Feeding and chewing monitoring method and system

Abstract

The invention relates to the technical field of health management, and relates to a feeding chewing monitoring method and system, and the method comprises the steps: obtaining an image of to-be-eaten food; based on the food image, physical attributes of food are determined through an analysis model, preset user personalized information is combined, a recommended chewing target containing recommended chewing times and recommended chewing quality threshold values is generated, and the recommended chewing quality threshold values comprise a myoelectric force threshold value and a breaking degree threshold value; the chewing action of the user is monitored in real time through an acquisition unit worn on the head and face of the user to obtain actual chewing data, and the actual chewing data comprises actual chewing times and actual chewing quality parameters; and comparing the actual chewing data with the recommended chewing target, and outputting instructive feedback information to the user based on a comparison result, the chewing process of the user is monitored through an intelligent means, and personalized chewing guidance is provided, so that the eating efficiency, nutrition absorption and oral health of the user are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of health management, more particularly, it relates to a method and system for monitoring chewing while eating. BACKGROUND

[0002] With the acceleration of social life pace and the improvement of people's health consciousness, the influence of eating habits on human health is increasingly valued. Studies have shown that sufficient chewing is an important part of the digestive process, not only helps to mechanically crush food and mix with saliva, thereby promoting the absorption of nutrients, but also enhances satiety to some extent, which has a positive significance for weight management, oral health and cognitive function. Epidemiological surveys show that when the average eating speed of adults is faster, the risk of indigestion and obesity increases by about 15%-30%; while individuals who eat slowly and chew sufficiently, the gastrointestinal burden is significantly reduced.

[0003] However, the problem of eating too fast and insufficient chewing is prevalent in modern society. For example, it is generally recommended that each mouthful of rice should be chewed at least 20 times, but surveys show that most people only chew an average of 10-12 times, which is significantly lower than the recommended level. Over a long period of time, this may lead to an increased burden on the gastrointestinal digestive function, and even cause obesity and metabolic diseases.

[0004] Therefore, it has certain application value and popularization prospect to develop a technical solution that can effectively monitor chewing behavior and guide users to gradually form good eating habits. For this purpose, technical personnel in the field have proposed various schemes for monitoring chewing behavior through wearable devices,

[0005] For example, the publication number CN108542387A discloses a human chewing signal detection system, which discloses a method of collecting facial masseter muscle signals using electromyography sensors and pulse counting. The actual count value is compared with the preset threshold value, and when the number is insufficient, the user is prompted by an indicator light. Another publication CN111557641A discloses a chewing monitoring method and storage medium for a neck massager, which proposes a method of collecting sound wave signals on both sides of the skull through bone conduction, and identifying the chewing side through energy difference analysis, thereby counting the number of chews and prompting the user to correct the unilateral chewing habit. This type of method can achieve basic monitoring of chewing events, but is prone to misjudgment in complex eating scenarios (such as eating while talking and swallowing).

[0006] To improve the accuracy of identification, some technologies attempt to distinguish similar actions such as chewing and swallowing. For example, the publication number CN106859653A discloses a diet behavior detection device and a diet behavior detection method, which proposes to analyze the stability and energy of the body vibration signal, and combines the signal duration to distinguish: the signal with shorter duration is determined as chewing, and the signal with longer duration is determined as swallowing. The accuracy of this method in the experimental environment can reach about 85%, but in the real dining scene, due to the existence of various interference signals, its accuracy still has room for improvement.

[0007] Considering the influence of individual differences, some technologies introduce personalized calibration and multi-source information fusion. For example, the publication number CN111012354B discloses a method for monitoring eating, a storage medium and an eating monitoring device, which proposes to use specific calibration actions (such as opening the mouth and chewing left and right) before use, so that the system optimizes the judgment algorithm according to the signal change to adapt to the physiological differences of different users. Another publication number CN111709282A discloses a method for representing food oral processing, which discloses a comprehensive method combining surface electromyography and infrared camera technology, which can simultaneously collect muscle activity and mandibular three-dimensional motion trajectory, thereby comprehensively evaluating the food oral processing process. In addition, the publication number CN120129491A discloses a chewing action analysis method, which proposes to use multi-site electromyography and machine learning algorithms to classify chewing actions more carefully, including biting, tongue movement, cheek movement, etc., and its classification accuracy can exceed 90% under experimental conditions.

