Electric toothbrush control method and system based on oral microecological balance

Abstract

The application discloses a kind of electric toothbrush control method and system based on oral microecological balance, its method includes: real-time acquisition multi-modal sensor oral environment data;The oral environment data collected is time registration and nonlinear adaptive filtering preprocessing, obtain the oral environment data after preprocessing;From the oral environment data after preprocessing, extract time domain feature and frequency domain feature, and determine the dynamic weight of each time domain feature and frequency domain feature using entropy method;According to the dynamic weight, health baseline value, sensitivity adjustment coefficient, historical influence factor, calculate microecological health index MHI value;According to the MHI value dynamically selected cleaning mode calculated in real time, and control motor by PID controller execution.The application realizes the comprehensive quantitative evaluation of oral health condition by multi-modal sensor fusion and innovative microecological health index model, overcomes the limitation of traditional single parameter detection.

Description

Technical Field

[0001] This invention belongs to the field of intelligent oral care devices. Specifically, it is an intelligent electric toothbrush control method and system that integrates multimodal sensors, can assess the balance of the oral microecology in real time, and dynamically adjust the cleaning strategy. Background Technology

[0002] With the increasing popularity of electric toothbrush technology, current products have evolved from basic mechanical cleaning to devices with multiple preset modes and intelligent timer functions. Traditional electric toothbrushes mainly rely on mechanical vibration or rotation for cleaning. While some high-end models have introduced simple intelligent functions such as pressure sensing and Bluetooth connectivity, they generally suffer from the drawback of static, patterned operation and cannot dynamically respond to real-time changes in the oral microbiome. These devices still have significant shortcomings in achieving truly personalized care and maintaining a balanced oral microbiome.

[0003] In existing technologies, such as the invention patent with publication number CN113116579A, an electric toothbrush that can detect oral gases and adjust its operating parameters according to gas composition is described. However, its sensor type is limited, and it can only reflect a limited range of oral conditions. Another technical solution proposes to collect and analyze oral images using a camera, but this cannot monitor changes in the microenvironment in real time and poses a risk of privacy breaches. The main shortcomings of current electric toothbrushes are: First, most products rely on preset programs and cannot adapt to dynamic changes in the oral environment in real time. The oral microenvironment is affected by various factors such as diet, sleep, and hormones, and fixed cleaning modes cannot meet personalized needs. Second, existing sensing technologies are mostly single-modal, such as only detecting pressure or time, lacking comprehensive evaluation indicators for the oral microecology. For example, the utility model patent with publication number CN217219262U uses ultraviolet light to disinfect the brush head, but it does not consider the actual needs of microbial community balance. Finally, some studies attempt to achieve gesture recognition using accelerometers, such as the invention patent with publication number CN107334555A, but this can only control the on / off state and cannot quantify oral health status. These technical solutions fail to address the fundamental issues of oral flora balance and microenvironment adaptation, resulting in limited cleaning effectiveness and potentially even disrupting the oral microbial balance.

[0004] Therefore, there is an urgent need for an intelligent electric toothbrush system that can monitor, quantitatively evaluate, and dynamically adjust cleaning strategies in real time. Summary of the Invention

[0005] A brief overview of embodiments of the invention is provided below to provide a basic understanding of certain aspects of the invention. It should be understood that this overview is not an exhaustive summary of the invention. It is not intended to identify key or essential parts of the invention, nor is it intended to limit the scope of the invention. Its purpose is merely to present certain concepts in a simplified form as a prelude to the more detailed description that follows.

[0006] The present invention aims to overcome the shortcomings of the prior art and provide an electric toothbrush control method and system that can comprehensively and quantitatively assess the oral microecological health status and achieve dynamic and personalized cleaning accordingly.

[0007] According to one aspect of this application, an electric toothbrush control method based on oral microecological balance is provided, comprising: Real-time acquisition of oral environment data from multimodal sensors; the multimodal sensors include gas sensors, pH sensors, temperature sensors, pressure sensors, and optical sensors; The collected oral environment data is preprocessed by time registration and nonlinear adaptive filtering to obtain preprocessed oral environment data; the nonlinear adaptive filtering adopts an algorithm that combines hyperbolic tangent function and signal change rate to suppress mechanical vibration noise and electromagnetic interference; Temporal and frequency domain features are extracted from the preprocessed oral environment data, and the dynamic weights of each feature are determined using the entropy method. The calculation of temporal features includes mean, variance, peak value, and average gradient, which are intended to reflect the stability and fluctuation of parameters. The frequency domain features are calculated by short-time Fourier transform to determine the dominant frequency component, specific frequency band energy, and spectral entropy, which are used to reflect the periodic changes and microbial activity of the oral environment. The Microbiome Health Index (MHI) is calculated based on the dynamic weights, health baseline value, sensitivity adjustment coefficient, and historical influence factor. This value maps multidimensional features to a dimensionless value between 0 and 1. The health baseline value is obtained statistically from steady-state sensor data of healthy individuals and corresponds to the benchmark level of each sensor. The sensitivity adjustment coefficient corresponds to the dispersion index (e.g., standard deviation) of the healthy population data for each sensor indicator and is used to control the response sensitivity when features deviate from the baseline. The historical influence factor is used to integrate the impact of the previous MHI change trend on the current assessment, reflecting the continuity of microbiome changes. The cleaning mode is dynamically selected based on the real-time calculated MHI value. The MHI value and its changing trend are mapped to the vibration frequency, oscillation amplitude and recommended brushing time for the appropriate working conditions through a multi-objective optimization function. The motor is then controlled by a fuzzy PID controller.

