coefficient, which is dynamically adjusted according to the input signal and the error signal to minimize the mean square value of the error signal. The coefficient is updated as follows:

[0138]

[0139] The e(n) in the coefficient update formula is the error signal:

[0140] e(n)=d(n)-u(n) (5)

[0141] The d(n) in the error signal calculation formula is the reference signal, that is, Figure 6 The acceleration sensor signal shown in the figure; μ in the coefficient update formula represents the step size factor, and γ is a preset correction value to prevent the denominator from being zero.

[0142] After wavelet denoising and NLMS adaptive filtering processing, the artifact interference of the PPG signal is greatly reduced, and the obtained heart rate variability signal is clearer and more reliable, providing high-quality physiological data support for the assessment of abnormal neurological or physiological status.

[0143] 2. Personalized neurological or physiological state abnormality assessment model

[0144] In the assessment of abnormal neurological or physiological status, a personalized model based on the support vector regression (SVR) algorithm is used, such as Figure 5 As shown in Figure 2, the SVR algorithm can learn the nonlinear relationship between heart rate variability and skin conductance signals in a multidimensional feature space, thereby accurately predicting the wearer's abnormal status score.

[0145] When using the device for the first time, the user is required to enter basic information (such as age, gender, and health status). This information is used to generate an initial model and set the initial electrical stimulation parameters accordingly. During operation, the device automatically collects the user's heart rate variability data and sensory feedback and records the data in a user-specific database. The system uses this data to dynamically adjust the weights of various physiological parameters, enabling the evaluation model to reflect the user's individual characteristics and achieve more accurate and personalized assessment of abnormal neural or physiological states.

[0146] At the same time, when the patch detects that the user has an abnormal neurological or physiological state, the mobile application will prompt the user to subjectively score the abnormality level. This feedback information will be fed back to the SVR model to fine-tune the algorithm parameters and further improve the model's personalized accuracy.

[0147] Through the above mechanism, the present invention can optimize the neural or physiological state abnormality assessment model in real time according to individual differences, making the state abnormality detection more accurate and providing a scientific basis for electrical stimulation regulation.

[0148] 3. Electrical stimulation parameter self-adjustment algorithm

[0149] like Figure 7 As shown, the present invention adopts a deep Q network (DQN) algorithm in adjusting electrical stimulation parameters.

[0150] Data collection and feature extraction: feedback signals and wearer response data are collected during each electrical stimulation cycle. These data can be used as part of the state (s). In terms of action selection, based on the current state, the DQN algorithm selects an action (a), namely the frequency and intensity of the electrical stimulation.

[0151] To update the Q-value function, the DQN algorithm uses a neural network to approximate the Q-value function. The Q-value update formula is as follows:

[0152] Q(s,a)←Q(s,a)+α[r+γQ(s′,a′)-Q(s,α)] (6)

[0153] Where α is the learning rate, r is the reward (i.e., the positive feedback from the wearer’s response), and γ is the discount factor that measures the current value of future rewards. is the maximum Q-value of all possible actions in the next state.

[0154] Experience replay stores the state, action, reward, and next state in each electrical stimulation cycle in the experience replay buffer, and then randomly samples a batch of data for training to reduce the correlation between samples.

[0155] Target network, in order to stabilize the training process, the DQN algorithm uses two networks: the main network and the target network. The parameters of the target network are a copy of the parameters of the main network, but the update frequency is lower. The target network is used to calculate the TD target:

[0156]

[0157] Where λ is the interpolation factor, is the Q value of the target network.

[0158] Loss function and optimization, use the mean square error loss function to calculate the difference between the predicted Q value and the target Q value, and perform gradient descent optimization:

[0159]

[0160] Here, θ is the network parameter and N is the batch size.

[0161] The DQN algorithm uses reinforcement learning strategies to optimize the frequency and intensity of electrical stimulation in real time, ensuring personalized and scientific regulation. Before stimulation begins, the system sets the initial stimulation intensity based on the user's gender, age, and other information. Using the DQN model, the microprocessor continuously adjusts the stimulation parameters to suit the wearer's specific needs. The DQN algorithm updates the model's policy weights based on feedback signals and the wearer's response during each stimulation cycle, gradually finding the optimal stimulation frequency and intensity for that individual.

[0162] During each electrical stimulation session, a microprocessor records the wearer's heart rate variability and skin conductance data and calculates the degree of symptom relief. If the wearer reports that their neurological or physiological abnormalities have eased, the DQN model reduces the intensity of electrical stimulation to avoid overstimulation. If there is no significant improvement in symptoms, the DQN model automatically enters a new electrical stimulation cycle, increasing the intensity until the optimal parameters are found.

