Ear vagus nerve stimulation regulation system and method

Abstract

This invention relates to a system and method for regulating auricular vagus nerve stimulation, specifically within the field of auricular vagus nerve stimulation technology. The specific technical solution includes: a stimulation unit for applying electrical stimulation to the auricular vagus nerve; a sensing unit for acquiring the user's photoplethysmography (PPG) signal; a control unit connected to both the stimulation unit and the sensing unit, used to identify the expiratory phase based on the PPG signal and control the stimulation unit to output electrical stimulation only during the expiratory phase; and to calculate a heart rate variability index based on the PPG signal, and dynamically adjust the intensity of the electrical stimulation based on the deviation of the heart rate variability index from a target vagal nerve tension threshold and the rate of change of the deviation. This disclosure can improve the therapeutic effect of auricular vagus nerve stimulation.

Description

Technical Field

[0001] This invention relates to the field of auricular vagus nerve stimulation technology, and in particular to an auricular vagus nerve stimulation regulation system and method. Background Technology

[0002] Transcutaneous vagus nerve stimulation (taVNS) is a non-invasive neuromodulation technique that modulates the function of the human autonomic nervous system by applying electrical stimulation to the vagus nerve distribution areas such as the cymba conchae and cavum conchae. It has broad application prospects in areas such as anxiety relief, sleep improvement, blood pressure regulation, and adjunctive treatment of epilepsy.

[0003] Traditional vagus nerve stimulation devices mostly employ an open-loop stimulation mode, using a fixed stimulation strategy to electrically stimulate the vagus nerve. This approach cannot dynamically adapt to the user's real-time physiological state. This fixed stimulation method cannot accommodate real-time changes in nerve tension and physiological rhythms, easily leading to insufficient stimulation, overstimulation, or long-term nerve adaptation attenuation. The resulting control effect is unstable and lacks individual adaptability, ultimately leading to poor treatment outcomes. Summary of the Invention

[0004] The purpose of this application is to provide an auricular vagus nerve stimulation modulation system and method that can improve the therapeutic effect of auricular vagus nerve stimulation.

[0005] To achieve the above objectives, this application provides the following solution: In a first aspect, this application provides an auricular vagus nerve stimulation modulation system, the system comprising: Stimulation unit, used to apply electrical stimulation to the vagus nerve of the ear; The sensing unit is used to collect the user's photoplethysmography (PPG) signal; A control unit, connected to both the stimulation unit and the sensing unit, is configured to identify the expiratory phase based on the photoplethysmography (PPG) signal and control the stimulation unit to output electrical stimulation only during the expiratory phase; and to calculate a heart rate variability index based on the PPG signal and dynamically adjust the intensity of the electrical stimulation based on the deviation of the heart rate variability index from the target vagal tone threshold and the rate of change of the deviation.

[0006] This disclosure achieves simultaneous respiratory phase recognition and heart rate variability feedback regulation through a control unit, forming a dual closed-loop control: the first closed loop ensures that electrical stimulation is delivered only during the expiratory phase when parasympathetic activity is highest, conforming to the respiratory sinus rhythm (RSA), improving therapeutic efficacy and reducing adverse reactions; the second closed loop dynamically adjusts the stimulation intensity based on heart rate variability indicators (such as RMSSD), making the stimulation parameters adapt to individual physiological states and avoiding overstimulation or understimulation. Compared to existing open-loop or single closed-loop systems, this disclosure significantly improves treatment precision and safety.

[0007] In one embodiment, in identifying the expiratory phase based on the photoplethysmography signal, the control unit is specifically configured to: The respiratory waveform is separated from the photoplethysmography signal; Locate the extreme points from the respiratory waveform; Calculate the slope of the envelope of the respiratory waveform; The moment when the slope of the envelope changes from a positive value to a negative value, and the extreme point of the respiratory waveform at the location of the slope change is the maximum value, is determined as the expiration start point of the expiration phase. The expiratory phase is determined based on the preset expiratory phase interval and the expiratory initiation point.

[0008] This disclosure presents a method for extracting the expiratory initiation point from photoplethysmography (PPG) signals (including respiratory component separation, extreme point localization, and envelope slope determination). This method can accurately pinpoint the expiratory initiation time, avoiding false triggering due to noise or waveform distortion, thereby ensuring the reliability of respiratory gating and providing an accurate time reference for the first closed loop.

[0009] In one embodiment, the sensing unit includes: a PPG sensor, the PPG sensor including a red light source and an infrared light source; and in the aspect of separating the respiratory waveform from the photoplethysmography signal, the control unit is configured to: Normalize the photoplethysmography (PPG) signals from the red and infrared channels. The respiratory source signal was separated from the normalized photoplethysmography signal. The respiratory source signal is denoised using a maximum correlation entropy Kalman filter to output a respiratory waveform. The measurement noise covariance of the Kalman filter is adjusted based on the signal quality assessment results of the respiratory source signal.

[0010] This disclosure employs a PPG sensor and a respiratory signal extraction method (normalization, blind source separation, and adaptive Kalman filtering). Multispectral blind source separation effectively suppresses motion artifacts and physiological interference by utilizing the difference between red and infrared channels. Adaptive Kalman filtering dynamically adjusts parameters according to signal quality, significantly improving the robustness of respiratory waveform extraction and ensuring that high-quality respiratory signals can still be obtained under conditions of stimulation interference or low perfusion.

[0011] In one embodiment, the system further includes: a six-axis inertial measurement unit and electrodermal activity (EDA) measurement. Before performing denoising processing on the respiratory source signal using maximum correlation entropy Kalman filtering, the control unit is further configured to: Based on the signals output by the six-axis inertial measurement unit and the skin conductance sensor, the motion artifact time periods in the photoplethysmography signal are identified and the corresponding signal segments are removed to obtain the respiratory source signal after artifact removal. In the aspect of denoising the respiratory source signal using maximum correlation entropy Kalman filtering, the control unit is specifically configured to: The artifact-removed respiratory source signal was denoised using a maximum correlation entropy Kalman filter.

[0012] This disclosure introduces a six-axis IMU and a skin conductance sensor to actively identify and remove motion artifact periods before Kalman filtering. This feature significantly reduces signal contamination caused by head movements, chewing, and poor sensor contact, preventing artifacts from propagating to subsequent respiratory phase detection and HRV calculation, thereby improving the stability and accuracy of the system in real-world usage scenarios (such as walking and daily activities).

[0013] In one embodiment, the system further includes a fuzzy PID controller, and the control unit is connected to the fuzzy PID controller; The control unit is used to calculate RMSSD as the heart rate variability index based on the photoplethysmography signal. The control unit is used to calculate the deviation between the RMSSD and the target vagal nerve tension threshold and the rate of change of the deviation; The fuzzy PID controller is used to determine the step adjustment amount based on the deviation and the rate of change of the deviation; The control unit is used to superimpose the current stimulation intensity with the step adjustment amount to determine the intensity of the target electrical stimulation, and control the stimulation unit to perform electrical stimulation at the intensity of the target electrical stimulation.

[0014] This disclosure presents an HRV feedback control architecture based on RMSSD, including deviation / rate of change calculation, fuzzy PID to determine step adjustment amount, and stimulus intensity update. This architecture quantifies the autonomic nervous state (parasympathetic tension) as RMSSD, and achieves nonlinear, adaptive adjustment through fuzzy PID, so that the stimulus intensity smoothly follows physiological needs, avoids drastic fluctuations, and improves treatment comfort and effectiveness.

[0015] In one embodiment, in determining the step adjustment amount based on the deviation and the rate of change of the deviation, the fuzzy PID controller is specifically configured to: Based on multiple preset deviation levels, the target deviation level of the deviation is determined; wherein the deviation value ranges corresponding to different deviation levels do not overlap. Based on multiple preset rate of change levels, the target rate of change level of the deviation is determined; wherein the ranges of rate of change values ​​corresponding to different rate of change levels do not overlap. The target step adjustment amount is determined according to a preset fuzzy rule table, which is used to indicate the mapping relationship between different target deviation levels, different rate of change levels and step adjustment amount; Specifically, when the target deviation level is a negative deviation level, the step adjustment amount is zero.

