Intelligent high-voltage pulse electrical stimulation system and control method thereof

Abstract

The invention relates to an intelligent high-voltage pulse electrical stimulation system and a control method, and solves the problems of poor individual suitability, low positioning precision and weak safety in the prior art. The system obtains body surface point cloud through a split scanning module (infrared / TOF), and dynamic mapping of an anatomical structure is achieved in combination with a medical image template library (the error is smaller than 1.2 mm); dynamically adjusting a high-voltage pulse (0-1200V) based on a BMI-impedance model, and adopting an improved KNN algorithm to match a discharge scheme; three-level safety protection is innovated, and dangerous area protection is achieved through impedance sudden change detection, displacement compensation and phase counteracting (electric field attenuation is larger than 99.9%). Clinical verifications show that the positioning precision is + / -0.5 mm, the pain relief rate of obese patients (BMI = 32) is 92%, and the false touch rate is less than 0.01%. Accurate targeting and safe regulation and control are achieved, and the method is suitable for the fields of muscle rehabilitation, pain management and the like.

Description

1. Technical Field

[0001] This invention lies at the intersection of intelligent medical devices and bioelectrical control, specifically a high-voltage pulsed electrical stimulation system based on dynamic anatomical mapping and its control method. This system achieves personalized neural control of deep tissues through multimodal sensor fusion (infrared / TOF / UWB), medical image-guided STL model registration, and anatomical feature-driven discharge scheme matching. Core technologies include:

[0002] 1. Anatomical coordinate system construction: Through the elastic registration (ICP+EKF) of the DICOM image template library and the real-time body surface point cloud, the millimeter-level mapping relationship between internal organs and body surface projections is established;

[0003] 2. Dynamic parameter optimization: Based on the cardiac safe zone boundary model (1.2×d_min) and nerve plexus impedance gradient tracking (ΔZ / Δx>15Ω / mm), three-level protection in the danger zone is achieved;

[0004] 3. Intelligent solution matching: An improved KNN algorithm (hybrid distance formula) is used to retrieve the optimal discharge parameters from the historical database, supporting precise adaptation for muscle rehabilitation, pain management, and sports assistance scenarios. 2. Background Technology

[0005] Defects of existing technology

[0006] 1. Insufficient individual adaptability:

[0007] According to a 2022 clinical study in Clinical Rehabilitation (sample size n=150), 42% of obese users with a BMI>30 reported that traditional electrical stimulation devices were ineffective, while the incidence of overstimulation in lean users with a BMI<18.5 reached 31%. The main reason is that the output parameters are not dynamically adjusted based on BMI, resulting in a mismatch between stimulation intensity and skin impedance. Traditional solutions use a fixed voltage output (such as CN20191023456B), which does not take into account the nonlinear relationship between skin impedance and BMI (impedance differences can reach 5-10kΩ).

[0008] 2. Low positioning accuracy:

[0009] The positioning error of existing integrated infrared / TOF sensors (such as Intel RealSense L515) reaches ±2cm (ISO19022 test standard), resulting in a deviation of the electrode activation area exceeding 15mm, which cannot meet the needs of neural targeted stimulation.

[0010] 3. Low efficiency of full body modeling:

[0011] The CN20191023456B solution requires users to manually move the device more than three times to complete a full-body scan (taking an average of 3 minutes and 15 seconds), and the cumulative registration error due to changes in body position reaches 8.7 mm.

[0012] 4. Weak security mechanisms:

[0013] US2020103456A1 does not solve the problem of motion displacement compensation, and the risk of false stimulation in the danger zone is >8%; the NFC fitting detection solution (CN20191023456B) has a misjudgment rate of up to 12.7%.

[0014] Technological gaps and innovation needs

[0015] Existing technologies lack the following core capabilities:

[0016] 1. BMI-driven dynamic parameter adaptation: Real-time adjustment of high-voltage pulse parameters based on the user's physiological characteristics.

[0017] 2. High-precision whole-body modeling: A single scan completes the mapping of the entire body’s anatomical structure with an error of less than 1mm.

[0018] 3. Multimodal safety protection: Integrates impedance, acceleration, and phase cancellation technologies to achieve a three-level response mechanism. 3. Summary of the Invention

[0019] Overall Technical Solution This invention proposes an intelligent high-voltage pulse electrical stimulation system that achieves precise, safe, and personalized neural regulation through the following innovations:

[0020] 1. High-voltage pulse generation and BMI adaptation system: A two-stage boost circuit combined with BMI dynamic adjustment outputs an adjustable pulse output range of 0-1200V, controls the frequency through a voltage-controlled oscillator, controls the pulse width through a digital delay line (DLL), controls the intensity through a constant current source + dynamic impedance compensation, controls the waveform through a waveform generator, and controls the phase difference through a multi-channel synchronous controller to achieve spatiotemporal modulation.

[0021] 2. Split positioning and scanning module: Integrates infrared structured light and TOF sensor to achieve ±0.5mm full-body modeling accuracy.

[0022] 3. The safety architecture includes a three-level safety protection system (a real-time response mechanism based on impedance mutation, electrode displacement and acceleration detection) and core safety indicators: single pulse energy: ≤4.8mJ (<10% of the ventricular fibrillation threshold), charge imbalance: <40nC (active balancing circuit), temperature rise control: ΔT ≤1.8°C (air cooling + semiconductor heat dissipation).

[0023] 4. Interactive optimization design: physical buttons and touch screen work together, and emergency response time is less than 10ms.

[0024] 5. Anatomy-surface mapping engine

[0025] This system establishes a three-dimensional anatomical coordinate system based on prior knowledge of medical imaging, realizing the intelligent association between body surface positioning and internal structure:

[0026] a) Multimodal data fusion modeling:

[0027] - Input 1: Real-time body surface point cloud obtained by the split scanning module (accuracy ±0.5mm, XYZ coordinates + normal vector)

[0028] - Input 2: Pre-existing DICOM standard anatomical template library (STL models grouped by BMI, including key structures such as the heart and nerve plexus)

[0029] - Input 3: Body surface reference points located by UWB beacon (umbilical point, seventh cervical vertebra, etc.)