[0008] Furthermore, the process of calculating the Microbiome Health Index (MHI) value includes: Multiple independent oral cavity zones were divided based on acceleration sensor signals; Based on the zoning results, the temporal and frequency domain features of each oral cavity zone are integrated at the sample level to form a unified feature sample set containing all zone data; the temporal and frequency domain features are standardized to eliminate dimensional differences. The feature weights are calculated based on the unified feature sample set, and the feature entropy is calculated based on the weights. The difference coefficient is calculated based on the feature entropy, and the dynamic weight of each feature is determined accordingly. Substituting the dynamic weights of each feature, the health baseline value, the sensitivity adjustment coefficient, and the historical impact factor into the exponential mapping formula yields the MHI value; where the historical impact factor controls the degree of influence of the previous MHI change trend on the current assessment.

[0009] As another refined approach, a method of independently calculating the weight and feature entropy within each partition can be adopted. Specifically, each partition calculates its own statistics. This involves defining a four-quadrant feature mapping function for the oral cavity using an oral cavity region recognition algorithm, and dividing the oral cavity region based on accelerometer data. Temporal and frequency domain features are then categorized into different regions (lower left, upper left, lower right, and upper right) based on the oral cavity region. The extracted temporal and frequency domain features are standardized to eliminate the influence of dimensions. The weights of the temporal and frequency domain features in different regions are then calculated, and the feature entropy of the temporal and frequency domain features is calculated based on these weights. Difference coefficients are calculated based on the feature entropy of the temporal and frequency domain features, and the weights of each feature are calculated based on these difference coefficients; features with smaller entropy values ​​have larger weights. Finally, the MHI value is calculated based on the weights of each feature, the healthy baseline value, the sensitivity adjustment coefficient, and historical influence factors.

[0010] Furthermore, the multi-objective optimization function takes the MHI value, MHI rate of change, and MHI integral as inputs, and outputs the target vibration frequency, target oscillation amplitude, and target remaining brushing time through a coefficient matrix and bias vector obtained offline by a particle swarm optimization algorithm.

[0011] Furthermore, the PID controller adopts fuzzy control, using the motor operation output deviation and the corresponding deviation change as fuzzy input. It infers the correction amount of the proportional, integral, and derivative parameters through multiple preset fuzzy subsets and the matching fuzzy inference rule base, and uses a standard defuzzification algorithm to realize the online adaptive adjustment of the PID parameters.

[0012] Furthermore, the dynamic selection of cleaning modes specifically includes: dividing the microecological health index into multiple health zones, and matching them with three gradient cleaning conditions: powerful cleaning, balanced cleaning, and gentle maintenance. Each condition corresponds to a different vibration frequency, brush head force, and zone cleaning duration. Specifically, when the microecological health index value is less than a first preset value, the powerful cleaning mode is entered; when the microecological health index value is greater than or equal to the first preset value and less than a second preset value, the balanced cleaning mode is entered; and when the microecological health index value is greater than or equal to the second preset value, the gentle maintenance mode is entered.

[0013] Furthermore, the method also includes a pressure protection step: when the pressure measured by the pressure sensor exceeds the safety threshold, the cleaning output intensity of the corresponding area is reduced, and the microecological health assessment weight of the area is corrected to reduce the cleaning intensity of the area and prevent gingival damage.

[0014] Furthermore, the method also includes: after the electric toothbrush stops working, generating a health report of this brushing session and synchronizing it to the mobile APP, while using the data from this brushing session to update the user's personalized model; the user's personalized model stores the user's exclusive oral benchmarks, feature weights, cleaning preferences and time-series health records, and is smoothly updated through an incremental learning mechanism.

[0015] Furthermore, the optical sensor employs two differentiated detection bands to perform dual-band reflectance spectral detection and analysis. The relative thickness of dental plaque is calculated by the ratio of the reflection intensities of the two bands, thereby eliminating interference from tooth color and ambient light.

[0016] According to a second aspect of this application, an electric toothbrush control system based on oral microecological balance is provided, comprising: The toothbrush head integrates multimodal sensors, including a gas sensor for detecting oral odor markers, a pH sensor for monitoring oral pH, a temperature sensor for measuring oral temperature in real time, a pressure sensor for detecting brushing force, and an optical sensor for analyzing plaque thickness through reflectance spectroscopy. The handle contains a microprocessor, a motor controller, and a power management module. The multimodal sensor transmits raw acquired data to the microprocessor via a bus; the microprocessor internally includes a filter calculation module, an oral cavity partitioning calculation module, an entropy health index calculation module, a multi-objective parameter mapping module, a fuzzy PID motor control module, and a personalized model storage and iterative update module. The microprocessor is configured to perform the steps of the electric toothbrush control method described above.

[0017] Furthermore, the brush head and the handle are magnetically connected, and the handle also has a wireless charging coil and a Bluetooth module embedded inside.

[0018] According to a third aspect of this application, a computer-readable storage medium is provided, on which a computer program is stored, wherein the computer program, when executed by a processor, implements the steps of the above-described electric toothbrush control method based on oral microecological balance.

[0019] Compared with the prior art, the present invention has the following beneficial effects: 1. By objectively calculating the dynamic weights of each feature using the entropy method, the bias of subjectively setting weights in traditional methods is avoided. This allows the system to automatically identify and prioritize features with more information, thereby achieving a precise quantitative assessment of oral health status.