[0016] This disclosure classifies the deviation / rate of change levels in fuzzy PID control, defines non-overlapping intervals and fuzzy rule tables, and explicitly sets the step adjustment to zero when there is a negative deviation. This feature achieves a precise mapping from physiological state to stimulus adjustment, and the negative deviation zeroing mechanism effectively prevents bradycardia induced by RMSSD exceeding the target value (over-relaxation), significantly enhancing the electrical safety and personalized adaptive capability of the system.

[0017] In one embodiment, the control unit is further configured to: after superimposing the current stimulus intensity with the step adjustment amount to obtain the target stimulus intensity, if the target stimulus intensity is greater than a preset upper threshold, then limit the target stimulus intensity to the upper threshold; if the target stimulus intensity is less than a preset lower threshold, then limit the target stimulus intensity to the lower threshold.

[0018] This disclosure defines a safe limit for the stimulation intensity (upper / lower threshold). This feature ensures that the output current is always within a medically safe range (e.g., 0.2~3.0mA), avoiding excessively strong or weak stimulation due to algorithm malfunctions or extreme physiological conditions, meeting medical device safety standards, and protecting users from electrical stimulation harm.

[0019] In one embodiment, the stimulation unit includes a stimulation pulse generator, which includes a pulse generation terminal and a synchronization signal output terminal; The stimulation pulse generator is controlled by the control unit and is configured to: under the control of the control unit, output an electrical stimulation signal through the pulse generation terminal, and output a synchronization signal through the synchronization signal output terminal at the same time as outputting the electrical stimulation signal; The sensing unit includes an analog front-end module, which has a synchronization signal input terminal; the synchronization signal output terminal of the stimulation pulse generator is directly connected to the synchronization signal input terminal of the analog front-end module via a hardwire. The analog front-end module is configured to: stop acquiring the photoplethysmography signal and maintain the current sampling value during the period when the synchronization signal is received.

[0020] This disclosure defines a hard-wired synchronous blanking mechanism between the stimulation pulse generator and the analog front-end module. This hardware-level solution immediately stops PPG signal acquisition and maintains the sampled value during stimulation pulse emission, blocking electromagnetic interference at its source with a microsecond-level response. Compared to pure software filtering or post-processing, this significantly improves the signal-to-noise ratio of physiological signals during stimulation, ensuring that the dual closed-loop regulation does not fail due to interference.

[0021] In one embodiment, the control unit is further configured to: Real-time acquisition of gain change parameters of the simulated front-end module; Based on the gain change parameters, the original sampled values ​​are normalized to convert the signal step caused by the gain jump into a continuous physiological waveform, so as to obtain the photoplethysmography (PPG) signal.

[0022] This disclosure specifies the real-time acquisition and normalization of analog front-end gain change parameters. This parameter-aware reconstruction algorithm eliminates step artifacts caused by gain switching, restores the signal to continuous smoothness, eliminates baseline abrupt changes, and ensures the accuracy of HRV calculation.

[0023] Secondly, this application provides a method for regulating vagus nerve stimulation, the method being applied to the vagus nerve stimulation system as described in any of the first aspects, the method comprising: The stimulation unit applies electrical stimulation to the vagus nerve of the ear; The sensing unit collects the user's photoplethysmography (PPG) signal; The control unit is connected to the stimulation unit and the sensing unit respectively, identifies the expiratory phase based on the photoplethysmography (PPG) signal, and controls the stimulation unit to output electrical stimulation only during the expiratory phase; and calculates a heart rate variability index based on the PPG signal, and dynamically adjusts the intensity of the electrical stimulation based on the deviation of the heart rate variability index relative to the target vagal nerve tension threshold and the rate of change of the deviation. Attached Figure Description

[0024] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0025] Figure 1 A block diagram of an auricular vagus nerve modulation system provided in an embodiment of this application; Figure 2 This is a schematic diagram of the respiratory waveform and stimulation triggering timing provided in an embodiment of this application; Figure 3 A schematic diagram of the PPG waveform (AFE Sync Signal) after hardware synchronization blanking provided in an embodiment of this application; Figure 4 This is a waveform diagram illustrating the combined hardware blanking and MCCKF processing. Figure 5 A schematic diagram comparing noise power spectral density; Figure 6 This is a schematic diagram illustrating the coherence factor comparison. Figure 7 This application provides an embodiment of the overall hardware architecture of an auricular vagus nerve modulation system; Figure 8 This is a flowchart of a method for regulating vagus nerve stimulation provided in an embodiment of this application. Detailed Implementation

[0026] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0027] Figure 1 A block diagram of an auricular vagus nerve modulation system provided in an embodiment of this application is shown below. Figure 1 As shown, the system includes: Stimulation unit, used to apply electrical stimulation to the vagus nerve of the ear.

[0028] The sensing unit is used to collect the user's photoplethysmography (PPG) signal.

[0029] A control unit, connected to both the stimulation unit and the sensing unit, is configured to identify the expiratory phase based on the photoplethysmography (PPG) signal and control the stimulation unit to output electrical stimulation only during the expiratory phase; and to calculate a heart rate variability index based on the PPG signal and dynamically adjust the intensity of the electrical stimulation based on the deviation of the heart rate variability index from the target vagal tone threshold and the rate of change of the deviation.

[0030] In one embodiment, the sensing unit may take the following forms and be worn in the following positions: 1. Wrist-worn sensors integrate sensing units into wrist-worn devices such as smartwatches and fitness trackers, utilizing the rich superficial vascular network of the wrist to collect PPG signals. This form factor is convenient for prolonged daily wear and is suitable for scenarios such as health monitoring, exercise recovery, and stress management.

[0031] 2. Finger clip sensor: The sensing unit is designed in the shape of a finger clip, which is clipped onto the end of the user's finger (usually the index or middle finger). It is mainly used in medical monitoring and home health devices, such as pulse oximeters.

[0032] 3. Ear-worn sensors integrate the sensing unit into ear-worn structures such as earplugs, ear hooks, or ear clips. The ear's blood flow is supplied by the external carotid artery, which is close to the anatomical location of the auricular branch of the vagus nerve (ABVN), making it convenient to integrate the stimulation electrode and PPG sensor into the same wearing structure.

[0033] 4. Patch-type sensors: The sensing unit is made into a flexible patch that can be attached to the user's chest, abdomen, or limbs. Patch-type sensors are suitable for continuous monitoring in specific scenarios, such as cardiopulmonary rehabilitation training and sleep monitoring.

[0034] 5. Ring sensor: The sensing unit is designed in the shape of a ring and worn at the base of the finger, which has the advantages of stable contact and comfortable wear.

[0035] In one specific implementation, the stimulation unit is integrated into an intraocular integrated electrode / sensor module, including stimulation electrodes placed in the tragus, cymba conchae, or cavum conchae, and a pulse generator connected thereto. The stimulation electrodes employ transcutaneous electrical stimulation, and stimulation parameters (such as pulse width, frequency, and current intensity) are dynamically set by a control unit.

[0036] In this embodiment, the sensing unit includes a photoplethysmography (PPG) sensor, which is used to acquire PPG signals in real time and convert the acquired signals into digital PPG signals for output.

[0037] From the perspective of the relative positions of the light source and detector, PPG sensors are divided into two types: transmissive and reflective. Transmissive sensors place the light source and detector on opposite sides of the tissue being measured (e.g., fingertip), measuring the intensity change of light after it penetrates the tissue. The signal amplitude is high, but it can only be used for thinner tissue areas such as the fingertip or frenulum. Reflective sensors place the light source and detector on the same side. Light penetrates the epidermis to the microvascular layer and is reflected back, but the signal is weaker and needs to resist noise from static reflections in the skin, bones, and muscles. They are suitable for thicker tissues (e.g., earlobe, wrist, forehead) and are easy to integrate into wearable devices. For example, when the sensor unit is a smartwatch, the smartwatch needs to obtain information from the wearer's wrist, so a reflective PPG sensor can be used.

[0038] The control unit is typically implemented using a low-power microcontroller (MCU) or digital signal processor (DSP) to execute a dual-loop control logic: the first loop identifies the expiratory phase based on the photoplethysmography (PPG) signal and controls the stimulation unit to output electrical stimulation only during the expiratory phase; the second loop calculates the heart rate variability index based on the PPG signal and dynamically adjusts the intensity of the electrical stimulation based on the deviation of the heart rate variability index from the target vagal nerve tension threshold and the rate of change of the deviation.