[0030] -Fusion algorithm: Use Extended Kalman Filter (EKF) to achieve spatiotemporal alignment of multi-source data and output anatomical mapping relationships with confidence

[0031] b) Dynamic registration and error compensation:

[0032] 1. Coarse registration stage: Procrustes analysis is used to calculate the rigid body transformation matrix (rotation R + translation t) between the STL model and the surface point cloud to minimize the distance between the reference points:

[0033] minΣ||R·p_i+t-q_i||2

[0034] Where p_i is the STL model reference point and q_i is the measured reference point

[0035] 2. Fine registration stage: The elastic constraint iterative closest point algorithm (point-surface matching) is used to optimize non-rigid deformation and introduce skin elasticity constraints:

[0036] Objective function: E = α·E_distance + β·E_stiffness

[0037] (α=0.7,β=0.3, E_stiffness is the Laplace smoothing term of adjacent points)

[0038] 3. Real-time compensation: When the IMU detects changes in body position, the transformation matrix is ​​updated through Lie group SE (3) to maintain the registration error <1.2mm c) Intelligent matching of discharge scheme:

[0039] -Anatomical feature encoding: Encode the geometric parameters (volume, center of mass coordinates, surface curvature) of key structures (such as the heart) into 128-dimensional feature vectors

[0040] -Scheme retrieval: Match the best discharge parameters from the scheme library based on cosine similarity, priority:

[0041] 1) Same BMI group historical plan (weight 0.6)

[0042] 2) Similar anatomical features scheme (weight 0.3)

[0043] 3) Default security scheme (weight 0.1)

[0044] d) Enhanced safety strategy: anatomy-body surface mapping system, including:

[0045] i) Standard STL template library: This library stores 3D models of key anatomical structures grouped by BMI, including the geometric center coordinates and safety buffer parameters of the heart, nerve plexus, and great blood vessels; uses DICOM image-guided STL model automatic registration technology to establish the association between the body surface coordinate system and the internal anatomical structure through the following steps: a standard STL model library containing key anatomical structures such as the heart and nerve plexus is established, and stored by BMI group (interval of 2.5 kg / m 2 ), the model accuracy reaches 0.1mm;

[0046] ii) Dynamic Registration Module: This module uses the ICP algorithm to align the user scanned point cloud with the STL template, and calculates the affine transformation matrix to establish the anatomical structure-body surface mapping relationship, satisfying the following requirements:

[0047] Registration error <1.5mm (RMSE)

[0048] Calculation time < 2s (i.MX 8M Plus processor) b)

[0049] Dynamic mapping algorithm:

[0050] 1) Obtain the user's full body point cloud data (XYZ coordinates + normal vector) through the split scanning module;

[0051] 2) Use the ICP registration algorithm to align the user point cloud with the standard STL model and calculate the affine transformation matrix:

[0052] T=argmin∑(R·p_i+t-q_i) 2

[0053] Where R is the rotation matrix, t is the translation vector, p_i is the user point, and q_i is the template point;

[0054] iii) Machine Learning Calibration: The 3D U-Net network processes the TOF depth map and multi-frequency impedance data, outputting a heat map of the probability of dangerous areas. The network architecture includes:

[0055] Input layer: 256×256×4 tensor (XYZ coordinates + 100kHz impedance value)

[0056] Encoder: 5-level downsampling (max pooling 2×2×2)

[0057] Decoder: 5-level deconvolution + skip connection

[0058] Loss function: Dice coefficient + cross entropy joint optimization

[0059] Machine learning-assisted correction: A 3D U-Net network is used to perform semantic segmentation on the TOF depth map (input size 256×256×32), outputting a probability heat map to mark the danger zone (threshold > 0.8). The network training uses a transfer learning strategy (ImageNet pre-training + 500 CT data sets fine-tuning)

[0060] iv) User interaction calibration: The touch screen supports selecting sensitive areas and automatically generating 3D bounding boxes. The boundary extension rules are:

[0061] Muscle area: 10mm outward expansion

[0062] Nerve area: 20mm outward expansion

[0063] Organ area: 30mm outward expansion

[0064] 6. Multimodal positioning and collaborative control system

[0065] This system integrates UWB / TOF / IMU multi-source positioning technology to build a dynamic spatial topology map:

[0066] a) Beacon deployment and signal processing:

[0067] Three UWB beacons are deployed on the body surface (umbilicus, cervical spine, and ankle), using the Decawave DW3000 chipset, operating in the 3.5-6.5 GHz frequency band and supporting TDoA positioning mode.

[0068] Signal processing flow:

[0069] 1. Received signal strength (RSSI) and time difference of arrival (TDOA) are combined to calculate the host coordinates;

[0070] 2. Extended Kalman filter fusion IMU data (200Hz sampling rate) to compensate for motion offset;

[0071] 3. Dynamic weight allocation: When the UWB signal-to-noise ratio is greater than 20dB, the positioning weight is 0.8; when the signal-to-noise ratio is ≤10dB, it switches to IMU dominance (weight 0.7);

[0072] b) Dynamic calibration of dangerous areas:

[0073] The finite element model was used to predict the electric field distribution: E_shield = E0 × exp(-0.35 × d^1.8). When the proximity of the heart area was detected, a reverse pulse was activated to achieve an electric field attenuation of >99.9%;

[0074] Level 3 priority protection: cardiac area > nerve plexus > large blood vessels, with phase offset performed in the overlapping areas according to the gradient;

[0075] 7. BMI-driven parameter optimization system

[0076] a) Dynamic conversion of body fat percentage to BMI:

[0077] Body fat percentage was estimated using dual-frequency impedance measurement (1kHz / 100kHz), and a corrected value was generated using the formula BMI = 0.68 × body fat percentage + 15.3;

[0078] Voltage compensation rule: V_base = 500V + 20V × (BMI - 18.5), output parameters are adjusted in three levels;

[0079] b) Anatomical feature coding and matching:

[0080] The improved KNN algorithm is used to retrieve historical solutions. The distance formula is:

[0081] D=0.5*cos_sim(F1,F2)+0.5 / (1+|ΔBMI / 5|);

[0082] Dynamic update of feature weight: w_i = w_i + 0.1 (actual efficacy - predicted efficacy) F_i;

[0083] 8. Innovation of split hardware architecture

[0084] a) Modular connection design:

[0085] The main unit and the split scanning module use an asymmetric N52 neodymium magnet interface (mechanical angle error <±2°) and are equipped with gold-plated Pogo Pin contacts (contact impedance <50mΩ).

[0086] Communication protocol optimization: Bluetooth 5.3 long-range mode, using frequency hopping spread spectrum (79 channels dynamically switched) and time division multiplexing (TDM) for anti-interference;

[0087] b) Security coordination mechanism:

[0088] Three-level response strategy:

[0089] Level 1: Abnormal distance between split modules + sudden impedance change → derating output;

[0090] Level 2: Continuous abnormality → low frequency pattern;

[0091] Level 3: Communication interruption → output cut-off + phase cancellation activation;

[0092] Technology comparison and advantages

[0093] 1. Quantitative analysis of existing technology defects

[0094] (1) Physiological basis of individual adaptability defects:

[0095] According to a study in IEEE Trans Biomed Eng, Vol. 68, 2021 (DOI: 10.1109 / TBME.2020.3012345), skin impedance (Z) has a significant nonlinear relationship with BMI. An exponential model was established by collecting data from 200 subjects (BMI 16-35):

[0096] Z=1250·exp(0.08·BMI)+320(R 2 =0.91)

[0097] The actual current density of the traditional fixed voltage scheme dropped to 32±7% of the target value when BMI>30, while it exceeded the target by 215±23% when BMI<18.5 (p<0.01), verifying the necessity of dynamic voltage compensation.

[0098] (2) Clinical impact of positioning error:

[0099] Citing a 2020 clinical trial (n=45) in the Journal of Neural Engineering, when the electrode positioning error is >10mm:

[0100] The success rate of common peroneal nerve stimulation decreased from 92% to 41%

[0101] Quadriceps activation efficiency decreased by 63%

[0102] The risk of accidentally hitting the sciatic nerve increases eightfold

[0103] The clinical value of the ±0.5mm positioning accuracy of the present invention was demonstrated.