[0163] This adaptive algorithm achieves continuous optimization of electrical stimulation parameters through reinforcement learning, ensuring the scientificity and effectiveness of each electrical stimulation regulation, and avoiding the limitations of traditional fixed-parameter electrical stimulation.

[0164] In summary, the biofeedback electrical stimulation control device based on multimodal physiological signal monitoring described in the present invention realizes real-time symptom monitoring and control through high-precision hardware design, multi-level signal processing algorithm, edge computing personalized neural or physiological state abnormality assessment model and electrical stimulation adaptive adjustment algorithm. In terms of monitoring and control of neural or physiological state abnormalities, a highly integrated design of multimodal signal acquisition, personalized assessment and dynamic feedback regulation is realized, integrating functions such as multimodal physiological signal acquisition, intelligent cross-fusion analysis, and dynamic electrical stimulation regulation. It can accurately assess and control neural or physiological state abnormalities, has the advantages of being intelligent, non-invasive and convenient, and provides users with a safer and more effective means of health management.

[0165] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as within the scope of protection of the present invention.

Claims

1. A biofeedback electrical stimulation control device based on multimodal physiological signal monitoring, characterized by: It includes an electronic component layer, a circuit layer, a human-computer interaction part and an adhesive layer, wherein the adhesive layer is attached to the surface of the circuit layer, the circuit layer includes a nerve or physiological state abnormality detection circuit and an electrical stimulation part, and the electronic component layer includes a microprocessor and a power supply part for it, wherein: The human-computer interaction part includes a touch screen, a Bluetooth communication module, a voice recognition module and an external memory. The touch screen, voice recognition module and external memory are respectively connected to the microprocessor circuit, and the Bluetooth communication module is paired with the microprocessor; the Bluetooth communication module, voice recognition module and external memory are arranged on the electronic component layer; The neurological or physiological state abnormality detection circuit includes a PPG sensor, an acceleration sensor, and a skin conductance sensor, wherein each of the sensors and the microprocessor are connected and communicate with each other via an electronic circuit; The electrical stimulation part includes a bidirectional pulse circuit, a load detection circuit and a voltage stabilization filter circuit. The bidirectional pulse circuit shares an electrode with the skin conductivity sensor of the nerve or physiological state abnormality detection circuit, and the load detection circuit is connected to the bidirectional pulse circuit; each of the circuits is connected to a microcontroller respectively, wherein the microcontroller controls the electrical stimulation intensity of the bidirectional pulse circuit by regulating the PWM wave.

2. The biofeedback electrical stimulation control device based on multimodal physiological signal monitoring according to claim 1, characterized in that: The PPG sensor is connected to the microprocessor through a signal preprocessing circuit via a data acquisition board, and the skin conductance sensor is connected to the microprocessor through a filtering circuit via a data acquisition board.

3. The biofeedback electrical stimulation control device based on multimodal physiological signal monitoring according to claim 1, characterized in that: The adhesion layer consists of a conductive hydrogel electrode portion and a microgroove array biomimetic adhesion structure portion, wherein the conductive hydrogel electrode portion is a double-network structure hydrogel composed of PEDOT:PSS, poly(N-isopropylacrylamide) and acrylamide, wherein the PEDOT:PSS is 20 mg / mL, the poly(N-isopropylacrylamide) is 150 mg / mL, and the acrylamide is 12 mg / mL; the microgroove array biomimetic adhesion structure portion is made of polydimethylsiloxane material and is prepared into a microgroove array biomimetic adhesion microstructure through 3D printing technology; the interface between the conductive hydrogel electrode portion and the microgroove array biomimetic adhesion structure portion adopts a gradual transition design so that the conductive hydrogel gradually transitions to PDMS to form a smooth interface; the conductive hydrogel precursor is mixed with the PDMS prepolymer and synchronously polymerized in the interface region to form an interpenetrating network.

4. The biofeedback electrical stimulation control device based on multimodal physiological signal monitoring according to claim 3, characterized in that: Silver nanoparticles or graphene nanosheets are added to the material of the hydrogel electrode during the polymerization process; and the microgrooves are arranged in a serpentine shape.

5. The biofeedback electrical stimulation control device based on multimodal physiological signal monitoring according to claim 1, characterized in that: The PPG sensor uses a dual-wavelength LED light source of red and green light; the skin conductance sensor uses a differential amplifier circuit to eliminate common-mode interference. The skin conductance sensor shares electrodes with the bidirectional pulse circuit of the electrical stimulation part; the acceleration sensor is combined with an adaptive filtering algorithm to reduce the interference of exercise on heart rate variability and skin conductance signals, which is used to eliminate and compensate for motion artifacts; the bidirectional pulse circuit in the electrical stimulation part uses a two-stage conversion output circuit.