[0039] In one embodiment, in identifying the expiratory phase based on the photoplethysmography signal, the control unit is specifically configured to perform the following steps A1-A5: A1. Separate the respiratory waveform from the photoplethysmography signal.

[0040] The photoplethysmography (PPG) signal is a voltage waveform resulting from photoelectric conversion of light intensity changes, containing multiple frequency components such as heart rate (0.8~3 Hz), respiration (0.1~0.5 Hz), and motion artifacts (ranging from 0.1~10 Hz). To extract the respiration component, this embodiment employs a method combining multispectral blind source separation and adaptive filtering.

[0041] In one specific implementation of this step, the sensing unit includes a PPG sensor, which includes a red light source and an infrared light source. The two wavelengths of light have different penetration depths in blood and tissues, and their responses to changes in blood volume and tissue displacement caused by respiratory movements are different. Therefore, the respiratory components can be better separated by using the signals from the two channels.

[0042] In the aspect of separating the respiratory waveform from the photoplethysmography signal, the control unit is configured to perform the following steps A11-A13: A11. Normalize the photoplethysmography (PPG) signals of the red and infrared channels.

[0043] By normalizing and unifying the dynamic range of the two PPG signals, red and infrared (IR), the baseline drift and amplitude inconsistency caused by hardware gain differences and LED current fluctuations are eliminated.

[0044] For example, Z-score standardization can be used: ; Where x(t) represents the sampled value of the original signal at time t; μ represents the mean of all sampled points within the sliding window (e.g., the first 10 seconds). N represents the sampled values ​​within the current sliding window. The mean reflects the DC bias level of the signal in that channel; subtracting the mean eliminates baseline drift. σ represents the standard deviation of all sampled points within the same sliding window. The standard deviation reflects the amplitude of signal fluctuations; dividing by the standard deviation compresses the signal to the order of unit variance, making the signals from different channels have similar dynamic ranges. ;x norm (t) represents the normalized signal value, with a mean of 0 and a standard deviation of 1.

[0045] A12. Separate the respiratory source signal from the normalized photoplethysmography pulse wave signal.

[0046] The PPG signal is acquired at a sampling rate of approximately 166 Hz to 200 Hz. The control unit constructs an observation matrix from the normalized signals of the red and infrared channels and uses multispectral blind source separation techniques (such as principal component analysis (PCA) or independent component analysis (ICA)) to extract the independent components most relevant to the respiratory rhythm from the mixed signal as the respiratory source signal. This respiratory source signal mainly contains the respiratory frequency component (0.1–0.5 Hz), and may also contain a small amount of residual pulse wave harmonics and noise.

[0047] The PPG signal was sampled at approximately 166–200 Hz. Multispectral blind source separation (PCA / ICA) was used to extract the respiratory signal, pulse wave signal, and motion artifact component from the red and infrared channels. Respiratory source signal (Resp) is mainly composed of blood volume changes caused by respiratory movements, with a frequency range of 0.1 to 0.5 Hz.

[0048] The pulse wave signal (Pulse) is a periodic component caused by heartbeats, with a frequency range of 0.8 to 3 Hz, and is used to calculate heart rate variability (HRV).

[0049] Motion artifacts, which are noise components introduced by head shaking, poor sensor contact, etc., will be suppressed later.

[0050] A13. The respiratory source signal is denoised using a maximum correlation entropy Kalman filter to output a respiratory waveform, wherein the measurement noise covariance of the Kalman filter is adjusted based on the signal quality assessment results of the respiratory source signal.

[0051] First, the quality of the respiratory signal is assessed. Let L be the number of respiratory signal sampling points within the current sliding window, denoted as r. j (j=1,2,...,L), calculate the mean within the window: ; Where r j Let μ be the j-th sampling point of the respiratory source signal. r Let L be the mean value within the window, and L be the number of sampling points within the window. The kurtosis of Gaussian noise is 3; a kurtosis much greater than 3 indicates the presence of spike noise.

[0052] The formula for calculating kurtosis is: ; The method for calculating power spectral entropy is as follows: The power spectrum P(f) is obtained by performing a Fourier transform on the respiratory source signal. m ), where f m (m=1,2,...,M) represents the m-th discrete frequency (in Hz), and M is the total number of frequency components. The power spectrum is normalized to a probability distribution: ; Then entropy H: ; High entropy indicates a lot of noise, while low entropy indicates good signal quality.

[0053] Increase the measurement noise covariance R when the quality is poor (high kurtosis or high entropy), and decrease R when the quality is good (kurtosis close to 3 and low entropy).

[0054] Maximum correlation entropy Kalman filtering (MCC-KF) is used to denoise the respiratory source signal, specifically suppressing non-Gaussian spike noise such as electrical stimulation pulses, and outputting a clean respiratory waveform. Traditional Kalman filtering assumes that noise follows a Gaussian distribution and is sensitive to pulse spikes; MCC-KF introduces the maximum correlation entropy criterion, where correlation entropy is defined as... Where X and Y are two random variables (representing the state prediction and actual observation of the filter, respectively), and E[] represents the expected value. Here, is the Gaussian kernel function, e=XY is the error variable, and r is the kernel width parameter that controls the sensitive range of the kernel function, usually taking a positive value.

[0055] The filtering process includes: state prediction, and establishing a first-order linear model using the quasi-periodic characteristics of the respiratory signal. ;in, This represents the state value at time k predicted based on the optimal estimate at time k-1. This is the optimal estimate at time k-1; Calculate prediction error during observation update ;where z k The actual observation value at time k (i.e., the sampled value of the respiratory source signal) is given, and ek is the prediction error; then the conventional Kalman gain is multiplied by the weighting factor calculated by the Gaussian kernel function. When the error e k When the value is large (i.e., a spike appears), the weight approaches 0, thus automatically reducing the impact of that sample on the filter update. Simultaneously, the filter uses the dynamically adjusted measurement noise covariance R from the previous step (R is a scalar representing the variance of the observation noise; a larger value indicates a less reliable measurement). The input is a respiratory signal, and the output is a clean respiratory waveform. , where t is time, used for subsequent expiratory phase locking; it can also output a clean PPG signal for HRV calculation.

[0056] In one embodiment, the system further includes: a six-axis inertial measurement unit and electrodermal activity (EDA) measurement. Before using maximum correlation entropy Kalman filtering to denoise the respiratory source signal, the control unit is further configured to perform the following steps: based on the signals output by the six-axis inertial measurement unit and the skin conductance sensor, identify the motion artifact time periods in the photoplethysmography signal and remove the corresponding signal segments to obtain the respiratory source signal after artifact removal. In the aspect of denoising the respiratory source signal using maximum correlation entropy Kalman filtering, the control unit is specifically configured to: denoise the artifact-removed respiratory source signal using maximum correlation entropy Kalman filtering.

[0057] Specifically, the system also includes a six-axis inertial measurement unit (IMU) and an electrical activity of the skin (EDA) sensor.

[0058] The IMU integrates a three-axis accelerometer and a three-axis gyroscope, outputting acceleration signals ax(t), ay(t), and az(t) in three orthogonal directions (unit: m / s²). 2 The respiratory signal is measured by the EDA sensor, which measures the skin conductivity level G(t) (µS), reflecting the skin's sweating and excitation state. Before denoising the respiratory signal using a maximum correlation entropy Kalman filter (MCC-KF), the control unit performs the following motion artifact identification and removal steps.

[0059] First, the control unit performs a sliding window analysis on the IMU signal. Assuming a window length of W (e.g., W = 256 sampling points, corresponding to approximately 1.5 seconds), the standard deviation of acceleration and angular velocity within each window is calculated. For acceleration, the composite amplitude is calculated: And calculate its standard deviation σA within the window; for angular velocity, take the maximum value of the standard deviation in the three directions. If σA > αA or σω > αω (αA and αω are preset thresholds, for example, α... A =0.2m / s 2 α ω If the value is 10 rad / s, then head motion is detected within the window, and the corresponding PPG signal segment is contaminated by motion artifacts.

[0060] Simultaneously, the EDA signal is analyzed. The difference dG / dt of the EDA is calculated. If dG / dt exceeds the threshold γG (e.g., γG = 0.5 μS / s) within 0.5 seconds, it indicates a rapid increase in skin conductance, which may be caused by motion friction or emotional stress. In this case, the corresponding PPG signal is also considered unreliable.