[0104] 2. Theoretical model construction

[0105] (1) Skin impedance-BMI-voltage compensation model:

[0106] V_comp=V_base×(1+k·ΔBMI)×(Z_meas / Z_ref)^0.5

[0107] in:

[0108] V_base = 500V (reference voltage)

[0109] k = 0.15 (compensation coefficient, optimized by gradient descent)

[0110] Z_ref = 2000Ω (reference impedance, measured value when BMI = 22)

[0111] Model verification shows that the output voltage error is reduced from ±35.2% of the traditional solution to ±7.8% (RMSE = 48V).

[0112]

[0113] 4. Equipment structure innovation

[0114] 1. High voltage pulse generation system

[0115] Two-stage boost circuit design:

[0116] First stage: The MP3429 IC (92% efficiency) boosts the 3.7V lithium battery voltage to 12V. A 4.7μH power inductor (Coilcraft XAL6060, 3A saturation current) and a 10μF low-ESR ceramic capacitor (X7R material) are used, achieving a ripple voltage of less than 50mV.

[0117] Second stage: The LM2733 IC boosts 12V to 36V, and combined with a 22μH inductor (TDK VLS6045) and a 10μF ceramic capacitor, the output stability reaches ±1%.

[0118] High-frequency transformer: EPCOS B78476A1103 (ratio 1:33), with a 0.1μF / 1500V energy storage capacitor (Kemet C1812) connected in series on the secondary side, outputting 0-1200V adjustable pulses with a current ≤10mA (compliant with IEC 60601-1).

[0119] Protection circuit design:

[0120] Input protection: An SMBJ150A TVS diode (breakdown voltage 150V ± 5%) and a 0.5A resettable fuse (Polyswitch PTR2012) provide dual-stage protection with a response time of less than 1μs.

[0121] Energy storage discharge: A 0.1μF capacitor is connected in parallel with a 1MΩ discharge resistor (±1% accuracy) to ensure that the voltage drops to the safety limit of 36V within 5 seconds after power failure.

[0122] Output isolation: The insulation distance between the primary and secondary sides is greater than 8mm, and has passed the 3000V AC / 1 minute withstand voltage test (IEC61010).

[0123] 2. Split scanning module

[0124] Magnetic interface design:

[0125] Asymmetric magnetic pole layout: Using N52 neodymium magnets (magnetic flux density 0.52T), calibrated by a 3D laser interferometer (Keyence LJ-V7000), the mechanical angle error is less than ±2°, preventing reverse installation.

[0126] Pogo Pin contacts: Gold-plated copper alloy (Au layer thickness 0.1μm, contact impedance <50mΩ, four-wire test), supporting a lifespan of 100,000 plug-in and pull-out cycles.

[0127] Scanning performance verification:

[0128] Equipment: Creaform HandySCAN 3D (compliant with ISO 10360-8 standard).

[0129] Accuracy:

[0130] Single point accuracy: ±0.05mm (static test).

[0131] Whole body modeling error: ±0.5mm (dynamic test, compared with ±15mm of traditional solution).

[0132] Efficiency: Complete a full-body scan in 15 seconds (traditional solutions take 3 minutes).

[0133] Sensor configuration:

[0134] Infrared structured light: Luminar Technologies LiDAR (wavelength 940nm, power 1.2W), which projects a pseudo-random speckle pattern to analyze skin surface deformation.

[0135] TOF sensor: ST VL53L5CX (8×8 array, range 0.1-4 cm, accuracy ±1 mm).

[0136] 3. Safety protection system

[0137] Three-level response mechanism:

[0138] First-level response: When BMI is less than 18.5 or impedance mutation is greater than 30%, the output intensity is reduced by 50%.

[0139] Secondary response: When the TOF detection electrode displacement is greater than 10 mm, it switches to 1 Hz low-frequency mode.

[0140] Level 3 response: When the ultrasound identifies the heart area (attenuation coefficient > 0.8dB / cm) or the acceleration is > 3g, the output is cut off within 10ms.

[0141] Phase cancellation technology:

[0142] Circuit design: The IR2104 H-bridge driver chip controls the output phase difference (±180°±2°) of the electrode pairs, and the AD8302 phase detection module (resolution 0.1°) performs real-time calibration.

[0143] Electric field attenuation: Electric field intensity in the heart area is less than 5V / m (measured value 0.2V / m), and attenuation efficiency is greater than 99.9%.

[0144] 4. Interaction optimization design

[0145] Side physical buttons:

[0146] Encoder: ALPS EC12E rotary encoder, short press to adjust intensity (±10% step), long press for 3 seconds to lock the device. Anti-slip design: Silicone surface texture (Ra=0.8μm), blind operation recognition rate >95%.

[0147] Touch screen: 2.1-inch TFT-LCD (resolution 240×320), real-time display of impedance value, output voltage and safety status.

[0148] 5. Beacon integration and status detection

[0149] a) Beacon fitting monitoring technology:

[0150] Multimodal detection solution:

[0151] 1. Dual-frequency micro-current injection (1kHz / 100kHz), contact impedance > 200Ω is considered to be disconnected;

[0152] 2. The MEMS accelerometer analyzes the vibration spectrum and the energy attenuation is >3dB when in contact with the skin;

[0153] 3. Thermistor monitors the temperature rise rate (normal > 0.5℃ / s) and triggers an alarm when abnormal;

[0154] b) Positioning circuit design:

[0155] The UWB receiver is equipped with a quad-helix antenna array (radiation efficiency > 65%), integrated with CRC32 checksum and MAC whitelist filtering;

[0156] Clock synchronization module: Based on the IEEE 802.15.4z standard high-precision time stamp (resolution ≤ 65ps), it performs two-way time synchronization every 5 seconds;

[0157] 6. High voltage safety generation system

[0158] a) Bridge drive topology:

[0159] The IR2104 H-bridge chip is used to generate forward and reverse alternating pulses with a phase difference of 180±2° and a reverse voltage ratio of -80%.

[0160] The energy storage capacitor is connected in parallel with a 1MΩ discharge resistor to ensure that the voltage drops to a safe value of 36V within 5 seconds after power failure;

[0161] b) Phase detection and compensation:

[0162] The AD8302 phase detection module calibrates the output waveform in real time, with a phase error of <±0.5°;

[0163] 5. Control method optimization

[0164] 1. BMI-Impedance Combined Compensation Algorithm

[0165] Formula and parameters:

[0166] Vbase=500V+20V×(BMI-18.5)Vbase=500V+20V×(BMI-18.5)

[0167] BMI classification:

[0168]

[0169]

[0170] Clinically validated:

[0171] Dual frequency impedance detection Figure 5 ,

[0172] Experimental design: Double-blind trial in Huashan Hospital affiliated to Fudan University (n=30, BMI 16-32).

[0173] Results: The difference in stimulation intensity decreased from ±35.2% to ±7.8% (p<0.001).

[0174] 2. Working mode selection

[0175] The system has preset a basic discharge mode, which means selecting the part to be discharged through the interactive terminal; it also has preset an advanced mode, which means positioning through beacon positioning or split scanning module, and the host can be moved arbitrarily on the body to discharge, and different discharge schemes are used for different parts.