[0020] 2. The Microbial Health Index (MHI) model is adopted to map multidimensional features into a single dimensionless index and innovatively introduces historical influencing factors to reflect the continuity of microecological changes, thus solving the problem that existing technologies cannot comprehensively assess the balance of oral flora and the state of the microenvironment.

[0021] 3. At the same time, a multi-objective optimization parameter mapping function was introduced to establish a real-time mapping mechanism from MHI value and its trend to cleaning parameters (frequency, force, time), enabling the toothbrush to dynamically adjust its cleaning strategy according to the real-time monitored changes in the oral microenvironment, thus realizing a fundamental shift from static, patterned cleaning to dynamic, adaptive cleaning. Attached Figure Description

[0022] The present invention can be better understood by referring to the description given below in conjunction with the accompanying drawings, in which the same or similar reference numerals are used throughout the drawings to denote the same or similar parts. These drawings, together with the following detailed description, are incorporated in and form part of this specification, and are used to further illustrate preferred embodiments of the invention and explain the principles and advantages of the invention. In the drawings: Figure 1 This is a flowchart of the electric toothbrush control method of the present invention. Detailed Implementation

[0023] Embodiments of the present invention will now be described with reference to the accompanying drawings. Elements and features described in one drawing or embodiment of the invention may be combined with elements and features shown in one or more other drawings or embodiments. It should be noted that, for clarity, representations and descriptions of components and processes unrelated to the present invention and known to those skilled in the art have been omitted from the drawings and description.

[0024] Example 1 This invention provides an electric toothbrush control system based on oral microecological balance, comprising: The toothbrush head integrates a multimodal sensor, which includes: a gas sensor array using metal oxide semiconductor technology (to detect oral odor markers such as VOCs and hydrogen sulfide), a miniature pH sensor (to monitor oral acid-base balance), a temperature sensor (to monitor changes in oral temperature in real time), a triaxial pressure sensor (to detect brushing force), and an optical sensor (to analyze plaque thickness through reflectance spectroscopy).

[0025] The handle houses the main control PCB board, which includes signal conditioning circuitry, a microprocessor unit, an adaptive motor controller, and a power management module. Data collected by the multimodal sensor is transmitted to the microprocessor unit (e.g., an ARM Cortex-M7 core) via an I2C bus, and it incorporates a specially optimized oral microbiome health analysis algorithm. In this embodiment, the electric toothbrush body adopts an ergonomic curved handle design, and the toothbrush head and handle are magnetically connected for easy replacement and to ensure reliable electrical connection. A wireless charging coil (supporting the Qi standard) and a Bluetooth 5.2 module can also be embedded inside the handle for data synchronization with smartphone applications. To ensure sensor accuracy, the air intake is covered with a waterproof and breathable membrane, and the optical sensor window uses a stain-resistant coating.

[0026] The optical sensor uses a specific wavelength LED light source (450nm blue light and 850nm infrared light) to quantify plaque thickness by analyzing the reflectance spectrum. 450nm blue light is easily absorbed by the plaque biofilm, and the reflected light intensity decreases with increasing plaque thickness; 850nm infrared light has strong penetrating power and is minimally affected by plaque, serving as a substrate compensation reference light; the ratio of the two-band reflected light intensity is used to eliminate interference from tooth color and illumination intensity. Let I... 450 The blue light reflection intensity at 450nm, I 850 Given an infrared reflectance intensity of 850nm; the relative plaque thickness = clinical calibration coefficient × (1-I 450 / I 850 The clinical calibration coefficients were obtained by fitting standard artificial plaque film gradient samples; the larger the relative thickness of the plaque, the more severe the local plaque accumulation, and this value was used as an optical feature input for MHI calculation.

[0027] The electric toothbrush features a multi-function button on the side for power on / off and mode selection. An RGB LED indicator light displays different operating states through color changes. A Bluetooth module enables data synchronization with a mobile app, transmitting sample data and health reports.

[0028] Example 2 This invention provides an electric toothbrush control method based on oral microecological balance, the execution process of which includes the following six steps: Step 1: System Initialization and Sensor Self-Test After the user turns on the electric toothbrush, the system first initializes. The built-in microprocessor unit starts up, and all multimodal sensors perform self-tests. The self-test program verifies the working status of each sensor. If any sensor malfunctions, an alarm is issued to the user via the LED indicator on the device or the connected mobile app. At the same time, the system establishes a connection with the user's smartphone app via Bluetooth to prepare for data synchronization.

[0029] Step 2: Real-time acquisition of multimodal oral environment data After the user starts brushing their teeth, each sensor begins to work at its own specific frequency, collecting raw oral environment data vectors in real time: ; in, , , , , These represent the readings of the gas sensor, pH sensor, temperature sensor, pressure sensor, and optical sensor at time t, respectively. To accurately identify brushing areas (such as the outer side of the upper left molar and the inner side of the lower incisors), the system also uses position sensors such as accelerometers and gyroscopes to establish a relative coordinate system in the oral cavity, tracking the position of the brush head in the oral cavity in real time.

[0030] In use, the readings of each gas sensor are mapped to a range of 0-1 and assigned different health weights. For example: hydrogen sulfide (H2S) concentration: weight w_H2S=0.4 (a major marker of halitosis); total volatile organic compounds (TVOC): weight w_TVOC=0.3 (representing overall microbial metabolic activity); methanethiol (CH3SH): weight w_MeSH=0.2 (related to periodontal disease); ammonia (NH3): weight w_NH3=0.1 (oral decomposition products). The weighted gas index is incorporated as a new time-domain feature into the MHI calculation. Specifically, the comprehensive gas index G = (w_H2S×S_H2S + w_TVOC×S_TVOC + w_MeSH×S_MeSH + w_NH3×S_NH3), where S_ The readings of each sensor are standardized (0-1). G is then incorporated into the time-domain feature vector, participating in the entropy-based weight calculation and the final MHI fusion. Each weight is determined based on the significance of the impact of each biomarker on oral health in clinical literature and is preset during system initialization, while allowing for gradual adjustment through personalized models.