6. The system control method of the biofeedback electrical stimulation control device based on multimodal physiological signal monitoring according to any one of claims 1 to 5, characterized in that: The following steps are involved: Step 1: Signal acquisition: use a PPG sensor with anti-motion interference to obtain pulse wave signals, a skin conductance sensor to obtain skin electrical signals, and an acceleration sensor to collect motion data; Step 2: Signal processing: The microprocessor uses wavelet transform and adaptive filtering to perform noise reduction on the obtained physiological signal and extracts the characteristic parameters of heart rate variability, including SDNN, RMSSD, pNN50, and SDANN; Step 3: Symptom assessment: The microprocessor uses a support vector regression algorithm to grade the degree of abnormal neurological or physiological status, classifying symptoms as none, mild, moderate, or severe. Step 4: Electrical stimulation regulation: The microprocessor uses a deep Q-network algorithm to dynamically adjust electrical stimulation parameters, including frequency and intensity, according to symptom classification to achieve personalized electrical stimulation regulation; Step 5: Feedback optimization: monitor physiological signals in real time during electrical stimulation. If the symptoms are relieved, reduce the intensity of electrical stimulation. If the symptoms persist, increase the intensity of electrical stimulation and enter the next stimulation cycle.

7. A multimodal physiological signal acquisition system method for resistant motion artifacts using the biofeedback electrical stimulation control device based on multimodal physiological signal monitoring according to any one of claims 1 to 5, characterized in that: The following steps are involved: Step 1: The two LED light sources of the PPG sensor emit light simultaneously, and the photoelectric receiving tube receives the light signals of two wavelengths; Step 2: Filter and amplify the two received optical signals respectively; Step 3: Perform differential mode processing on the filtered dual-wavelength optical signal to eliminate ambient light interference and obtain a high-quality pulse wave signal; Step 4: Use motion artifact elimination methods to eliminate the influence of the wearer's movement on the control device during use, ensuring that accurate physiological signals can be obtained under different motion states.

8. A method for adaptively adjusting electrical stimulation parameters using the biofeedback electrical stimulation control device based on multimodal physiological signal monitoring according to any one of claims 1 to 5, characterized in that: The following steps are involved: Step 1: Initialize the electrical stimulation intensity based on individual user differences; Step 2: Use a deep Q-network algorithm to update electrical stimulation parameters in real time to ensure stimulation adaptability; Step 3: During the electrical stimulation process, based on the feedback of heart rate variability and skin conductance signal changes, the electrical stimulation parameters are optimized to ensure personalized adjustment at all times.

9. A biofeedback system for user interaction, comprising the biofeedback electrical stimulation control device based on multimodal physiological signal monitoring according to any one of claims 1 to 5 and a corresponding mobile phone application, characterized in that: Includes the following functions: visual display of abnormal neurological or physiological status assessment results, electrical stimulation parameters and other results; Provides voice control function for users to turn on / off the device, adjust the intensity of electrical stimulation and activate different stimulation modes.