[0061] Based on the above determination, the control unit generates a binary mask sequence m(t): when any item detects an artifact, m(t) = 0 at the corresponding time (indicating that the signal segment is invalid); otherwise, m(t) = 1. Then, the time periods where m(t) = 0 in the original respiratory source signal (the signal obtained after blind source separation) are directly removed. For the removed missing segments, the system can use linear interpolation or retain the previous valid value to fill them, but usually, abnormal segments are directly discarded without filling, because subsequent HRV calculations require continuous and clean segments. Finally, the respiratory source signal after artifact removal is obtained, denoted as r. clean (t).

[0062] In the aspect of denoising the respiratory source signal using maximum correlation entropy Kalman filtering, the control unit is specifically used to: convert the aforementioned r clean (t) is used as input and fed into the MCC-KF filter. The kernel width parameter γ and measurement noise covariance R of the MCC-KF are still dynamically adjusted according to the aforementioned quality assessment results (Note: the quality assessment should be performed after artifact removal, because the signal after removal is cleaner and the assessment is more accurate). The MCC-KF outputs the final clean breathing waveform for subsequent expiratory phase locking. By first removing large motion artifacts and then using a robust filter to suppress residual spikes, the system significantly improves the accuracy of breathing phase recognition in real-world scenarios (such as walking and head turning).

[0063] The respiratory gating system precisely locates the expiratory initiation point from a clean respiratory waveform and outputs an expiratory phase window trigger signal, including the following steps.

[0064] A2. Locate the extreme point from the respiratory waveform.

[0065] First, a sliding window is used to perform local quadratic polynomial fitting on the respiratory waveform: within each local window, the fitting function is... Where t is time (in seconds), y is the amplitude of the breathing waveform, and a, b, and c are fitting coefficients (a determines the direction of the parabola's opening, b determines the position of the axis of symmetry, and c is the intercept), then the coordinates of the vertex (extreme point) of this quadratic curve are: If a < 0, the peak is a maximum value (peak), corresponding to the end of inhalation / the beginning of exhalation; if a > 0, the minimum value (trough) is a minimum value, corresponding to the end of exhalation / the beginning of inhalation.

[0066] A3. Calculate the slope of the envelope of the respiratory waveform.

[0067] Next, the envelope of the respiratory waveform is calculated. The envelope can be obtained through the Hilbert transform: Let the respiratory waveform be r(t), and its Hilbert transform be H[r(t)]. Then the instantaneous amplitude of the analytic signal is... This is the envelope; the first derivative of the envelope, dE(t) / dt, is used as the slope and is calculated using the central difference method.

[0068] A4. The moment when the slope of the envelope changes from a positive value to a negative value, and the extreme point of the respiratory waveform at the position of slope change is the maximum value, is determined as the expiration start point of the expiration phase.

[0069] A5. Determine the expiratory phase based on the preset expiratory phase interval and the expiratory start point.

[0070] Simultaneously, a typical expiratory phase interval (θstart, θend) is statistically determined based on recent respiratory cycles. For example, the expiratory phase accounts for 40%–50% of the respiratory cycle, where θstart and θend are the start and end times (in seconds) of the expiratory phase, respectively. This interval is dynamically updated based on the average value of the most recent 3–5 respiratory cycles. The moment is determined as the expiratory start point when the following three conditions are met simultaneously: the envelope slope changes from positive to negative (zero crossing); the extreme point of the respiratory waveform corresponding to this zero crossing moment is a maximum value (peak); and this moment falls within the preset expiratory phase interval (θstart, θend).

[0071] Based on the detected exhalation start point t exp_start Starting from the pre-set expiratory phase duration T, continue forward. exp It can be fixed at 1.2 seconds, or dynamically calculated as T based on the current respiratory rate. exp =0.45 / f resp , where f resp This is the real-time respiratory rate (in Hz, obtained by calculating the reciprocal of the time interval between adjacent exhalation initiation points).

[0072] This yields the complete expiratory phase window [t] exp_start ,t exp_start +T exp The control unit outputs a trigger signal within the window, allowing the stimulation unit to output electrical stimulation; outside the window, the output is shut off, thus achieving a first closed-loop control that outputs electrical stimulation only during the exhalation phase.

[0073] Figure 2 The diagram illustrates the principle of respiratory waveforms and stimulus triggering timing, as follows: Figure 2 As shown, the blue curve represents the respiratory signal, the green horizontal line represents the trigger threshold, and the orange rectangle represents the adaptive stimulus output.

[0074] Taking one complete respiratory cycle as an example: 1. Inhalation phase (stimulus off).

[0075] The respiratory signal rises from the trigger threshold, reaches its peak, and then falls back. During this phase, the signal is above the trigger threshold, which the system interprets as the inspiratory phase. The stimulation output remains at a low level (off state) to avoid treatment / stimulation during the inspiratory phase when the chest wall movement is large, thus reducing errors.

[0076] 2. Stimulation trigger point (stimulus activation).

[0077] When the respiratory signal drops and crosses the green trigger threshold line (falling edge), it reaches the Trigger Point. The system recognizes the start of exhalation and immediately switches the stimulation output to a high level (on state).

[0078] 3. Expiratory gating window (continuous stimulation).

[0079] The stimulus remains high for a duration of T. exp This stage corresponds to the stable period of exhalation, when the position of the chest is relatively fixed. At this time, outputting stimulation / treatment can significantly reduce positional deviations caused by respiratory movements.

[0080] 4. Window closed (stimulus off).

[0081] When T exp Once the timing ends, the stimulation output immediately switches back to a low level (off state), waiting for the trigger of the next respiratory cycle.

[0082] 5. Repeat in the next cycle.

[0083] The respiratory signal enters the next inspiratory-expiratory cycle. When it crosses the trigger threshold again from high to low, the stimulation is turned on again, thus achieving the cyclical control of "expiratory synchronous stimulation".

[0084] This disclosure reconstructs the respiratory cycle by collecting physiological signals and precisely pinpoints the start of expiration using a trigger threshold, triggering vagal nerve electrical stimulation only during the expiratory phase. Leveraging the higher sensitivity of the vagus nerve during expiration, it achieves precise timing of stimulation, enhancing neuromodulation while avoiding side effects such as heart rate fluctuations and physical discomfort caused by stimulation at inappropriate times.

[0085] In one embodiment, the system further includes a fuzzy PID controller, and the control unit is connected to the fuzzy PID controller; The control unit is used to calculate the RMSSD as the heart rate variability index based on the photoplethysmography (PPS) signal, where RMSSD is the root mean square of the difference between adjacent heartbeat intervals. The specific calculation formula is as follows: Let the extracted pulse interval sequence be PP1, PP2, ..., PP... Q (Q is the number of intervals, in milliseconds), then .

[0086] In one possible implementation, the heart rate variability index may also include high-frequency (HF) energy, which is calculated by performing a fast Fourier transform on the heartbeat interval sequence to obtain the power spectral density, and integrating it in the frequency range of 0.15–0.4 Hz to obtain the high-frequency power PHF. This index reflects the intensity of parasympathetic nerve activity.

[0087] In this embodiment, HF energy is mainly used to verify the consistency of RMSSD calculation results. Every second, the system calculates the RMSSD value over the past 5 seconds, denoted as RMSSD(k), where k=1,2,3,... are the calculation sequence numbers (i.e., updated every second). Simultaneously, the HF energy within the corresponding window is calculated. If the RMSSD shows relaxation (value increases) but the HF energy is abnormally low (e.g., below 50% of the individual baseline value), it is determined that the RMSSD calculation may be unreliable due to motion artifacts or poor signal quality. In this case, the system temporarily suspends the adjustment of stimulus intensity, i.e., maintains the current stimulus intensity unchanged, and waits for the next calculation sequence number. Subsequent step adjustments are only performed when the HF energy and RMSSD change trends are consistent. The control steps of the second closed loop are described in detail below.

[0088] The control unit is used to calculate the deviation between the RMSSD and the target vagal tone threshold, as well as the rate of change of the deviation. Let the preset target vagal tone threshold be T. target (Unit: ms). For the current k-th frame (each frame is a 5-second sliding window), the real-time calculated RMSSD value is RMSSD(k). The deviation D(k) at the current moment represents the gap between the current tension and the relaxation target: ; Calculate the rate of change of deviation C(k) at the current moment, which represents the trend of escalating or easing tension: ; Where D(k-1) is the deviation of the previous calculation window. For the first window (k=1), the rate of change C(1) can be set to zero.