[0031] Step 3: Data Preprocessing and Spatiotemporal Registration Since the response characteristics of each sensor are different, this step synchronizes and optimizes the acquired raw data, mainly including two parts: time registration and nonlinear adaptive filtering.

[0032] (1) Time registration: Due to the differences in sampling frequency and response time among various sensors (e.g., gas sensor: 10Hz, pH sensor: 100Hz, temperature sensor: 50Hz, pressure sensor: 200Hz, optical sensor: 30Hz), a unified time reference must be established. An algorithm based on the cross-correlation function (a general method) is used to calculate and compensate for the time offset caused by the difference in sampling frequency among different sensors, ensuring that all data correspond to the same time point.

[0033] (2) Nonlinear adaptive filtering: for synchronized data Filtering is performed to eliminate mechanical vibration noise and electromagnetic interference from the brush head. The filtering algorithm is as follows: ; This is the filtered environmental data vector at time t. Let t be the original oral environment data vector (multimodal sensor data vector) at time t, and its components are the real-time readings of each sensor. , These are the mean and standard deviation of the data within the sliding window. and To achieve adaptive weighting coefficients, offline grid optimization was performed based on multiple sets of toothbrush vibration noise sample datasets. The dual optimization objectives were maximizing the signal-to-noise ratio (SNR) of the filtered signal and preserving the integrity of the bacterial community feature signals. This process iterated through... ∈[0,1]、 All combinations of ∈[0,1]; when =0.4, When the coefficient is 0.6, the mechanical vibration noise suppression effect is optimal, and the attenuation of oral pH, gas, and optical effective characteristics is less than 5%. Therefore, this set of coefficients is fixed. for Taking the derivative with respect to t, this step ensures the quality of the data upon which subsequent analyses are based. , The calculation formulas are as follows: N=50; .

[0034] i is the time index variable, and R(i) iterates through the original oral environment data vector from time t-N+1 to t. That is, R(i) represents the vector used to calculate statistical features. , A collection of raw oral environment data vectors representing a series of historical moments within a window. , Based on these original oral environment data vector sets, nonlinear adaptive filtering processing is performed on the current R(t).

[0035] The above-mentioned nonlinear adaptive filtering algorithm innovatively combines the nonlinear hyperbolic tangent function with the signal rate of change (derivative term), specifically targeting the suppression of mechanical vibration noise and electromagnetic interference in the working environment of electric toothbrushes. It not only preserves the trend of effective signal changes but also ensures the data quality upon which subsequent analysis is based.

[0036] Step 4: Feature Extraction and Microbial Health Index Calculation From preprocessed data Key features characterizing oral health status were extracted, including time-domain features and frequency-domain features.

[0037] (1) Time domain characteristics: The calculation includes mean, variance, peak value and average gradient, reflecting the stability and fluctuation of the parameters.

[0038] Mean characteristics: , This is the time from when the toothbrush starts to the current execution state, in seconds, and the same applies to the following.

[0039] Variance characteristics: ; Peak characteristics: ; Average gradient: ; (2) Frequency domain features: The dominant frequency component, specific frequency band energy (such as the 3Hz microbial metabolic activity band), and spectral entropy are calculated using short-time Fourier transform to reflect the periodic changes and microbial activity of the oral environment. The frequency domain feature extraction based on Fourier transform (general formula and function) is as follows: ; The main frequency component is: ; Frequency band energy distribution: ; Spectral entropy (characterizing signal complexity): ; in, ; (3) This process is an optional approach: Based on the data from the accelerometer, an oral region recognition algorithm is used to extract brushing position features, and an oral four-quadrant feature mapping function is defined: ; This region-based judgment function, serving as a pre-judgment function in the calculation of the Microbial Health Index (MHI), is primarily used to separate and process data collected from different locations, reducing the system's computational load. Its specific judgment rules employ a general method, based on acceleration thresholds and sign characteristics. This involves statistical analysis of a large amount of brushing action data, as well as oral anatomy, where the mandible is typically lower than the maxilla, leading to differences in accelerometer Z-axis readings. For example, the triaxial acceleration is defined as follows: ax is the lateral acceleration; ay is the anterior-posterior longitudinal acceleration; az is the vertical acceleration; then the region-based judgment threshold rules are as follows: Lower left region: az > Tz, ax < -Tx (left mandibular dentition); Upper left region: az < -Tz, ax < -Tx (left maxillary dentition); In the lower right region: az > Tz, ax > Tx (right mandibular dentition); Upper right region: az < -Tz, ax > Tx (maxillary right dentition); Tx and Tz are pre-calibrated acceleration thresholds, obtained through statistical analysis of a large number of standard brushing motion samples.

[0040] After obtaining the features, feature weighted fusion and MHI calculation are performed: The time-domain features and frequency-domain features are divided into different regions (lower left region, upper left region, lower right region, and upper right region). The dynamic weights of time-domain and frequency-domain features in different regions are determined using the entropy method. The calculation method is as follows: There are M samples (representing multiple brushing sessions), and each sample has K features (features in the time domain and frequency domain). First, the original feature matrix is ​​standardized to eliminate the influence of dimensions.