[0176] 3. Dynamic channel modeling and anti-interference

[0177] a) Adaptive positioning engine:

[0178] Signal priority strategy:

[0179] UWB RSSI>-60dBm→TDOA algorithm is preferred;

[0180] RSSI ≤ -80dBm → switch to TOF+IMU fusion positioning;

[0181] RF environment compensation: Dynamic update of channel attenuation parameters (path loss exponent n = 2.3 to 3.5);

[0182] 4. Anatomical Calibration and Machine Learning

[0183] a) 3D U-Net dynamic calibration:

[0184] Network input: 256×256×4 tensor (XYZ coordinates + 100kHz impedance value);

[0185] Encoder-decoder structure: 5-level downsampling (max pooling) + deconvolution jump connection;

[0186] Loss function: Dice coefficient + cross entropy joint optimization, segmentation accuracy Dice>0.92; 6. Specific Implementation Methods

[0187] Example 1: Treatment of Menstrual Pain (BMI Adaptation)

[0188] Hardware configuration:

[0189] The main unit is clipped to the waist belt and activates the E2-E3 electrode pair.

[0190] The pen cap module is placed on the opposite side of the abdomen and infrared structured light scanning is started.

[0191] Operation process:

[0192] The user inputs BMI=22, and the system is loaded with 800V / 50Hz / 40% duty cycle.

[0193] The AD5933 detection impedance increases from 2kΩ to 4kΩ, and the pulse width increases by 20%.

[0194] LIS2DH12 detects 5g acceleration and cuts off the output within 10ms.

[0195] Example 2: Full body scanning and danger zone protection

[0196] Steps:

[0197] The pen cap module scans and generates a full-body 3D model (completed in 15 seconds).

[0198] The depth of the quadriceps femoris muscle (2.8 cm) was identified by ultrasound, and electrodes E5-E6 were activated.

[0199] When the heart area is detected, the phase cancellation mode is activated (the electric field is attenuated to <5V / m).

[0200] Example 3: Sports Injury Rehabilitation

[0201] Scenario: Patient with quadriceps injury and BMI=27.

[0202] process:

[0203] TOF identifies the depth of the rectus femoris muscle (2.5 cm) and outputs 1200 V / 40 Hz.

[0204] The IMU detected the knee flexion movement (θ = 30°) and corrected the electrode offset by 1.2 mm.

[0205] When the acceleration is greater than 3g, the output is cut off within 10ms.

[0206] Example 4: Multi-user adaptation test

[0207] Experimental Design:

[0208] Subjects: 30 subjects (BMI 17.2-31.1), used continuously for 30 days.

[0209] result:

[0210] Indicator of the present invention traditional solution

[0211] Pain relief rate 92%±3% 78%±15%

[0212] Danger zone accidental touch 0 / 300 times 1.2 times / person

[0213] Example 5: Basic discharge mode

[0214] Short press the power button of the host for three seconds to turn on the host, select discharge to the outside of the thigh from the touch screen menu, tie the host to the outside of the thigh with the strap, and press the intensity increase and intensity decrease buttons for three seconds at the same time. The host will start to detect skin electricity and output low-intensity high-voltage pulse electricity according to the preset range. Press the intensity increase or intensity decrease button to adjust the desired output intensity. After discharging is completed, press the intensity increase and intensity decrease buttons for three seconds at the same time to stop discharging. Remove the host and turn off the power.

[0215] Example 6: UWB beacon positioning and dynamic discharge control

[0216] Scenario: Users need to perform dynamic electrical stimulation treatment on multiple areas of the waist during sports rehabilitation, and use UWB beacons to achieve precise positioning and adaptive parameter adjustment.

[0217] 1. Hardware Configuration

[0218] 1. Beacon deployment:

[0219] Beacon 1: attached 10 cm below the navel (pelvic reference point, coordinate P1).

[0220] Beacon 2: Attached to the lower corner of the left shoulder blade (coordinate P2).

[0221] Beacon 3: attached to the right radial styloid process (coordinate P3).

[0222] Beacon parameters: Decawave DW3000 chip, operating frequency 6.5GHz, transmit power -14dBm, TDoA positioning mode.

[0223] 2. Host configuration:

[0224] MCU: GD32F407 (integrated UWB receiver module).

[0225] Electrode pairs: E3-E4 (lumbar region), E5-E6 (shoulder region).

[0226] Safety module: LIS2DH12 accelerometer (detection range ±8g), AD5933 impedance detection chip.

[0227] 2. Operation process

[0228] 1. Beacon activation and pairing:

[0229] 2. Short press the host power button for 3 seconds to turn on the device. The touch screen will display "UWB Positioning Mode".

[0230] 3. The host automatically scans and binds to three beacons (Bluetooth 5.2 pairing takes less than 5 seconds), and the screen displays the three-dimensional coordinates of the beacon locations (P1, P2, P3).

[0231] 4. Positioning and discharge start:

[0232] The user wears the host on the waist, and the system calculates the real-time coordinates of the host through the TDoA algorithm (accuracy ±3cm).

[0233] The touch screen automatically loads the "Lumbar Dynamic Rehabilitation" mode and displays the preset parameters:

[0234] Voltage: 1200V (adaptive if BMI>25).

[0235] Frequency: 30Hz (dynamically adjusted ±5Hz according to impedance).

[0236] Double-click the side intensity increase or decrease button to start discharging (initial intensity 50%).

[0237] 5. Dynamic adjustment and safety monitoring:

[0238] Impedance feedback: The AD5933 detects skin impedance every 10 seconds (1kHz / 100kHz dual frequency). If R0>3kΩ, the pulse width is automatically increased by 20%.

[0239] Motion compensation: IMU detects the waist twist angle θ, and the elastic model corrects the electrode offset (Δx = 0.8mm, formula: Δx = kθd 2 / (E·t 3)).

[0240] Emergency cut-off: If the accelerometer detects a sudden bend (>3g), the output will be cut off within 10ms and a vibration alarm will sound.

[0241] Multi-zone switching:

[0242] 1. The user moves the host to the shoulder, UWB real-time positioning updates the coordinates, and the system automatically switches to "shoulder mode" (E5-E6 electrodes are activated and the voltage drops to 800V).

[0243] 2. The touchscreen displays the impedance value and safety status of the new zone, and the user can fine-tune the intensity (±10% steps) by rotating the encoder.

[0244] 3. Security Protection Mechanism

[0245] Beacon exception handling:

[0246] 1. If any beacon signal is lost (RSSI < -80dBm), the system switches to IMU inertial navigation mode. If it does not recover after 30 seconds, it enters safe standby.

[0247] 2. When the beacon falls off (capacitance detection fluctuation > ±5pF), a first-level response is triggered: the output intensity is reduced by 50%.

[0248] Dangerous area protection: When the host approaches the heart projection area (coordinate mapping error <5cm), the phase-controlled electrodes E1-E2 start the reverse pulse (-80% intensity), and the electric field intensity decays to <5V / m.

[0249] 4. Data Verification and Effect

[0250] Positioning accuracy test (n=50 times):

[0251] Static error: ±2.8cm (UWB beacon positioning).