[0008] In summary, existing technologies have been able to realize counting from basic chewing events, to distinguishing different oral actions, to personalized calibration and detailed analysis combined with multi-source information. However, in practical applications, these technologies are mostly based on pre-set unified standards or user-set targets (such as "20 times per mouthful") for feedback. Such methods do not fully consider the differences in physical properties such as hardness, toughness, and fiber content of different foods. For example, actual test data shows that it takes an average of 30-40 times to chew a piece of cooked beef to achieve sufficient crushing, while for tofu, it may be within 10 times to meet the basic digestive needs. If a single target value is uniformly used, it may lead to insufficient or excessive chewing of some foods, thereby affecting the guidance value of monitoring. SUMMARY

[0009] Therefore, the purpose of the present application is to provide an eating chewing monitoring method and system, which aims to monitor the user's chewing process through intelligent means and provide personalized chewing guidance, thereby improving the user's eating efficiency, nutrient absorption and oral health.

[0010] To achieve the above object, the present application provides the following technical scheme: a method for monitoring chewing of food, characterized in that it comprises the following steps:

[0011] S1, acquiring an image of food to be eaten;

[0012] S2, determining physical properties of the food based on the food image through an analysis model, and combining preset user individual information to generate a recommended chewing target containing recommended chewing times and recommended chewing quality thresholds, wherein the recommended chewing quality thresholds include muscle power intensity thresholds and fragmentation thresholds;

[0013] S3, real-time monitoring of chewing actions of a user through a collection unit worn on the head and face of the user to obtain actual chewing data, wherein the actual chewing data includes actual chewing times and actual chewing quality parameters; and comparing the actual chewing data with the recommended chewing target, and outputting guiding feedback information to the user based on the comparison result.

[0014] The present application is further provided that: the physical properties include hardness, fiber degree, toughness and brittleness calculated from the food image; the user individual information includes age, health goals and historical eating habit data of the user, and a user individual correction coefficient is generated accordingly . The present application is further provided that: the recommended chewing times:

[0015] ;

[0016] wherein is the basic number, H is the hardness, F is the fiber degree, V is the estimated volume of the food, and are preset weight coefficients, is the user individual correction coefficient;

[0017] Muscle power intensity threshold: ;

[0018] wherein is the toughness, and are preset weight coefficients;

[0019] Recommended fragmentation threshold: ;

[0020] wherein is the brittleness, is the preset weight coefficient.

[0021] The present application is further provided that: the step of obtaining actual chewing data through the collection unit comprises: identifying swallowing events through vibration sensor signals in the collection unit, and dividing the eating period accordingly;

[0022] In a feeding cycle, the single chewing events are identified and counted by combining the myoelectric sensor signals and the vibration sensor signals to obtain the actual chewing frequency; and

[0023] For each chewing event, the corresponding myoelectric signal is analyzed to obtain a parameter representing the chewing intensity, and the corresponding vibration signal is analyzed to obtain a parameter representing the food crushing degree, which are collectively used as the actual chewing quality parameter.

[0024] The application is further provided that the step of comparing the actual chewing data with the recommended chewing target includes matching the actual chewing quality parameter of a single chewing with the recommended chewing quality threshold to determine whether it is an effective chewing, and calculating the proportion of effective chewing in a feeding cycle.

[0025] A feeding monitoring system, comprising:

[0026] An image analysis module configured to acquire and analyze food images to determine the physical properties of the food;

[0027] A dynamic target generation module configured to generate a recommended chewing target containing recommended chewing frequency and recommended chewing quality threshold by combining the physical properties and user personalized information;

[0028] A collection unit worn on the head and face of the user, configured to monitor the chewing action of the user in real time,

[0029] A data processing and evaluation module configured to process the received signals to obtain actual chewing data including actual chewing frequency and actual chewing quality parameter, and compare it with the recommended chewing target in real time;

[0030] A feedback module configured to output guiding feedback information to the user according to the comparison result.