[0041] ; in It is the j-th feature value of the i-th sample. It is its standardized value.

[0042] Next, the weights are calculated, specifically the weight of the j-th feature value of the i-th sample. .

[0043] ; Then, the feature entropy is calculated, specifically the entropy value of the j-th feature. .

[0044] ; in, It is the normalization factor, ensuring 0 ≤ ≤1, The total number of samples.

[0045] Finally, the weights are calculated based on the entropy value of the j-th feature. .

[0046] ; parameter This is called the difference coefficient. Entropy value. The smaller the value, the greater the information provided by the feature, the larger its difference coefficient, and the higher its weight. The larger it is, the greater K is. K is the total number of features, and k is the index variable in the formula.

[0047] Then, the microecological health index at time t was calculated: ; in, is the baseline health value of the j-th feature derived from clinical data, obtained through large-scale clinical data statistics. K is the total number of features. It is the sensitivity adjustment coefficient of the j-th feature, which controls the response range of each parameter. It is a historical influencing factor, reflecting the continuity of microecological changes. It was the previous moment The gradient is used to capture changing trends. () is an exponential function with the natural logarithm as its base. It is a dimensionless value between 0 and 1, and the closer it is to 1, the healthier the oral microecology.

[0048] Among them, health baseline value Data was obtained through a large-scale clinical trial. Steady-state oral sensor data were collected from N (e.g., 200) subjects who were professionally assessed as having healthy oral health (no periodontal disease, no halitosis, and balanced oral flora) upon waking (before eating, drinking, or brushing teeth). Outliers were removed (using the 3σ criterion), and the statistical median of each sensor reading (pH value, specific gas concentration) in the valid sample data was calculated as the baseline health value for that sensor. For example, if the statistical median of the pH sensor is 7.0, then the healthy baseline value is... =7.0.

[0049] Sensitivity adjustment coefficient The sensitivity of the parameter's contribution to the oral temperature response (MHI) when it deviates from the baseline value is defined. Its value is obtained by analyzing the data distribution of healthy and mildly abnormal states in clinical data, and is typically set as the standard deviation of the sensor data in a healthy population. For example, if the standard deviation of oral temperature in a healthy population is 0.3℃, then... It can be set to 0.3.

[0050] Historical Impact Factor The influence of historical health status on the current assessment is controlled, with a value ranging from (0,1). Its optimal value is determined through offline training: using a clinical dataset containing time series data, the goal is to optimize the accuracy of MHI predictions of clinical assessment results in the future (such as at the next brushing time) using gradient descent. The method was determined through offline training: using a clinical dataset containing time series data, and aiming at the accuracy of MHI in predicting future clinical assessment outcomes, the method was optimized using gradient descent. , Typical values ​​are approximately 0.2-0.4.

[0051] The calculation is as follows: Loss function: Loss = (1 / N) × Σ(MHI_p(t) - MHI_c(t)) 2 Where MHI_p(t) is the MHI calculated according to the formula at the current time, and MHI_c(t) is the clinical assessment value (such as plaque index, gingival index, etc.) of a professional dentist at the same time. Training dataset composition: 100 subjects (aged 20-60 years, covering different oral health conditions) were recruited and required to use this electric toothbrush every morning and evening for 30 consecutive days to record all sensor data. At the same time, a professional dentist conducted an oral clinical assessment every 3 days (recording plaque thickness, pH value, gingival bleeding index, etc.). The total training sample is approximately 100×30×2=6000 brushing sessions, each session containing time-series MHI and related clinical scores. γ optimization interval: γ∈[0,1], initial value set to 0.3. Convergence criterion: Training stops when the loss function decreases by less than 1e-5 within 10 consecutive epochs, or when the maximum number of epochs of 1000 is reached. Optimization algorithm: The Adam optimizer is used, with a learning rate set to 0.01. Validation and testing: Divide the data into training, validation and test sets in a 7:2:1 ratio to prevent overfitting.

[0052] Step 5: Adjusting Dynamic Cleaning Parameters Based on MHI Based on the above calculation process, the control system obtains the MHI value through real-time calculation and then performs real-time evaluation, dividing the dynamic cleaning parameters into three modes, as shown below: The intensive cleaning mode (MHI < 0.3) indicates a higher risk of oral microecological imbalance (such as plaque thickening and signs of inflammation). In this mode, the algorithm maps to a higher vibration frequency (e.g., 350-400Hz), slightly stronger force, and a longer cleaning time, aiming to powerfully remove pathogenic biofilms.

[0053] The balanced cleaning mode (0.3 ≤ MHI < 0.6) indicates that the oral cavity is in a sub-healthy or transitional state. The algorithm uses moderate intensity parameters to effectively clean while protecting tooth enamel and gums.

[0054] The gentle maintenance mode (MHI≥0.6) indicates that the oral cavity is in a healthy and balanced state. The goal is to maintain rather than to cause strong disturbances, so low-frequency vibrations (such as 200-250Hz), gentle force, and standard duration are used.

[0055] Under the constraints of the corresponding mode, the operating parameters of the motor are dynamically adjusted through a preset multi-objective optimization function.

[0056] (1) Parameter mapping: Mapping MHI to specific cleaning instructions: ; in, It is the target vibration frequency. It refers to the amplitude (or force) of the swing. This is the suggested remaining brushing time. (Coefficient) It is a matrix, which was optimized through a large amount of experimental data. It is a bias vector that takes into account the user's personalized preferences. The system ensures that the output parameters are within a safe and comfortable range, such as the vibration frequency. ∈[8000,40000]rpm, brush head pressure ∈[0.5,3.0]N.