[0089] The fuzzy PID controller is used to determine the step adjustment amount based on the deviation and the rate of change of the deviation.

[0090] Specifically, in determining the step adjustment amount based on the deviation and the rate of change of the deviation, the fuzzy PID controller is specifically configured to perform the following steps B1-B3: B1. Determine the target deviation level of the deviation based on multiple preset deviation levels; wherein the deviation value ranges corresponding to different deviation levels do not overlap.

[0091] B2. Based on multiple preset rate of change levels, determine the target rate of change level of the deviation; wherein the rate of change value intervals corresponding to different rate of change levels do not overlap.

[0092] The system presets a threshold range and directly maps continuous deviations D(k) and rates of change C(k) to discrete physiological state levels, thus achieving fuzzification processing.

[0093] The deviation D(k) is divided into four levels: PB (severe tension / extremely large gap, e.g., D(k)>5ms); PS (Mild tension / small gap, e.g., 1) <D(k)≤5ms); ZO (perfect relaxation / equilibrium, e.g., -1≤D(k)≤1ms); NB (overactivation / risk of bradycardia, e.g., D(k) < -1ms).

[0094] The rate of change C(k) is divided into three levels: PB (tension increases rapidly, e.g., C(k) > 1ms / window); ZO (state stationary, e.g., -1≤C(k)≤1ms / window); NB (Tense is rapidly relieved, for example, C(k) < -1ms / window).

[0095] The numerical ranges corresponding to different levels do not overlap, ensuring that each input uniquely corresponds to one level.

[0096] B3. Determine the target step adjustment amount according to the preset fuzzy rule table, wherein the preset fuzzy rule table is used to indicate the mapping relationship between different target deviation levels, different rate of change levels and step adjustment amount; wherein, when the target deviation level is a negative deviation level, the step adjustment amount is zero.

[0097] Based on the deviation level and rate of change level obtained in the previous step, the step adjustment amount ΔP is determined using a preset fuzzy rule table. cur (k) (unit mA). Preset fuzzy rules are shown in Table 1: Table 1 Note: When the deviation level is NB, it means that the real-time RMSSD has greatly exceeded the target value. In order to prevent the induction of bradycardia, the system ignores any rate of change trend and forces the step adjustment to zero, and the corresponding current output is directly reduced to zero.

[0098] In one embodiment, the control unit is further configured to: after superimposing the current stimulus intensity with the step adjustment amount to obtain the target stimulus intensity, if the target stimulus intensity is greater than a preset upper threshold, then limit the target stimulus intensity to the upper threshold; if the target stimulus intensity is less than a preset lower threshold, then limit the target stimulus intensity to the lower threshold.

[0099] The control unit is used to superimpose the current stimulation intensity with the step adjustment amount to determine the intensity of the target electrical stimulation, and control the stimulation unit to perform electrical stimulation at the intensity of the target electrical stimulation.

[0100] The step adjustment amount ΔP obtained from the table is... cur (k), calculate the final output stimulation current P in this window. cur (k): ; Among them, P cur (k-1) represents the stimulation current output from the previous window (initial value can be set to 0.2 mA). To ensure medical electrical safety, the system sets an output limit: if P cur If (k) > 3.0mA, then the amplitude limit is 3.0mA; if P cur If (k) < 0.2 mA and the current state is not zero (i.e., the deviation level is not NB), then the limit is 0.2 mA; if P is caused by NB returning to zero... cur If (k) is below 0.2mA, then output 0mA directly (safety suspend). Finally, the control unit will limit the P... cur (k) is the target stimulus intensity, and the stimulation unit is controlled to output electrical stimulation at that intensity.

[0101] Through the aforementioned adaptive step-by-step control algorithm based on fuzzy PID, combined with the consistency verification of HF energy, the system can dynamically adjust the stimulation current according to the user's real-time vagal nerve tension. When tense, the stimulation intensity is gradually increased to promote relaxation, and when excessively relaxed, the stimulation intensity is decreased or reduced to zero to avoid bradycardia. At the same time, it avoids erroneous adjustment caused by poor signal quality, thus achieving safe and effective second closed-loop control.

[0102] In one embodiment, the stimulation unit includes a stimulation pulse generator, which includes a pulse generation terminal and a synchronization signal output terminal; The stimulation pulse generator is controlled by the control unit and configured to: output an electrical stimulation signal through the pulse generation terminal under the control of the control unit, and simultaneously output a synchronization signal through the synchronization signal output terminal; the stimulation pulse generator is a circuit module capable of generating controllable electrical stimulation pulses, integrating functions such as digital-to-analog conversion, waveform shaping, and power output. The stimulation pulse generator has two independent output terminals: a pulse generation terminal and a synchronization signal output terminal. The pulse generation terminal outputs a high-voltage, limited-current stimulation pulse waveform (e.g., a biphasic exponential wave or a symmetrical square wave) directly applied to the ear stimulation electrodes; the synchronization signal output terminal outputs a low-voltage logic level signal (typically TTL level, active high, width the same as or slightly wider than the stimulation pulse width) that is strictly time-aligned with the stimulation pulse. This signal itself does not provide stimulation energy but serves only as a timing indicator. The stimulation pulse generator is controlled by the control unit. The control unit determines when a stimulation pulse needs to be output based on dual closed-loop control logic (expiratory gating and HRV feedback). Once the control unit issues a trigger command, the stimulation pulse generator is configured to: under the control of the control unit, output an electrical stimulation signal through the pulse generation terminal; and simultaneously, at the start moment (rising edge) of the output electrical stimulation signal, immediately output a synchronization signal through the synchronization signal output terminal. These two actions occur simultaneously at the hardware level, with a time deviation of less than 1 microsecond.

[0103] The sensing unit includes an analog front-end module, which has a synchronization signal input terminal; the synchronization signal output terminal of the stimulation pulse generator is directly connected to the synchronization signal input terminal of the analog front-end module via a hardwire; the analog front-end module is configured to: stop acquiring the photoplethysmography signal and maintain the current sampling value during the receipt of the synchronization signal.

[0104] The analog front-end module (AFE) is a key circuit for acquiring PPG signals. It converts photocurrent into voltage and performs amplification, filtering, and analog-to-digital conversion. This AFE has a dedicated synchronization signal input (also referred to as a "blank pin" or "sample-and-hold control pin" in some chip datasheets). The function of this input is to force the signal acquisition chain within the AFE into a special state when a valid level (e.g., high) is received.

[0105] The synchronization signal output terminal of the stimulation pulse generator and the synchronization signal input terminal of the analog front-end module are directly connected via a separate hardwire. "Direct hardwire connection" means that there are no intermediate processing units (such as control units, level converters, or software logic) between them; they are directly connected through copper foil traces on the circuit board. This connection method ensures extremely low latency (nanosecond level) and high reliability in synchronization signal transmission.

[0106] The analog front-end module is configured to: stop acquiring the photoplethysmography (PPG) signal and maintain the current sampled value during the receipt of the synchronization signal. Specifically, when the synchronization signal is valid (e.g., high level), the analog switch inside the AFE disconnects the sensor input path, preventing the weak current signal from the PPG photodetector from entering the amplification stage; simultaneously, the AFE's output data register is locked to the last valid sampled value before the arrival of the synchronization signal and is no longer updated. The AFE only resumes normal acquisition when the synchronization signal is withdrawn (goes low level).

[0107] The working principle and timing of this embodiment are as follows: When the control unit decides to output the stimulation pulse, the stimulation pulse generator simultaneously outputs the stimulation pulse and a synchronization signal. Due to the direct hardwired connection, the synchronization signal arrives at the synchronization signal input of the AFE almost instantaneously. The AFE immediately enters the "sample and hold" state, which lasts for the entire stimulation pulse period (e.g., 200μs to 500μs). During this period, the AFE no longer samples the PPG signal, which is subject to strong electromagnetic interference from the stimulation pulse, and its output remains at the last valid value before the interference. After the stimulation pulse ends, the synchronization signal also disappears, and the AFE resumes acquisition. At this time, the stimulation trail in the PPG signal has decayed, and subsequent sampling will obtain a clean waveform.