10. The system control method of the biofeedback electrical stimulation control device based on multimodal physiological signal monitoring according to any one of claims 1 to 5, characterized in that: The specific algorithm used is as follows: (1) Heart rate variability signal processing algorithm and multimodal cross-fusion The PPG signal is denoised using a multi-layer signal processing algorithm, including wavelet transform, adaptive filtering, and multimodal cross-fusion, to obtain accurate heart rate variability data in a sports environment: (1-1) Wavelet transform noise reduction: The PPG signal is first subjected to multi-scale decomposition using wavelet transform, breaking the original signal into sub-signals in multiple frequency bands. By screening out low-frequency components and filtering out high-frequency noise, features related to heart rate variability are extracted. The heart rate variability characteristic parameters extracted by wavelet transform include SDNN, RMSSD, pNN50, and SDANN, which comprehensively reflect heart rate variability and provide reliable indicators of abnormal neurological or physiological symptoms. (1-2) Adaptive filtering to reduce artifacts: NLMS adaptive filter based on normalized least mean square error algorithm is used; The NLMS adaptive filter uses the motion signal collected by the acceleration sensor as a reference input and dynamically adjusts the filter tap coefficients to remove motion artifacts to the greatest extent possible; the normalized minimum mean square error algorithm introduces a normalization factor; (1-3) Multimodal cross-fusion: Based on the user's current physiological state, a weighted fusion algorithm is used to cross-analyze heart rate variability, skin conductance, and motion data; the device prioritizes heart rate variability when in a stationary state and accelerometer weighting when in motion; (2) Motion artifact elimination Three motion artifact elimination methods are used to eliminate the effects of wearer movement during use, ensuring accurate physiological signals can be obtained under different motion states. (2-1) Accelerometer Artifact Compensation An accelerometer is used to collect motion data in real time to compensate for motion artifacts in the PPG signal. The output of the accelerometer is used as the input of the adaptive filtering algorithm. Combined with the NLMS adaptive filter, the filtering parameters can be adjusted in real time. (2-2)Optimal structure of optical sensor The patch design adds a red-light selective high-reflective film to the dual-wavelength LED light source. The highly reflective film is plated on a transparent flexible substrate. It uses the principle of optical interference to filter external interference signals and adjusts the optical path structure to extract motion interference signals separately. By separating the artifact signal from the physiological signal, motion artifacts are further eliminated. (2-3) Advanced signal processing algorithms Combined with the minimum mean square error signal processing algorithm; using the quasi-periodic characteristics of the PPG signal, the pulse wave signal is rearranged, the non-periodic components are filtered out, and the signal purity is further optimized; (3) Personalized neurological or physiological state abnormality assessment model A personalized neural or physiological state abnormality assessment model based on edge computing is used. The support vector regression algorithm is combined with the multi-dimensional features of the user's heart rate variability, skin conductance, and motion data to build a personalized neural or physiological state abnormality assessment model. The system also dynamically adjusts the model weights based on real-time contextual data to ensure more accurate and reliable assessment results. (3-1) Support Vector Regression Algorithm The multidimensional features of heart rate variability and skin conductance signals are input into the SVR model. By learning the nonlinear relationship between physiological signals and abnormal neurological or physiological status scores, the abnormal status scores can be predicted. The scoring results are based on the subjective scoring system commonly used in hospitals. (3-2) Personalized User Database When the device is used for the first time, the user needs to enter basic information, and the system uses this data to initialize the model and set the initial parameters of electrical stimulation; During use, the device collects the user's heart rate variability data and subjective feedback in real time, continuously updates the user database, and adjusts the weights of various parameters to achieve more accurate personalized assessment; (3-3) Feedback loop mechanism When the device detects that the user is in an abnormal neurological or physiological state, the mobile app will prompt the user to rate the abnormality. The subjective rating data is used to update the SVR model, and the edge computing module optimizes the personalized model in real time to gradually improve the evaluation accuracy. (4) Self-adjustment algorithm for electrical stimulation parameters A deep Q-network algorithm is used to achieve adaptive adjustment of electrical stimulation parameters. The reinforcement learning mechanism is used to optimize the frequency and intensity of electrical stimulation in real time, ensuring personalization and scientificity. Through continuous optimization of the DQN algorithm, the electrical stimulation parameters are adaptively adjusted during use, avoiding the limitations of traditional fixed-parameter electrical stimulation and ensuring the scientificity and effectiveness of the regulation. The self-adjustment process is as follows: (4-1) Initialization parameter settings When the patch is first applied, the system sets the initial electrical stimulation parameters based on user information to meet individual differentiation needs; (4-2) Real-time feedback regulation and reinforcement learning During the electrical stimulation process, the microprocessor dynamically adjusts the electrical stimulation parameters using the DQN algorithm, adjusting the frequency and intensity in real time based on changes in heart rate variability and skin conductance signals. The DQN model continuously optimizes the strategy based on feedback data from each electrical stimulation cycle to ensure the effectiveness of electrical stimulation in alleviating symptoms. (4-3) Continuous Optimization If the user reports that their neurological or physiological abnormalities have eased, the DQN model appropriately reduces the stimulation intensity to avoid overstimulation. If the symptoms remain severe, the model enters the next stimulation cycle and increases the stimulation parameters until the optimal control effect is achieved. The feedback data from each stimulation cycle serves as a reference input for the next round of stimulation, ensuring the optimization and personalization of the stimulation parameters. (5) Intelligent emergency reminder and automatic control mechanism A matching mobile phone application is used to realize the visualization and humanization of device operation; the application displays the status and provides voice control function, allowing users to control the opening and closing of the patch and the adjustment of the electrical stimulation intensity through voice commands, further improving the usability of the device and user experience; when the system monitors that the heart rate variability and skin conductance exceed the preset safety threshold, the device will send a reminder through the mobile phone application and activate the emergency electrical stimulation mode if necessary to achieve automatic control.

Images