[0057] Among them, coefficient and bias vector Offline training was performed using a constrained particle swarm optimization (PSO) algorithm. The fitness function was: fit = first weight coefficient × ΔMHI + second weight coefficient × user subjective comfort score, where ΔMHI is the improvement in MHI after cleaning, and the user subjective comfort score ranges from 0 to 1; the first and second weight coefficients are less than 1, and their sum equals 1. Constraints included: target vibration frequency F: 8000-40000 rpm; brush head pressure P: 0.5-3.0 N; recommended brushing time T: 30-180 seconds; population size: 50-100 (typical value, 80 recommended); iteration termination condition: a maximum of 200 iterations, or a fitness function change of less than 1e-4 over 20 consecutive generations.

[0058] Training data is derived from experiments simulating different oral health states (using artificial oral models or subjects), recording the dynamic changes in oral hygiene index (MHI) under different cleaning parameters (F, P, T). There are two optimization objectives: 1) maximizing the improvement in MHI after cleaning; 2) maximizing the user's subjective comfort score. The PSO algorithm searches within the parameter constraint space, ensuring that the cleaning parameters calculated using the above formula simultaneously and optimally approximate the "gold standard" parameters for both objectives. Finally, it converges to obtain the optimal values ​​of matrix Q and vector B.

[0059] Training dataset source: Using artificial oral cavity models (silicone dental arches, simulated saliva, plaque substitutes), different initial oral cavity conditions (such as low pH, high bacterial concentration) were set, and MHI change curves and simulated user feedback were recorded under different combinations of cleaning parameters.

[0060] Weights Q and bias vector B: The optimal matrix Q (3×3) and vector B (3×1) searched by the PSO algorithm are such that the cleaning parameters calculated according to the formula can maximize the fitness function.

[0061] Motor control: The above parameter mapping algorithm calculates the target cleaning parameters (target frequency) at the current moment based on the real-time MHI. Target strength Then, a PID controller is used to convert the mapped parameters into PWM signals for the motor, precisely controlling the motor's speed and angle. For example, when a large amount of plaque is detected in a certain area (low MHI), the cleaning frequency and intensity of that area will be automatically increased.

[0062] The error input of the PID controller is: , ; in, and The frequency and force at the current moment are obtained from feedback from the motor encoder and pressure sensor.

[0063] ; ; in, , These are the control quantities for frequency and intensity, respectively; , , These are the proportional, integral, and differential parameters of the frequency, respectively.

[0064] The parameters of the above PID controller adopt fuzzy PID control logic: Input / output variables: frequency error and error change rate The input is the PID parameter correction, and the output is the PID parameter adjustment. , , ; Domain partitioning: and The universe of discourse is divided into seven fuzzy subsets {NB, NM, NS, ZO, PS, PM, PB}, and the membership function adopts triangular or Gaussian type. Fuzzy rule base: 49 fuzzy rules (e.g., IF) is NB and is NB THEN isPB), and the rules are derived from expert experience and simulation debugging; Defuzzification: The centroid method is used to convert the fuzzy output into a precise correction value; Parameter update: Actual PID parameters = Initial parameters + Correction amount, i.e. =initial value+ , =initial value+ , =initial value+ .

[0065] Implementation method: The above fuzzy rule table is pre-stored in the microprocessor, and the correction amount is calculated based on the current error in each control cycle.

[0066] , , The default values ​​are 1, 0.03, and 0.1. , , These are the proportional, integral, and derivative parameters of the force, preset to 0.9, 0.04, and 0.2, respectively. These are the factory default values; the equipment can be fine-tuned online according to the user's customized model.

[0067] Furthermore, if the pressure exceeds a safety threshold (3.0N), the motor power is automatically reduced to prevent gum damage. Therefore, in the motor control chain, the pressure sensor signal is introduced as a feedback correction term into the multi-objective optimization function. Specifically, after parameter mapping, a pressure correction step is added: Pressure correction factor: P_c = max(0, 1 - (P_a - 1.5) / 1.5), where P_a is the measured pressure (N), and the threshold of 1.5N is the upper limit of ideal force (3.0N is the safe upper limit, and half of it is taken here as the upper limit of comfort). Corrected parameters: Actual target frequency = target frequency × Pressure correction factor P_c; Actual target strength = min(target strength) (Pressure correction factor P_c×3.0); A pressure feature is added to the MHI calculation. When the pressure exceeds the limit, the system considers the current MHI evaluation to be distorted (because excessive force may damage the micro-ecosystem), and temporarily reduces the MHI of the area by 0.1 as a penalty, prompting the system to reduce the cleaning intensity of the area.

[0068] After standardizing the pressure characteristics, they are incorporated into the time-domain characteristics and used in the entropy method weight calculation, making pressure itself a factor in the MHI assessment.

[0069] Step Six: Finish Brushing and Generate Health Report Once the preset total brushing time (e.g., 2 minutes) is reached or the system determines that all oral areas have been cleaned to the required standard, the toothbrush will automatically stop working or prompt the user to finish.

[0070] (1) Data synchronization and report generation: All MHI data, cleaning parameter adjustment records and brushing time of each area during this brushing process are encrypted and synchronized to the mobile APP via Bluetooth.

[0071] (2) User-personalized model update: The system adopts an incremental learning mechanism to incorporate the data from the current session into the user-personalized model for updates, optimizing weight coefficients and baseline values ​​so that the algorithm can adapt to long-term changes in the user's oral environment. The mobile app will generate a visualized health report, including oral health trends, cleaning effect assessment, and personalized care recommendations.