[0108] The core technological advantages of this mechanism lie in its microsecond-level response, achieved entirely in hardware, which completely eliminates synchronization deviations caused by software response delays. It also involves actively shielding the sensor channel during the duration of the stimulation pulse, blocking interference from entering the signal chain at its source, resulting in higher reliability and signal-to-noise ratio compared to pure software post-processing filtering. Experimental verification shows that after hardware blanking synchronization using this embodiment, combined with subsequent blind source separation and Kalman filtering algorithms, the recognition rate of respiratory / heart rate components in the PPG signal during stimulation can be increased to over 90%, significantly improving the stability and accuracy of the dual-loop control system.

[0109] Figure 3 This is a schematic diagram of the PPG waveform (AFE Sync Signal) after hardware synchronization blanking only, as shown below. Figure 3 As shown, the white waveform represents the physiological signal with only AFE hardware synchronization blanking enabled and without MCC-KF digital filtering. Relying on stimulation synchronization sampling and holding, stimulation pulse spike interference has been removed, but the waveform still contains small residual noise, with many waveform spikes and large baseline fluctuations.

[0110] Figure 4 This is a waveform diagram illustrating the combined hardware blanking and MCCKF processing, as shown below. Figure 4 As shown, compared to Figure 3 The waveform is filtered out significantly, and the main waveform of breathing / pulse is completely preserved with a smooth outline, which is an effective and clean signal for the backend to identify the expiratory phase and calculate RMSSD.

[0111] Figure 5 This is a diagram illustrating the comparison of noise power spectral density, as shown below. Figure 5 As shown, the light green curve "Original Noise" represents the noise power of the original signal containing stimuli and interference, with high peak noise energy in the characteristic frequency band; the dark blue curve "Filtered Signal" represents the noise power after AFE+MCC-KF processing, with a significant drop in noise energy across the entire frequency band, effectively suppressing high-frequency interference while preserving the inherent frequency band energy of the physiological signal. This intuitively demonstrates that the combined scheme significantly suppresses noise power.

[0112] Figure 6 This is a diagram illustrating the coherence factor comparison, as shown below. Figure 6 As shown, the blue "No Filter" indicates that the original signal without filtering has low coherence and poor integrity of the effective physiological signal; the green "MCC-KF Filtered" indicates that the coherence coefficient is significantly increased after filtering, representing that the filtered physiological signal has higher consistency with the real human physiological waveform and less distortion.

[0113] Depend on Figures 3-6The four sets of test results show that: AFE hardware blanking alone can eliminate large-amplitude pulse interference, and the superimposed MCCKF adaptive filtering can further filter out residual noise and completely preserve physiological waveforms; the hardware and software collaborative noise reduction scheme of this invention can preserve the effective PPG information to the maximum extent while suppressing stimulation interference, and ensure the accurate extraction of respiratory phase and HRV indicators, providing data support for the reliable operation of the dual closed-loop regulation of this invention.

[0114] In one embodiment, the control unit is further configured to perform the following steps C1-C2: C1. Real-time acquisition of gain change parameters of the analog front-end module.

[0115] During PPG signal acquisition, to adapt to different individuals' skin translucency, blood perfusion levels, and sensor coupling states, the analog front-end module (AFE) in the sensing unit may automatically or according to instructions switch amplification gain. For example, when the signal amplitude is too small, the AFE may switch the gain from ×100 to ×200; when the signal is too large and may saturate, it will switch the gain back from ×200 to ×100. This gain switching changes the scaling ratio of the original ADC sample value SADC(t), directly causing non-physiological step jumps in the output waveform.

[0116] The control unit communicates via an internal bus (such as SPI or I2C). 2 The control unit reads the current gain setting value from the AFE's internal register in real time via the C interface, denoted as G(t) (dimensionless, e.g., G=100, 200, 400, etc.). Simultaneously, the control unit also acquires other relevant parameters, including: System offset voltage V offset (Unit: V or LSB, obtained through factory calibration, representing the zero-point offset when there is no signal input). LED drive current I LED (Unit: mA, set by the control unit, known value); Input bias current I bias (Unit: nA, provided by AFE chip datasheet).

[0117] These parameters together constitute a complete description of the gain variation.

[0118] C2. Based on the gain change parameters, the original sampled values ​​are normalized to convert the signal step caused by the gain jump into a continuous physiological waveform, so as to obtain the photoplethysmography (PPG) signal.

[0119] The control unit performs a normalization calculation on each raw sample point SADC(t), mapping it to a uniform scale independent of gain. The normalization formula is as follows: ; Wherein: S ADC (t): Raw ADC sampled value (dimensionless, usually a 12-bit or 16-bit integer value); V offset : ADC code value (dimensionless) corresponding to the system offset voltage, obtained through calibration, used to eliminate zero-point drift; G(t): Gain setting value at the current moment (dimensionless); K cal (I LED ,I bias ): Correction factor, a function related to LED drive current and bias current, which can be obtained through factory calibration (e.g., K). cal =I ref / I LED , where I ref (This is a reference current value) used to compensate for the impact of light source intensity fluctuations on signal amplitude.

[0120] When the gain changes (e.g., from G=100 to G=200), a step with a sudden amplitude change occurs before and after the change point in the unnormalized original signal. After the normalization process described above, this step is eliminated because the gain factor G(t) is removed, and the signal becomes a smooth, continuous waveform. Experiments show that this method can reduce the baseline abrupt change rate caused by gain switching from over 50% to below 5%, eliminating more than 95% of artifacts.

[0121] The normalized signal Snorm(t) is a continuous photoplethysmography (PPG) signal, which can be further used for respiratory phase recognition and heart rate variability calculation. This parameter-aware signal reconstruction algorithm does not depend on specific hardware implementation details. It can accurately compensate for hardware gain changes in the digital domain using only real-time read gain parameters and known calibration information, thereby ensuring the accuracy of subsequent physiological parameter extraction.

[0122] This system effectively solves the problems of interference and false triggering in traditional ear VNS devices through both hardware-level and algorithm-level measures. The hardware-synchronized blanking mechanism combined with algorithmic denoising (such as maximum correlation entropy Kalman filtering) increases the recognition rate of respiratory / heart rate components of the PPG signal to over 90% during stimulation; traditional open-loop or simple filtering often fails to recognize these components. Parameter normalization reduces the baseline abrupt change rate before and after ADF gain jumps from >50% to <5%, eliminating over 95% of artifacts and avoiding erroneous control caused by false heart rate spikes. By using IMU+EDA linkage to eliminate motion artifacts, this system achieves full-dimensional interference removal, providing safety degradation protection for sensitive devices. When combined with respiratory gating triggering and HRV feedback regulation, it can conform to respiratory sinus rhythm (RSA) and autonomic rhythm, achieving a resonant enhancement of neural stimulation and endogenous regulatory mechanisms. In summary, this disclosure significantly improves the accuracy and safety of treatment and is suitable for various applications such as anxiety relief, sleep regulation, and blood pressure control.

[0123] Figure 7 The hardware architecture of an auricular vagus nerve modulation system provided in one embodiment of this application is as follows: Figure 7 As shown, it includes: The integrated in-ear electrode sensing module integrates four types of devices: PPG photosensitive sensor, Stim electrode, six IMUs, and EDA (electrodermal absorption controller), all integrated into the ear-worn end. One path transmits the PPG physiological signal via a shielded wire and a TIA (transimpedance amplification) to the adaptive AFE (Automatic External Wire) front end, with the shielding wire reducing spatial electromagnetic interference. The other path leads to a hardware synchronization trace, synchronously transmitting stimulation and blanking trigger signals to achieve time-series linkage of the entire device. The IMUs and EDA acquire motion and electrodermal physiological data, which are subsequently used to eliminate PPG motion artifacts.

[0124] The adaptive AFE blanking front end, as the PPG signal analog processing unit, receives two hardware synchronization levels from the pulse generator and the in-ear module. When the system outputs electrical stimulation, the synchronization signal immediately triggers the AFE to enter blanking mode. The channel pauses data sampling and maintains the original sampling value, isolating the stimulation pulse from interference with PPG acquisition at the hardware level. The filtered clean physiological signal is then sent to the back-end MCU.

[0125] The pulse generator (TavnsStim) is the electrical stimulation output unit of this system. It is controlled by low-power MCU instructions. On the one hand, it generates electrical stimulation waveforms that are applied to the vagus nerve via in-ear electrodes. On the other hand, it outputs a hard-wired synchronization signal to synchronously output blanking trigger level to the front end of AFE, and synchronously realizes the linkage logic of "stimulation output, sampling immediately turned off".