[0031] The application is further provided that the collection unit includes a flexible substrate for attaching to the user's masseter region, a flexible printed circuit board placed on the flexible substrate, and an electromyographic sensor, a vibration sensor, an inertial measurement unit connected to the flexible printed circuit board, and a system on chip, which has an analog front end, an analog-to-digital converter and a wireless communication module integrated inside, for pre-processing the sensor signals and wireless transmission.

[0032] The application is further provided that the data processing and evaluation module distinguishes chewing and swallowing events by analyzing the signals of the vibration sensor, and evaluates the intensity of a single chewing and the food crushing degree by fusing the signals of the myoelectric sensor and the vibration sensor.

[0033] The feedback module comprises a visual interactive interface configured to map the effective chewing action of the user into a disintegration animation or a progress bar of the food image to provide intuitive progress feedback.

[0034] The application further provides a calibration module configured to guide the user to complete action collection to establish a reference model for subsequent signal processing and evaluation when the application is used for the first time or specified by the user.

[0035] Compared with the prior art, the application has the following beneficial effects:

[0036] By analyzing the physical properties (hardness, fiber, brittleness, toughness, volume) of the food and the individual differences (age, health goals, historical habits) of the user, the recommended chewing frequency and chewing quality (force, fragmentation) threshold can be generated, which is quantitative and practical, making the guidance more scientific and targeted.

[0037] By combining the electromyography sensor and the vibration sensor, the swallowing event can be accurately identified, the eating cycle can be divided, and the single chewing event can be accurately located and counted. By fusing and analyzing the electromyography signal and the vibration signal, the force of chewing and the fragmentation degree of food can be quantified at the same time, providing a more comprehensive chewing quality evaluation than the previous monitoring of frequency or duration.

[0038] Based on the comparison between the real-time monitoring data and the generated target, the system can accurately determine the effective chewing and calculate the effective chewing ratio. The feedback information is no longer a general "chew more", but a targeted guidance combined with specific problems (such as "insufficient biting force" and "food not chewed"). The visual feedback form (such as food disintegration animation) enhances the user's participation and interest, which helps the user to develop good chewing habits.

[0039] Encouraging users to achieve the recommended chewing frequency and quality targets can promote the thorough grinding and salivary secretion of food, improve the digestibility of food, and enhance the efficiency of nutrient absorption, thereby reducing the burden on the stomach and intestines. BRIEF DESCRIPTION OF DRAWINGS

[0040] Figure 1 The flowchart of the application is shown. DETAILED DESCRIPTION

[0041] REFERENCE Figure 1 The application further describes an eating and chewing monitoring method and system.

[0042] The eating and chewing monitoring method is implemented by the following steps:

[0043] S1, Obtain the image of the food to be eaten: This step aims to obtain the visual information of the food to be eaten by the user, so as to analyze its physical properties subsequently. In terms of specific implementation, a camera fixed on a camera support, a mobile phone camera operated by the user himself, or a micro camera integrated in a smart tableware can be used to shoot the image of the food. The image should be as clear as possible and have high resolution to facilitate subsequent image analysis.

[0044] S2, Determine the physical properties of the food based on the image of the food, and generate a recommended chewing target in combination with the individualized information of the user. This step converts the visual information of the food into quantifiable physical properties, and generates individualized chewing guidance targets in combination with the individual differences of the user. Since the physical properties of the food directly affect the chewing force, the number of chewing times and the chewing efficiency that the user needs to exert, quantifying these properties helps to more accurately formulate the chewing target.

[0045] Determination of food physical properties: By analyzing the texture, color, shape, edge and other visual features in the food image, and using a pre-set analysis model, the hardness, fiber degree, toughness and brittleness of the food are estimated.

[0046] Hardness is determined by analyzing the roughness of the surface texture of the food, the color depth, and whether there are "indentations" or "deformations" in the model. For example, texture feature extraction (such as LBP, GLCM) combined with support vector machine (SVM) or convolutional neural network (CNN) model can be used for recognition and quantification.