[0072] The user personalization model (or simply model) is a data structure stored locally on the device and capable of being synchronized with the cloud. It mainly includes: (1) Personalized health baseline vector: Over long-term use by users, the general baseline is gradually updated with the average value of sensor data when the user wakes up in the morning and is in a healthy state; (2) Personalized feature weight vector: Records the long-term average weight calculated by the entropy method, reflecting the degree of differentiation of the user's health status by different sensor indicators.

[0073] (3) Personalized cleaning bias vector: Based on the user's feedback on the preference of cleaning intensity and mode (set through APP), the general bias vector B is finely adjusted.

[0074] (4) Regionalized historical MHI profile: Records the historical MHI average and trend of different oral regions (such as the left upper molar and lower anterior teeth) for regional assessment and reminder.

[0075] The model is updated through incremental learning: after each brushing, new and valid data is used to smoothly update the parameters mentioned above.

[0076] When using an electric toothbrush with the above-described control method, the system performs a sensor self-test after power-on. When the user begins brushing, the sensors collect data at a frequency of 100Hz. After preprocessing, the data is input into the MHI calculation module. The core control algorithm selects the optimal cleaning mode from a predefined strategy library based on the current MHI value and its trend. The motor drive module uses pulse width modulation technology to precisely control the vibration frequency and amplitude. Simultaneously, the system monitors pressure sensor data in real time; if the pressure exceeds a safe threshold (3.0N), it automatically reduces motor power to prevent gum damage. After brushing, a health report for the session is generated, including the MHI curve, cleaning effectiveness score, and care recommendations. The report is synchronized to a mobile app via Bluetooth to update the user's personalized model.

[0077] Example 3: Clinical Validation Trial (1) Research subjects: 120 volunteers were recruited and divided into three groups (40 people in each group): Healthy group (gingival index GI < 1, plaque index PI < 1) Mild periodontal abnormalities group (GI 1-3, PI 1-3, mild periodontitis) Severe bacterial imbalance group (GI>3, PI>3, significant halitosis, thick dental plaque) (2) Experimental method: Each group was randomly divided into two subgroups (20 people each): Experimental group: Using the electric toothbrush of this invention (adaptive to current MHI) Control group: Using commercially available high-end electric toothbrushes (fixed cleaning mode) Brush your teeth twice daily for 2 minutes each time for 4 weeks. Measure the following parameters at baseline, week 2, and week 4: oral pH (pH sensor), hydrogen sulfide concentration (gas sensor), plaque thickness (optical sensor), and MHI value.

[0078] (3) Results (Example Data):

[0079] (4) Conclusion: The electric toothbrush of the present invention is significantly better than the fixed mode toothbrush in improving oral pH balance, reducing hydrogen sulfide concentration, reducing plaque thickness and improving MHI (p<0.05), especially for people with mild abnormalities and severe imbalances, which verifies the superiority of the dynamic adaptive cleaning strategy.

[0080] This invention provides an electric toothbrush control method based on oral data perception and dynamic feedback for oral microecological balance. It dynamically assesses the user's oral microecological balance using a proposed multi-source data fusion algorithm and a Microecological Health Index (MHI) quantification model. The core innovation of this invention lies in using nonlinear adaptive filtering and feature-weighted fusion algorithms to process heterogeneous sensor data and establishing a multi-objective optimization function to map the MHI to optimal cleaning parameters (vibration mode, duration, and intensity). The toothbrush's built-in control system dynamically adjusts the motor output based on the real-time calculated MHI value using a fuzzy PID controller, achieving personalized oral care. This invention overcomes the shortcomings of existing electric toothbrushes in lacking quantitative assessment and dynamic adjustment of the oral microecology, effectively maintaining oral flora balance and health. Compared with existing technologies, it has the following advantages: 1. Comprehensive Quantitative Assessment of Oral Health: Through multimodal sensor fusion and an innovative microecological health index model, a comprehensive quantitative assessment of oral health status is achieved, overcoming the limitations of traditional single-parameter detection. This index comprehensively reflects multidimensional information such as oral flora balance, pH environment, and inflammation status, providing a more scientific basis for care.

[0081] 2. Dynamic Adaptive Cleaning Strategy: Based on a parameter mapping algorithm using nonlinear adaptive filtering and multi-objective optimization, the toothbrush can adapt to changes in the oral microenvironment in real time. This dynamic adjustment mechanism ensures that the optimal cleaning strategy is adopted under different health conditions (such as gingivitis, pH imbalance), effectively cleaning while avoiding over-stimulation.

[0082] 3. Continuous Learning and Personalized Optimization: The algorithm possesses continuous learning capabilities, adapting to changes in users' lifestyles through incremental learning mechanisms, such as changes in the oral environment caused by dietary adjustments or smoking cessation. This personalized adaptation is far superior to static preset programs, and its effectiveness continues to improve with long-term use.

[0083] 4. Balancing practicality and security: The sensor array features a waterproof and stain-resistant design, ensuring long-term reliable operation; all health data is locally encrypted, providing strong privacy protection; wireless charging and modular design enhance the user experience. Compared to dental examinations that require professional equipment, this invention makes routine home monitoring possible, which is of great significance for the prevention of oral diseases.

[0084] It should be emphasized that the term "including / comprises" as used herein refers to the presence of a feature, element, step, or component, but does not exclude the presence or addition of one or more other features, elements, steps, or components.