[0126] The low-power MCU (the core of the whole machine algorithm and control) is internally divided into three functional layers: ①S1 Integrated parameter sensing: Receives PPG, IMU, and EDA data output from AFE, identifies and removes motion interference signal segments based on IMU and EDA data, completes PPG normalization and blind source separation, and separates respiratory and pulse components; ②S2 MCC-KF feature extraction: Conducts signal quality assessment based on kurtosis and spectral entropy, dynamically corrects Kalman filter noise parameters, and outputs clean breathing waveforms and usable PPG signals after MCC-KF filtering; ③S3 Dual closed-loop decision-making: The first loop is a respiratory gating closed loop, which determines the exhalation start point through quadratic polynomial fitting and envelope slope, and only opens stimulation output during the exhalation phase; The second loop is a HRV fuzzy PID closed loop, which calculates the RMSSD index from PPG, performs fuzzy classification based on index deviation and deviation change rate, automatically calculates the stimulation intensity step size and sets upper and lower limit amplitude values, and finally sends control commands to the pulse generator.

[0127] At the upper level, the MCU establishes data interaction with the upper level via BLE Bluetooth. The upper computer receives and displays the device's HRV data, respiratory phase status, and current stimulation conditions in real time for parameter viewing and device debugging.

[0128] In one embodiment, when the system includes an AFE, a six-axis inertial measurement unit (IMU), and an electrical activity sensor (EDA), the above-mentioned normalization processing of the photoplethysmography (PPG) signals of the red and infrared channels includes the following sub-steps: S1. Acquire multispectral PPG raw signal: Use 660nm red light and 940nm infrared dual optical paths to acquire photoplethysmography (PPG) signal, set the sampling rate to 166Hz~200Hz, and input the raw PPG signal to the AFE adaptive cancellation front end.

[0129] S2 and AFE, in conjunction with a hardware-synchronized blanking mechanism, perform sample-and-hold blanking during the electrical stimulation pulse output period to filter out spike interference introduced by the stimulation pulse and output the hardware-preprocessed PPG raw data X. raw .

[0130] S3. Motion state recognition is achieved by acquiring acceleration and angular velocity data through a six-axis IMU, and motion interference such as sensor displacement and head slippage is determined. Artifact classification is performed using an EDA module to distinguish between physiological fluctuations and electrode loosening artifacts, and a motion compensation coefficient K is generated. ca .

[0131] S4, EDA, skin electrical activity, collects skin conductance change signals, distinguishes the causes of artifacts, such as physiological fluctuations caused by emotional fluctuations or hardware artifacts caused by loose or poor contact of sensors, and helps to distinguish between valid signals and invalid noise.

[0132] S5, according to the formula Complete signal correction; where X norm X is the PPG output signal after normalization correction. raw This refers to the raw PPG sampling data after AFE hardware blanking and without parameter compensation; G afe I is the AFE gain parameter. bias I is the reference rated bias current for the LED. led For LED drive current parameters, K is used to compensate for baseline offset caused by front-end gain jumps and light source current fluctuations. ca Combining artifact recognition results from IMU and EDA to compensate for motion noise; S6. Output normalized PPG signal: After the above correction, the PPG baseline jump is reduced from more than 50% of the ADC range to within 5% of the ADC range, and the normalized photoplethysmography signal is obtained, which is then transmitted to carry out blind source separation and respiratory source signal extraction.

[0133] In one exemplary embodiment, such as Figure 8 As shown, a method for modulating vagus nerve stimulation is provided. This method is applied to the vagus nerve stimulation system as described in any of the above embodiments. The method includes the following steps S101-S103: S101, the stimulation unit applies electrical stimulation to the vagus nerve of the ear.

[0134] S102, The sensing unit collects the user's photoplethysmography (PPG) signal.

[0135] S103, the control unit is connected to the stimulation unit and the sensing unit respectively, identifies the expiratory phase based on the photoplethysmography (PPG) signal, and controls the stimulation unit to output electrical stimulation only during the expiratory phase; and calculates the heart rate variability index based on the PPG signal, and dynamically adjusts the intensity of the electrical stimulation based on the deviation of the heart rate variability index relative to the target vagal nerve tension threshold and the rate of change of the deviation.

[0136] In one embodiment, identifying the expiratory phase based on the photoplethysmography signal includes: The respiratory waveform is separated from the photoplethysmography signal; Locate the extreme points from the respiratory waveform; Calculate the slope of the envelope of the respiratory waveform; The moment when the slope of the envelope changes from a positive value to a negative value, and the extreme point of the respiratory waveform at the location of the slope change is the maximum value, is determined as the expiration start point of the expiration phase. The expiratory phase is determined based on the preset expiratory phase interval and the expiratory initiation point.

[0137] In one embodiment, the sensing unit includes: a PPG sensor, the PPG sensor including a red light source and an infrared light source; the step of separating the respiratory waveform from the photoplethysmography signal includes: Normalize the photoplethysmography (PPG) signals from the red and infrared channels. The respiratory source signal was separated from the normalized photoplethysmography signal. The respiratory source signal is denoised using a maximum correlation entropy Kalman filter to output a respiratory waveform. The measurement noise covariance of the Kalman filter is adjusted based on the signal quality assessment results of the respiratory source signal.

[0138] In one embodiment, the system further includes: a six-axis inertial measurement unit and electrodermal activity (EDA) measurement. Before performing denoising processing on the respiratory source signal using maximum correlation entropy Kalman filtering, the method further includes: Based on the signals output by the six-axis inertial measurement unit and the skin conductance sensor, the motion artifact time periods in the photoplethysmography signal are identified and the corresponding signal segments are removed to obtain the respiratory source signal after artifact removal. In the aspect of denoising the respiratory source signal using maximum correlation entropy Kalman filtering, the control unit is specifically configured to: The artifact-removed respiratory source signal was denoised using a maximum correlation entropy Kalman filter.

[0139] In one embodiment, the system further includes: a fuzzy PID controller, and the control unit is connected to the fuzzy PID controller; the method further includes: The control unit calculates RMSSD as the heart rate variability index based on the photoplethysmography pulse wave signal. The control unit calculates the deviation between the RMSSD and the target vagal nerve tension threshold, as well as the rate of change of the deviation; The fuzzy PID controller determines the step adjustment amount based on the deviation and the rate of change of the deviation; The control unit superimposes the current stimulation intensity with the step adjustment amount to determine the intensity of the target electrical stimulation, and controls the stimulation unit to perform electrical stimulation at the intensity of the target electrical stimulation.

[0140] In one embodiment, determining the step adjustment amount based on the deviation and the rate of change of the deviation includes: Based on multiple preset deviation levels, the target deviation level of the deviation is determined; wherein the deviation value ranges corresponding to different deviation levels do not overlap. Based on multiple preset rate of change levels, the target rate of change level of the deviation is determined; wherein the ranges of rate of change values ​​corresponding to different rate of change levels do not overlap. The target step adjustment amount is determined according to a preset fuzzy rule table, which is used to indicate the mapping relationship between different target deviation levels, different rate of change levels and step adjustment amount; Specifically, when the target deviation level is a negative deviation level, the step adjustment amount is zero.

[0141] In one embodiment, the method further includes: after superimposing the current stimulus intensity with the step adjustment amount to obtain a target stimulus intensity, if the target stimulus intensity is greater than a preset upper threshold, then limiting the target stimulus intensity to the upper threshold; if the target stimulus intensity is less than a preset lower threshold, then limiting the target stimulus intensity to the lower threshold.

[0142] In one embodiment, the stimulation unit includes a stimulation pulse generator, which includes a pulse generation terminal and a synchronization signal output terminal; The stimulation pulse generator is controlled by the control unit. Under the control of the control unit, it outputs an electrical stimulation signal through the pulse generation terminal and outputs a synchronization signal through the synchronization signal output terminal at the same time as outputting the electrical stimulation signal. The sensing unit includes an analog front-end module, which has a synchronization signal input terminal; the synchronization signal output terminal of the stimulation pulse generator is directly connected to the synchronization signal input terminal of the analog front-end module via a hardwire; during the receipt of the synchronization signal, the acquisition of the photoplethysmography signal is stopped and the current sampling value is maintained.