[0252] Dynamic error: ±3.5cm (including IMU compensation).

[0253] Clinical effects:

[0254] Pain relief rate: 89% (VAS score from 5.6 to 1.2).

[0255] False stimulation events: 0 (out of 300 operations).

[0256] 5. Beneficial Effects

[0257] Accurate dynamic positioning: UWB+TDoA algorithm achieves ±3cm real-time positioning and supports multi-zone adaptive switching.

[0258] Safety enhancement: The three-level response mechanism reduces the false touch rate in the danger zone to <0.01%.

[0259] Energy efficiency optimization: The DW3000 chip consumes 3.5mA of power and the system has a battery life of >72 hours.

[0260] Example 7: Multimodal Fusion Positioning and Adaptive Discharge Control

[0261] Scenario: During sports rehabilitation, users need to dynamically switch between multiple treatment areas (such as shoulders, waist, and legs). The system uses UWB beacons, TOF / infrared, IMU, and split scanning modules for collaborative positioning to achieve precise adaptive discharge.

[0262] 1. Hardware Configuration and Deployment

[0263] 1. Positioning system deployment:

[0264] UWB beacon (macro positioning):

[0265] Beacon 1: 10 cm below the navel (pelvic reference point, Decawave DW3000 chip, positioning accuracy ±3 cm).

[0266] Beacon 2: spinous process of the seventh cervical vertebra (spinal reference point).

[0267] Beacon 3: right ankle lateral malleolus tip (lower limb reference point).

[0268] Split scanning module (full body modeling):

[0269] The pen-cap-shaped module integrates the ST VL53L5CX TOF sensor (accuracy ±1mm) and infrared structured light (Intel RealSense L515). It is placed next to the treatment bed and completes full-body 3D modeling within 15 seconds (error ±0.5mm).

[0270] IMU (motion compensation):

[0271] The host has a built-in MPU6050 (sampling rate 200Hz, angular resolution 0.01°) to detect limb movement posture in real time.

[0272] Host configuration:

[0273] Electrode array: 8 pairs of Ag / AgCl flexible electrodes (spacing 5-20mm adaptively adjusted).

[0274] Safety module: LIS2DH12 accelerometer (detection range ±8g), AD5933 dual-frequency impedance detection (1kHz / 100kHz).

[0275] 2. Operation process

[0276] 1. Initialization and full body modeling:

[0277] The user places the split scanning module at the bedside, starts infrared + TOF scanning, and generates a full-body point cloud model (including bone landmarks and skin contours).

[0278] The host receives model data via Bluetooth 5.2 and integrates it with the UWB beacon coordinates to build a dynamic spatial topology map.

[0279] 2. Mode selection and positioning start:

[0280] Long press the touch screen to select "Advanced Mode", and the system will prompt you to wear the host to the target area (such as shoulders).

[0281] Macro positioning: UWB beacon calculates the host's real-time position through the TDoA algorithm (update frequency 10Hz, accuracy ±3cm).

[0282] Micro-correction: The TOF sensor scans shoulder skin deformation (accuracy ±1mm), infrared structured light identifies the deltoid muscle contour, and corrects the electrode activation position.

[0283] Dynamic discharge and parameter adaptation:

[0284] The host automatically loads parameters based on BMI data (user preset BMI=24):

[0285] Voltage: 800 V (base value), frequency: 50 Hz, electrode distance: 10 mm.

[0286] Impedance feedback: The AD5933 detects skin impedance R0 = 2kΩ, and the system fine-tunes the pulse width to 200μs (PID closed-loop control). Motion compensation: The IMU detects upper limb abduction (θ = 45°), and the elastic model calculates the electrode offset Δx = 1.2mm, dynamically adjusting the output focus of the E3-E4 electrode pair.

[0287] Multi-position switching and safety protection:

[0288] The user moves the host to the waist, the UWB beacon updates its position to the lumbar L4 area, and the TOF scan identifies the depth of the erector spinae muscle (2.5 cm).

[0289] The system automatically switches to "waist mode", the parameters are adjusted to: 1200V / 30Hz, and the electrode spacing is extended to 20mm (suitable for BMI>25).

[0290] Three-level security response:

[0291] If the TOF detection electrode displacement is greater than 10mm, switch to 1Hz low-frequency mode;

[0292] The accelerometer detects a sudden bend (>3g) and cuts off the output within 10ms;

[0293] When the split module scans and approaches the heart (distance < 5 cm), it activates the E1-E2 reverse pulse (phase difference 180°), and the electric field attenuation is > 99%.

[0294] 3. Discharge end and data storage:

[0295] Press the intensity increase and decrease keys simultaneously for 3 seconds to stop the discharge, and the host saves the treatment data (impedance curve, positioning trajectory, and motion compensation amount).

[0296] The split module enters low-power mode (current < 1mA), and the UWB beacon continuously monitors the device's location until it shuts down.

[0297] 3. Technical Advantages and Verification Data

[0298] 1. Positioning performance test:

[0299] Fusion positioning accuracy: static ±0.8mm, dynamic ±1.5mm (including IMU compensation).

[0300] Switching response time: Position switching delay <0.5 seconds (traditional solution >2 seconds).

[0301] Clinical effect verification (n=30):

[0302] Treatment consistency: Difference in stimulation intensity among users of different body sizes is <±8% (±35% for traditional regimen).

[0303] Safety: There were no false touch events in the danger zone during 300 operations, and the acceleration cut-off success rate was 100%.

[0304] Energy efficiency performance:

[0305] The UWB beacon consumes 3.5mA of power, and the split module has a battery life of >72 hours (CR2032 button battery).

[0306] The host standby power consumption is less than 8μA, and the continuous use time is greater than 84 hours.

[0307] 4. Summary of beneficial effects

[0308] 1. Multimodal positioning fusion: UWB+TOF+IMU+split scanning achieves full scene coverage, and positioning error is reduced to millimeter level.

[0309] 2. Dynamic adaptive capability: Real-time closed-loop control combining BMI, impedance, and motion data improves treatment accuracy.

[0310] 3. Balance between safety and efficiency: A three-level response mechanism ensures safety in extreme scenarios, and the split scanning module increases operational efficiency by three times.

[0311] 4. Medical-consumer compatibility: Lighter-grade portable design (78×45×15mm), supports fast switching between 12 clinical scenarios.

[0312] Example 8: BMI Dynamic Adaptation and Body Fat Percentage Calculation

[0313] The user wears the split scanning module and starts dual-frequency impedance detection (1kHz / 100kHz);

[0314] The system separates the skin resistance R0 = 1.8kΩ and the capacitance C = 22nF according to the Cole-Cole model;

[0315] According to the formula, the body fat percentage is 28.5%, and the derivation of BMI is 0.68 × 28.5 + 15.3 = 34.7;

[0316] Automatic loading high BMI mode: voltage compensation -15%, electrode spacing extended to 20mm;

[0317] Example 9: Redundant Positioning with Multiple Beacon Failures

[0318] Simulate beacon 2 (cervical vertebrae point) detachment, capacitance detection fluctuation > ±5pF triggers an alarm;

[0319] The system switches to IMU+TOF fusion positioning, and the Kalman filter compensates for posture offset;

[0320] If the signal is not restored for 30 seconds, the safety standby is activated and the user is prompted to check the device;

[0321] Example 10: Phase-offset electric field protection

[0322] When the host approaches the heart projection area (distance < 5cm), the UWB positioning coordinates trigger an alarm;

[0323] The H-bridge circuit generates reverse pulses (phase difference 180°, amplitude -80%);

[0324] The measured intensity of the electric field probe decayed from 1200V / m to 0.2V / m, meeting the safety threshold;

[0325] Example 11: Achieve diverse neural regulation effects through dynamic coordinated modulation of five-dimensional parameters and intelligent feedback control, and produce rich sensory experiences without moving the electrode position.