[0085] Furthermore, the method of the present invention is not limited to being executed in the chronological order described in the specification, but may also be executed in other chronological orders, in parallel, or independently. Therefore, the execution order of the method described in this specification does not constitute a limitation on the technical scope of the present invention.

[0086] Although the invention has been disclosed above through the description of specific embodiments, it should be understood that all the embodiments and examples described above are exemplary and not restrictive. Those skilled in the art can design various modifications, improvements, or equivalents to the invention within the spirit and scope of the appended claims. These modifications, improvements, or equivalents should also be considered to be included within the protection scope of the invention.

Claims

1. A method for controlling an electric toothbrush based on oral microecological balance, characterized in that, include: Real-time acquisition of oral environment data from multimodal sensors; the multimodal sensors include gas sensors, pH sensors, temperature sensors, pressure sensors, and optical sensors; The collected oral environment data were preprocessed by time registration and nonlinear adaptive filtering to obtain preprocessed oral environment data. Temporal and frequency domain features were extracted from the preprocessed oral environment data, and the dynamic weights of each temporal and frequency domain feature were determined using the entropy method. The Microbiome Health Index (MHI) value is calculated based on the dynamic weights, health baseline value, sensitivity adjustment coefficient, and historical influence factors. The health baseline value corresponds to the statistical benchmark value of the healthy population for each sensor, the sensitivity adjustment coefficient corresponds to the dispersion index of the healthy population data for each sensor index, and the historical influence factor is used to fuse the time-series change trend and control the degree of influence of the previous MHI change trend on the current assessment. The cleaning mode is dynamically selected based on the real-time calculated MHI value. The MHI value and its changing trend are mapped to the vibration frequency, oscillation amplitude and recommended brushing time for the appropriate working conditions through a multi-objective optimization function. The motor is then controlled by a PID controller.

2. The electric toothbrush control method according to claim 1, characterized in that, The process of calculating the Microbiome Health Index (MHI) includes: Multiple independent oral cavity zones were divided based on acceleration sensor signals; Based on the partitioning results, the temporal and frequency domain features of each oral cavity partition are integrated at the sample level to form a unified feature sample set containing all partition data; The feature weights are calculated based on the unified feature sample set, and the feature entropy is calculated based on the weights. The difference coefficient is calculated based on the feature entropy, and the dynamic weight of each feature is determined accordingly. Substituting the dynamic weights of each feature, the health baseline value, the sensitivity adjustment coefficient, and the historical influence factor into the exponential mapping formula yields the MHI value.

3. The electric toothbrush control method according to claim 1, characterized in that, The multi-objective optimization function takes the MHI value, MHI rate of change, and MHI integral as inputs, and outputs the target vibration frequency, target oscillation amplitude, and target remaining brushing time through a coefficient matrix and bias vector obtained offline by a particle swarm optimization algorithm.

4. The electric toothbrush control method according to claim 1, characterized in that, The PID controller adopts fuzzy control, which uses the motor operation output deviation and the corresponding deviation change as fuzzy input. It infers the correction amount of the proportional, integral and derivative parameters through multiple preset fuzzy subsets and the matching fuzzy inference rule base, and uses a standard defuzzification algorithm to realize the online adaptive adjustment of PID parameters.

5. The electric toothbrush control method according to claim 1, characterized in that, The dynamic selection of cleaning modes specifically includes: Based on the microecological health index, multiple health zones are divided, and three gradient cleaning conditions are matched respectively: strong cleaning, balanced cleaning, and gentle maintenance. Each condition corresponds to a different vibration frequency, brush head force, and zone cleaning time.

6. The electric toothbrush control method according to claim 1, characterized in that, It also includes a pressure protection step: when the pressure measured by the pressure sensor exceeds the safety threshold, the cleaning output intensity of the corresponding area is reduced, and the microecological health assessment weight of the area is corrected to reduce the cleaning intensity of the area and prevent gum damage.

7. The electric toothbrush control method according to claim 1, characterized in that, Also includes: After the electric toothbrush stops working, a health report for this brushing session is generated and synchronized to the mobile app. At the same time, the user's personalized model is updated using the data from this brushing session. The user-personalized model stores the user's exclusive oral health baseline, feature weights, cleaning preferences, and time-series health records, and updates them smoothly through an incremental learning mechanism.

8. The electric toothbrush control method according to claim 1, characterized in that, The optical sensor uses two differentiated detection bands to complete dual-band reflectance spectral detection and analysis. The relative thickness of dental plaque is calculated by the ratio of the reflection intensity of the two bands, so as to eliminate the interference of tooth color and ambient light.

9. An electric toothbrush control system based on oral microecological balance, characterized in that, include: The toothbrush head integrates multimodal sensors, including a gas sensor for detecting oral odor markers, a pH sensor for monitoring oral pH, a temperature sensor for measuring oral temperature in real time, a pressure sensor for detecting brushing force, and an optical sensor for analyzing plaque thickness through reflectance spectroscopy. The handle contains a microprocessor, a motor controller, and a power management module. The multimodal sensor transmits raw acquired data to the microprocessor via a bus; the microprocessor internally includes a filter calculation module, an oral cavity partitioning calculation module, an entropy health index calculation module, a multi-objective parameter mapping module, a fuzzy PID motor control module, and a personalized model storage and iterative update module. The microprocessor is configured to perform the steps of the electric toothbrush control method as described in any one of claims 1 to 8.

10. The electric toothbrush control system according to claim 9, characterized in that, The brush head and handle are magnetically connected, and the handle also has a wireless charging coil and a Bluetooth module embedded inside.

Images