[0143] In one embodiment, the method further includes: Real-time acquisition of gain change parameters of the simulated front-end module; Based on the gain change parameters, the original sampled values ​​are normalized to convert the signal step caused by the gain jump into a continuous physiological waveform, so as to obtain the photoplethysmography (PPG) signal.

[0144] This disclosure integrates an ear stimulation electrode, a PPG sensor, motion and skin conductance sensors, and a low-power MCU control unit. It can acquire respiratory phase and heart rate variability (HRV) information in real time and employs a "dual closed-loop" strategy: the first closed loop triggers vagal nerve electrical stimulation only during the expiratory phase; the second closed loop adaptively adjusts the stimulation current using a fuzzy PID controller based on the real-time calculated RMSSD deviation and rate of change under HRV. The system employs parallel hardware blanking and algorithmic filtering (including PCA / ICA blind source separation and Kalman filtering) to suppress motion artifacts and stimulation interference, thereby improving the robustness of signal extraction.

[0145] Taking a daily anxiety management scenario as an example, the system continuously collects the wearer's PPG and IMU / EDA signals. The MCU first detects signs of sympathetic activation (e.g., a continuously decreasing RMSSD and an increased LF / HF, indicating an anxiety state). Then, the first closed loop is activated, identifying the expiratory initiation point through quadratic fitting; low-frequency electrical stimulation (e.g., initial intensity 0.2mA) is triggered at the beginning of each expiratory phase window. Simultaneously, the second closed loop is entered, where a fuzzy PID controller adjusts the stimulation current based on the real-time RMSSD deviation and its rate of change (e.g., increasing the current stepwise if RMSSD is low, otherwise decreasing it). When physiological indicators return to the individual baseline (RMSSD recovers, LF / HF recovers), the system automatically reduces or pauses stimulation.

[0146] Taking meditation and relaxation training as an example, this disclosed system is integrated into a wearable earpiece for neuromodulation during meditation and relaxation. In this embodiment, the system utilizes the meditation instructor's cue cycle to trigger a single stimulus (i.e., S1 skipping) and monitors the short-term response of HRV in real time. Depending on the needs of different meditation stages, soothing stimulation (enhancing parasympathetic response) can be actively applied during the exhalation phase, or a weak stimulus can be applied during the inhalation phase (slightly activating sympathetic response). Fuzzy PID rules can be fine-tuned for this purpose, for example, reducing stimulation when RMSSD is excessive and slightly increasing stimulation when RMSSD is low.

[0147] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0148] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A vagus nerve stimulation and modulation system, characterized in that, The system includes: Stimulation unit, used to apply electrical stimulation to the vagus nerve of the ear; The sensing unit is used to collect the user's photoplethysmography (PPG) signal; A control unit, connected to both the stimulation unit and the sensing unit, is configured to identify the expiratory phase based on the photoplethysmography (PPG) signal and control the stimulation unit to output electrical stimulation only during the expiratory phase; and to calculate a heart rate variability index based on the PPG signal and dynamically adjust the intensity of the electrical stimulation based on the deviation of the heart rate variability index from the target vagal tone threshold and the rate of change of the deviation.

2. The system according to claim 1, characterized in that, In identifying the expiratory phase based on the photoplethysmography (PPG) signal, the control unit is specifically configured to: The respiratory waveform is separated from the photoplethysmography signal; Locate the extreme points from the respiratory waveform; Calculate the slope of the envelope of the respiratory waveform; The moment when the slope of the envelope changes from a positive value to a negative value, and the extreme point of the respiratory waveform at the location of the slope change is the maximum value, is determined as the expiration start point of the expiration phase. The expiratory phase is determined based on the preset expiratory phase interval and the expiratory initiation point.

3. The system according to claim 2, characterized in that, The sensing unit includes a PPG sensor; and in terms of separating the respiratory waveform from the photoplethysmography signal, the control unit is configured to: Normalize the photoplethysmography (PPG) signals from the red and infrared channels. The respiratory source signal was separated from the normalized photoplethysmography signal. The respiratory source signal is denoised using a maximum correlation entropy Kalman filter to output a respiratory waveform. The measurement noise covariance of the Kalman filter is adjusted based on the signal quality assessment results of the respiratory source signal.

4. The system according to claim 3, characterized in that, The system also includes: a six-axis inertial measurement unit and skin electrical activity measurement; Before performing denoising processing on the respiratory source signal using maximum correlation entropy Kalman filtering, the control unit is further configured to: Based on the signals output by the six-axis inertial measurement unit and the skin conductance sensor, the motion artifact time periods in the photoplethysmography signal are identified and the corresponding signal segments are removed to obtain the respiratory source signal after artifact removal. In the aspect of denoising the respiratory source signal using maximum correlation entropy Kalman filtering, the control unit is specifically configured to: The artifact-removed respiratory source signal was denoised using a maximum correlation entropy Kalman filter.

5. The system according to claim 1, characterized in that, The system further includes a fuzzy PID controller, and the control unit is connected to the fuzzy PID controller; The control unit is used to calculate RMSSD as the heart rate variability index based on the photoplethysmography signal. The control unit is used to calculate the deviation between the RMSSD and the target vagal nerve tension threshold and the rate of change of the deviation; The fuzzy PID controller is used to determine the step adjustment amount based on the deviation and the rate of change of the deviation; The control unit is used to superimpose the current stimulation intensity with the step adjustment amount to determine the intensity of the target electrical stimulation, and control the stimulation unit to perform electrical stimulation at the intensity of the target electrical stimulation.

6. The system according to claim 5, characterized in that, In determining the step adjustment amount based on the deviation and the rate of change of the deviation, the fuzzy PID controller is specifically configured to: Based on multiple preset deviation levels, the target deviation level of the deviation is determined; wherein the deviation value ranges corresponding to different deviation levels do not overlap. Based on multiple preset rate of change levels, the target rate of change level of the deviation is determined; wherein the ranges of rate of change values ​​corresponding to different rate of change levels do not overlap. The target step adjustment amount is determined according to a preset fuzzy rule table, which is used to indicate the mapping relationship between different target deviation levels, different rate of change levels and step adjustment amount; Specifically, when the target deviation level is a negative deviation level, the step adjustment amount is zero.

7. The system according to claim 5, characterized in that, The control unit is further configured to: after superimposing the current stimulus intensity with the step adjustment amount to obtain the target stimulus intensity, if the target stimulus intensity is greater than a preset upper limit threshold, then limit the target stimulus intensity to the upper limit threshold. If the target stimulus intensity is less than a preset lower threshold, then the target stimulus intensity is limited to the lower threshold.

8. The system according to claim 3, characterized in that, The stimulation unit includes a stimulation pulse generator, which includes a pulse generation terminal and a synchronization signal output terminal. The stimulation pulse generator is controlled by the control unit and is configured to: under the control of the control unit, output an electrical stimulation signal through the pulse generation terminal, and output a synchronization signal through the synchronization signal output terminal at the same time as outputting the electrical stimulation signal; The sensing unit includes an analog front-end module, which has a synchronization signal input terminal; the synchronization signal output terminal of the stimulation pulse generator is directly connected to the synchronization signal input terminal of the analog front-end module via a hardwire. The analog front-end module is configured to: stop acquiring the photoplethysmography signal and maintain the current sampling value during the period when the synchronization signal is received.

9. The system according to claim 8, characterized in that, The control unit is also configured to: Real-time acquisition of gain change parameters of the simulated front-end module; Based on the gain change parameters, the original sampled values ​​are normalized to convert the signal step caused by the gain jump into a continuous physiological waveform, so as to obtain the photoplethysmography (PPG) signal.

10. A method for modulating vagus nerve stimulation, characterized in that, The method is applied to the auricular vagus nerve stimulation modulation system as described in any one of claims 1-9, and the method includes: The stimulation unit applies electrical stimulation to the vagus nerve of the ear; The sensing unit collects the user's photoplethysmography (PPG) signal; The control unit is connected to the stimulation unit and the sensing unit respectively, identifies the expiratory phase based on the photoplethysmography (PPG) signal, and controls the stimulation unit to output electrical stimulation only during the expiratory phase; and calculates a heart rate variability index based on the PPG signal, and dynamically adjusts the intensity of the electrical stimulation based on the deviation of the heart rate variability index relative to the target vagal nerve tension threshold and the rate of change of the deviation.

Images