[0326] 1. Core components of the hardware architecture

[0327] Main control unit

[0328] Using the STM32H7 series MCU, running a real-time operating system (RTOS), responsible for:

[0329] Fast calculation and synchronous control of five-dimensional parameters (frequency / pulse width / intensity / waveform / phase)

[0330] Receive biofeedback signals (PPG / EMG / temperature) and perform closed-loop regulation

[0331] Safety monitoring (charge balance, temperature rise, impedance mutation)

[0332] Dual-frequency impedance detection module

[0333] Low frequency channel (1kHz): monitors electrode-skin contact impedance to prevent falling off (accuracy ±5%)

[0334] High-frequency channel (100kHz): Assess deep tissue status (such as inflammation / edema) and distinguish resistance / capacitance components through phase angle analysis

[0335] Multi-channel output ASIC

[0336] Integrated 16-bit high-precision DAC, supporting:

[0337] Voltage dynamic range: 5V-1500V (adaptive output in constant current mode)

[0338] Waveform generation: 12 standard waveforms (square wave / sine wave / sawtooth wave, etc.) are pre-stored, and user-defined waveforms can be uploaded.

[0339] The safety monitoring unit measures tissue temperature (NTC sensor), charge accumulation (integrator circuit), and local blood flow (PPG) in real time

[0340] Hardware-level protection: Output is cut off within 2ms in case of abnormality (10 times faster than software response)

[0341] 2. Sensory modulation technology under fixed electrodes

[0342] 1. Five-dimensional parameter coordination strategy

[0343] When the electrodes are fixed, different sensations are achieved by combining the following parameters:

[0344] Constant base frequency: Maintain a base frequency of 50Hz (to avoid neural adaptation)

[0345] Pulse width dynamic adjustment:

[0346] 50μs narrow pulse width → targeting Aδ fibers (sharp pain)

[0347] 300μs wide pulse width → activates C fibers (burning sensation)

[0348] Real-time waveform switching:

[0349] Square wave: strong electric shock feeling (used for sports rehabilitation)

[0350] Sine wave: soft massage sensation (for anxiety relief)

[0351] Multi-channel phase difference control:

[0352] Dual electrodes with a 90° phase difference → creating a virtual sense of diffusion

[0353] Four electrodes with gradual phase change → simulate "moving" stimulation effect

[0354] 2. Time coding technology (breaking physical limitations)

[0355] Pulse Position Modulation (PPM):

[0356] The carrier frequency is fixed at 50 Hz, and different sensations are encoded by changing the pulse time interval;

[0357] 3. Summary of Technical Advantages

[0358] No need to move electrodes:

[0359] The combination of five-dimensional parameters can produce ≥20 different sensations, avoiding frequent adjustments to electrode positions.

[0360] Nerve selective activation:

[0361] Pulse width-frequency synergy achieves precise targeting of Aβ / Aδ / C fibers (error <15%).

[0362] Dynamic adaptability:

[0363] Dual-frequency impedance monitoring + biofeedback enables "set-and-forget" automatic adjustment.

[0364] Safety redundancy design:

[0365] Double protection of hardware-level protection (2ms response) and algorithm prediction (dopamine-dependent model).

[0366] Personalized extensions:

[0367] Supports importing user MRI / CT data to generate stimulation plans adapted to anatomical structures.

[0368] Typical application scenarios:

[0369] After training, athletes can complete the entire process of "analgesia → relaxation → muscle strength recovery" by fixing the electrodes

[0370] Chronic pain patients: A single patch provides all-day intelligent pain management VII. Description of the Figures

[0371] Figure 1:System architecture diagram (including high-voltage circuit, split module, MCU and electrode array), showing the connection relationship between high-voltage circuit, split module, MCU and electrode array.

[0372] Figure 2 : The three-level safety response logic diagram (BMI-impedance-acceleration coordination) shows the BMI-impedance-acceleration coordinated response process.

[0373] Figure 3 and Figure 5 : Schematic diagram of the dual-frequency impedance detection circuit, showing the frequency switching logic of the AD5933 driving the electrode pair.

[0374] Figure 4 : Schematic diagram of the elastic mechanics compensation model, showing the relationship formula between Δx, θ, d and the correction process.

[0375] All embodiments and drawings do not constitute limitations on this patent; all pen caps refer to the stimulator adopting a split structure, and the pen cap represents the split scanning module in the split structure.

Claims

1. An intelligent high-voltage pulse electrical stimulation system, characterized in that include: a) a multi-stage high-voltage generation unit, which boosts the input voltage to therapeutic-grade high voltage in stages through at least two stages of voltage conversion, and includes a combined architecture of a DC boost module and a high-frequency transformer; b) a multi-channel electrode system comprising at least two pairs of electrodes; c) Dynamic control system, integrating at least one of the following functional modules:

1. Parameter control module: supports user setting and / or automatic adjustment of at least one of the five parameters including pulse frequency, pulse width, intensity, waveform, and spatiotemporal modulation; 2. Biofeedback module: Dynamically adjusts output intensity by real-time detection of electrode-skin impedance; 3. Regional adaptation module: intelligently matches activated electrode pairs based on pre-stored anatomical feature data; d) Composite safety system, including:

1. Hardware protection layer: current / voltage clamping is achieved through current limiting circuits and / or transient suppression devices; 2. Software protection layer: triggers output interruption or derating based on impedance mutation detection; e) Two-way interactive interface, supporting local input (touch, key, voice) and / or remote control (wireless communication protocol).

2. The high-voltage pulse electrical stimulation system according to claim 1, characterized in that Also include at least one of the following: a) Motion sensing module, including a three-axis accelerometer. When an acceleration greater than 3g is detected and sustained for a set period of time, the safety system is triggered to cut off the output or reduce the output intensity; b) The BMI adaptation module dynamically adjusts the pulse parameters by at least one of the following methods: i) Estimation of body fat percentage using multi-frequency impedance measurement, and calculation of the correction value using the formula BMI = 0.68 × body fat percentage + 15.3; ii) Using a TOF sensor or infrared structured light to generate a 3D body point cloud and predicting BMI using machine learning; c) Positioning module, using at least one of the following technologies: i) Infrared structured light triangulation, which calculates skin deformation using coded gratings and CMOS sensors; ii) TOF-UWB fusion positioning, combined with IMU data to compensate for motion offset; iii) The TOF sensor measures the electrode-skin distance and determines the wearing position using a gradient algorithm; iv) Beacon positioning system, including at least two wireless beacons deployed on the body surface, supporting UWB, BLE, ultrasonic or infrared coding types, and achieving three-dimensional positioning through TDOA / TOF / optical flow tracking algorithms; d) Bridge drive topology, generating alternating forward and reverse pulses, supporting multi-electrode collaborative or independent working modes; e) Modular interface, supporting rapid replacement of carbon fiber, silicone or gel electrode pads or electrodes; f) Power module: battery, charging circuit, charging interface, and / or power management module and / or thermal management module and / or waterproof structure; g) The high-voltage pulse electrical stimulation system uses an independent power supply from a high-voltage generation unit: the boost circuit is isolated from the MCU power domain and / or the energy storage capacitor is connected in parallel with the discharge circuit to ensure that the voltage drops to a safe value after power failure; h) Phase cancellation circuit: The H-bridge driver chip works in conjunction with the phase detection module, with an optimal reverse pulse voltage ratio of -80%; i) A discharge plan matching system based on anatomical feature coding, comprising at least one of the following:

1. Anatomical structure dynamic mapping subsystem, including: - A split 3D scanning module integrating an infrared structured light emitter (wavelength 940±10nm) and a TOF sensor array (8×8 pixels), configured to acquire full-body point cloud data (accuracy ±0.5mm) within 15 seconds; -Medical imaging database, storing images grouped by BMI (interval 2.5 kg / m 2 ) human anatomy STL model, including safety buffer parameters of the heart and nerve plexus; - A dynamic registration processor configured to perform: i) Non-rigid registration of user point cloud and STL template is achieved by improving ICP algorithm, with objective function E = 0.7*E_distance+0.3*E_stiffness; ii) Lie group SE (3) was used to update the model to compensate for body position changes and maintain the registration error < 1.2 mm; 2. Adaptive pulse generation subsystem, including: - Two-stage boost circuit, boosting 3.7V input to 1200V adjustable output (±5V step accuracy); - Phase-controlled H-bridge circuit, supporting the generation of oppositely offset pulses with a phase difference of 180±2°; 3. Multimodal security protection subsystem, integrating: -Three-level response mechanism: When impedance mutation ΔZ / Δt>30% / s is detected, the first level derating is activated; when electrode displacement is>10mm, the second level low-frequency mode is activated; when acceleration is>3g, the output is cut off within 10ms; - an electric field cancellation module, configured to attenuate the electric field strength in the target area to <5 V / m (attenuation efficiency >99.9%) when the heart area is detected to be approaching; 4. Intelligent decision-making subsystem, configured as: - Retrieve discharge plans from the historical database based on the improved KNN algorithm. The distance formula is: D=0.5*cos_sim(F1,F2)+0.5 / (1+|ΔBMI / 5|); -Dynamically update feature weight w_i = w_i + 0.1*(actual efficacy - predicted efficacy)*F_i 5. Calibration of anatomical structures: Using at least one of the following methods, preferably automatically extracting STL models of key anatomical structures (heart / neural plexus) from DICOM images and mapping them to a body surface coordinate system and / or using machine learning for dynamic calibration using a 3D U-Net network architecture with a 256×256×4 tensor input layer (XYZ coordinates + 100kHz impedance value), an encoder with five levels of downsampling, and a decoder using deconvolution and skip connections. Alternatively, user-defined anatomical structures can be selected via a touchscreen, with the system recording the 3D bounding box.

3. The high-voltage pulse electrical stimulation system according to claim 1, characterized in that A split-type wearable electrical stimulation device comprising at least one of the following: a) Split connection structure: i) The main unit and the split scanning module are connected via a magnetic interface. The magnet layout is an asymmetric N52 neodymium magnet, and the mechanical angle error is preferably <±2°; ii) Adaptive adjustment of transmission power (0-10dBm) between the host and the split module based on Bluetooth RSSI signal strength, preferably ensuring communication stability within 3 meters; b) The main unit has a built-in auxiliary TOF sensor and the split module has a built-in battery; c) When the distance between the host and the split module causes the scanning field of view to overlap or the data integrity to be insufficient, the safety mode is triggered or the output is cut off; d) Data transmission module: After the split module scans and generates a full-body 3D point cloud model, it is transmitted to the host computer after encryption and / or forward error correction. The data compression adopts at least one of the following methods: i) spatial coordinate difference prediction coding; ii) hierarchical voxel encoding; iii) Statistical entropy coding; iv) Chunked streaming; e) Level 3 security response: i) Level 1: Reduce output intensity when the distance between split modules is abnormal and impedance suddenly changes; ii) Level 2: Switch to low-frequency mode when the abnormality persists; iii) Level 3: When communication is interrupted, the output is cut off and phase cancellation is activated; f) Split module integrating infrared structured light emitter and / or TOF sensor array; g) The host and / or split scanning module have built-in UWB chips and / or synchronous positioning beacons and / or Bluetooth modules and / or auxiliary infrared light sources to assist in locating the positions of the host and split scanning modules and / or assist in imaging; h) Bluetooth module (preferably Bluetooth 5.3 long-range mode) for transmitting 3D point cloud data; frequency hopping spread spectrum technology (FHSS) for interference resistance, including dynamic switching between multiple channels within the 2.4 GHz band to avoid Wi-Fi / microwave oven interference; and / or time division multiplexing (TDM): the split modules synchronize their clocks with the host, alternately transmitting and receiving according to a preset cycle to avoid signal collisions; i) The infrared structured light transmitter supports hardware auto-focus and / or wavefront coding technology (WFC).

4. The high-voltage pulse electrical stimulation system according to claim 1 comprises a multimodal beacon positioning system, characterized in that At least one of the following: a) Beacon deployment and data collection:

1. At least two wireless beacons are attached to the body surface, supporting at least one of UWB, BLE, ultrasonic, or infrared coding types.

2. The electrical stimulator has a built-in multi-mode receiver to acquire the signal characteristic parameters of each beacon in real time, including: time difference of arrival (TDOA) and / or received signal strength indicator (RSSI) and / or channel state information (CSI) phase offset. b) Adaptive positioning engine, dynamically selecting the optimal signal source and using a hybrid algorithm (TDOA / TOF / optical flow tracking); c) Beacon collaborative management: cross-protocol networking and dynamic topology calibration, updating channel model parameters; d) Enhanced security strategy: Enable redundant positioning when a beacon fails; if the RSSI value of any beacon suddenly drops, respond immediately, preferably by cutting off the output and issuing a system alarm; e) The dynamic calibration method for hazardous areas includes at least one of the following:

1. Impedance-temperature joint monitoring: When a specific area meets the following conditions simultaneously: When ΔZ>30% (5-second sliding window) and ΔT>1.5°C Automatically mark as temporary danger zone and generate phase offset instructions 2. Establish a hazardous area propagation model, predict the electric field distribution through finite element analysis, and generate a three-dimensional shielding matrix: E_shield=E0×exp(-k×d^n) (k = 0.35, n = 1.8 is the biological tissue attenuation coefficient, d is the Euclidean distance to the boundary of the danger zone) 3. Risk zone priority management strategy: When multiple overlapping risk zones are detected, gradient protection is implemented in the order of cardiac area > nerve plexus > large blood vessels > muscles; f) Dynamic positioning algorithm:

1. Calculate the relative distance between the electrical stimulator and each beacon based on the signal propagation model. The formula is: d_i=c·Δt_i+ε (c is the signal propagation speed, Δt_i is the arrival time difference, and ε is the environmental noise correction term); 2. Solve the coordinates (x, y, z) of the electrical stimulator using the least squares method. The objective function is: minΣ(d_i-\sqrt{(x-x_i)^2+(y-y_i)^2+(z-z_i)^2})^2 g) The beacon integrates at least one of the following fit status detection technologies:

1. Bio-impedance detection unit: preferably injects 1kHz / 100kHz dual-frequency microcurrent into the skin through the conductive contact ring at the bottom of the beacon to measure the contact impedance. If the threshold is exceeded, it is judged as shedding; 2. Capacitive coupling sensing: Detects the coupling capacitance between the beacon and the skin, triggering an alarm when the fluctuation exceeds a preset value; 3. Vibration damping analysis: The beacon vibration spectrum is collected through a MEMS accelerometer. If the energy attenuation during skin contact is less than a preset value, it is determined to be not properly adhered.

4. NFC near-field verification: When the NFC signal strength (RSSI) between the beacon and the stimulator is less than the preset value, it is determined that the beacon is too far away from the skin; 5. Temperature gradient detection: The beacon’s built-in thermistor (e.g., NTC 10kΩ) monitors the rate of temperature rise when in contact with the skin (normal > 0.5°C / s). Abnormal rates are marked as falling off. h) The electrical stimulator has one of the following co-located modules built into it:

1. UWB receiving chipset, including: -Decawave DW3000 chip, operating in the 3.5-6.5GHz frequency band, supports TDoA positioning mode; -Quadruple helix antenna array, radiation efficiency> 65%, half-power beamwidth ≥ 120°; 2. Inertial navigation unit, including: -MPU6050 six-axis IMU, configured to collect acceleration and angular velocity data at a 200Hz sampling rate; - Extended Kalman filter, fusing IMU data with UWB positioning results to dynamically compensate for motion offset; 3. The security verification module performs at least one of the following verifications: Beacon ID whitelist verification: only respond to the MAC addresses of pre-registered beacons; Signal integrity check: Detect signal tampering through CRC32 and parity bits; Topology rationality judgment: When the distance between beacons suddenly changes by more than 10%, an abnormal alarm is triggered; i) A high-voltage pulse electrical stimulation system, characterized in that the collaborative positioning method of the UWB receiving chipset and the IMU includes:

1. Dynamic weight allocation strategy: -When the UWB signal noise ratio is greater than 20dB, the positioning weights W_uwb = 0.8 and W_imu = 0.2; -When the UWB signal noise ratio is ≤10dB, W_uwb=0.3,W_imu=0.7; 2. Motion compensation algorithm: Calculate the pose transformation matrix using IMU data: ΔT=∫(a+g)dt 2 +1 / 2∫ω×v dt (a is acceleration, g is gravity compensation, ω is angular velocity, and v is velocity) Apply ΔT to the UWB original coordinates and output the corrected stimulator position: P_corrected=ΔT·P_uwb 3. Exception handling mechanism: -When the UWB positioning residual is greater than 3σ, the IMU autonomous navigation mode is activated and lasts for ≤30 seconds; -If the residual error persists, a level 3 safety response is triggered and a fault code is recorded; j) The electrical stimulator has a built-in phase synchronization module to achieve clock calibration with the beacon, including:

1. High-precision timestamp based on the IEEE 802.15.4z standard (resolution ≤ 65ps); 2. Temperature compensated crystal oscillator (TCXO), frequency stability ±1ppm; 3. Two-way timing protocol, performing clock deviation correction every 5 seconds to maintain time base error <100ns.

5. The high-voltage pulse electrical stimulation system according to claim 2, characterized in that One of the following: a) The biofeedback module uses dual-frequency impedance detection (1kHz / 100kHz) to separate skin resistance (R0) and coupling capacitance (C). Dual-frequency impedance detection is based on the Cole-Cole model to separate skin resistance and capacitance, with frequencies of 1kHz and 100kHz being preferred. b) Dynamically adjust the high-voltage output according to the formula V_base = 500V + 20V × (BMI-18.5), and adjust the output in three levels according to the BMI value, including low BMI mode (BMI < first threshold), standard mode, and high BMI mode (BMI > second threshold).

6. The high-voltage pulse electrical stimulation system according to claim 2, characterized in that One of the following: a) Phased electric field cancellation technology: When the heart area is detected, the H-bridge driver chip is activated to generate reverse pulses with a phase difference of ±180°±5°, and the electric field strength is attenuated to <5V / m; b) The electric field strength decays to <5V / m, and the decay efficiency is >99.9%; c) The main electrode outputs a positive pulse (V1), the offset electrode outputs a reverse pulse (0.8V1), and the electric field gradient decays according to distance.

7. The high-voltage pulse electrical stimulation system according to claim 4, characterized in that One of the following: a) Adaptive positioning engine signal priority strategy: When the UWB signal strength is greater than -60dBm, the TDOA algorithm is prioritized; when it is less than -80dBm, it switches to TOF+IMU fusion positioning; b) Kalman filtering integrates multi-source data, and IMU feedback angular velocity and acceleration; c) Dynamic channel modeling includes RF fading parameters, acoustic velocity temperature compensation, and optical background suppression.

8. The high-voltage pulse electrical stimulation system according to claim 1, characterized in that One of the following: a) The physical buttons on the side work in conjunction with the touch screen: short press to adjust the intensity, long press for a preset time to turn pages, and double-click for emergency stop; b) Anti-slip silicone pattern on the button surface; c) The touch screen displays real-time parameters and safety status prompts.

9. The high-voltage pulse electrical stimulation system according to claim 1, characterized in that At least one of the following: a) Flexible PCB integrated with graphene conductive layer, electrode array thickness <0.15mm, bending radius <15mm, and fitting error <1.2mm; b) Contact impedance <200Ω, supports wrist, waist and leg multi-part adaptation; c) Standby power consumption <8μA, battery life >84 hours, and meets IP67 waterproof rating; d) Contains a touch screen structure and / or a rotatable fixing clip.

10. The high-voltage pulse electrical stimulation system according to claim 1 includes a method for dynamic optimization of discharge schemes, characterized in that Include at least one of the following: a) User registration and preference management: i) Differentiate individual users through biometrics (fingerprint / face) or user ID to build a personalized database; ii) Storing user historical preference parameters, including commonly used stimulation intensity, treatment site, and time period preferences; b) Record actual efficacy data, including pain score, impedance change rate and user feedback; c) Reverse update feature weights, the optimization formula is: w_i=w_i+α(actual efficacy-predicted efficacy)F_i (α is the learning rate, F_i is the eigenvector component); d) Personalized solution generation: Based on the user's registered preference parameters and real-time anatomical features, the top three candidate solutions are generated. The priority rules are as follows: i) Preference matching weight: 40% ii) Anatomical feature similarity weight: 40% iii) Safety weight: 20%; e) When the feature weight change exceeds a threshold, the solution library is triggered to retrain.

Images