[0047] Fiber degree is determined by analyzing whether the internal structure or surface of the food has obvious fibrous structure, such as muscle fibers in meat, veins in vegetables, etc. This can be achieved by edge detection algorithm (such as Canny) combined with morphological processing to identify the density and length of line features.

[0048] Toughness is determined by combining the color, surface glossiness and other information of the food, and considering the typical morphology of different foods. For example, darker color and smoother surface may be associated with higher toughness. Quantification of this property may require more complex texture and color analysis models.

[0049] Brittleness is determined by analyzing whether the surface of the food has visual features that are prone to cracking or chipping, such as whether the edges of the food are sharp or whether the surface has small cracks. High-frequency texture analysis and edge detection can be used to evaluate.

[0050] Estimation of food volume is achieved by estimating the approximate volume of the food based on the image size of the food and the distance information of the user (if available), or by the user's preliminary segmentation (such as the user marking the region of interest).

[0051] Integration of user personalization information and generation of user personalization correction coefficient (C ): The user's age, health goals (such as weight loss, muscle gain, improved digestion), and historical eating habits data (such as average chewing times, eating duration) are key factors in determining the user's optimal chewing strategy. By analyzing this information, a user-specific correction coefficient is generated to adjust the basic target calculated based on the physical properties of the food.

[0052] As age increases, chewing ability may decline, requiring more gentle meals and appropriate chewing guidance. For example, a lower is set for the elderly.

[0053] If the user's health goal is to promote digestion, the number of chews and the quality of chewing may need to be increased, and may be appropriately increased. If the goal is to eat quickly (such as during specific training), adjustments may be needed.

[0054] Analyze the user's past chewing frequency, chewing times per mouthful, chewing duration, etc. For example, if the user is used to swallowing quickly, a lower is needed to gradually guide them to change their habits.

[0055] The calculation of can use weighted average, decision tree, or rule-based system to calculate, which can map from age, health goals, and historical habits to the function of the corrected coefficient.

[0056] Since each user's chewing ability and physiological needs are different. The personalized correction coefficient can ensure that the recommended target not only meets the characteristics of the food, but also fits the user's actual situation, thereby improving the effectiveness and safety of the guidance.

[0057] Generate recommended chewing target:

[0058] Recommended chewing times:

[0059] is the basic number, representing the most basic chewing rhythm, a constant;

[0060] and are preset weight coefficients, used to adjust the degree of influence of hardness and fiber on recommended chewing times, which can be optimized through experimental data.

[0061] H is the hardness of the food, the higher the hardness of the food, the more chewing times are usually needed to fully break down.

[0062] F is the degree of food fiber, foods with high fiber content (such as some vegetables, meat) require more meticulous grinding, so increase the number of chewing.

[0063] V is the estimated volume of food, the larger the volume, the more times you need to chew. By slow down the exponential growth brought by too large volume, avoid the target too large.

[0064] User individual correction coefficient, used to adjust the recommended number of times according to individual differences of users.

[0065] The formula takes into account the degree of difficulty of chewing food itself (H, F, V) and the individual condition of the user , through the weighted superposition and multiplication correction of the basic number of times, a relatively reasonable recommended chewing number is obtained, and the use of logarithmic function is to make the influence of volume more gentle.

[0066] Recommended muscle strength threshold:

[0067] And is the preset weight coefficient, used to adjust the influence degree of hardness and toughness on the recommended muscle strength threshold.

[0068] R is the toughness of food, the tougher the food, the more sustained and greater biting force is needed to effectively tear and grind.

[0069] The strength of the biting muscle is the key to breaking food. This formula sets a minimum muscle signal intensity threshold by combining the hardness and toughness of food, to ensure that the user applies enough force to effectively handle food with hard or tough texture.

[0070] Recommended fragmentation threshold: ;

[0071] λ is the preset weight coefficient, used to adjust the influence degree of brittleness on the recommended fragmentation threshold;

[0072] C is the brittleness of food, the more brittle the food, the easier it is to be quickly broken into small particles during chewing.

[0073] Fragmentation reflects the degree of food decomposition during chewing, which is an important indicator of chewing efficiency. Brittleness is the main physical property that causes food to break, so a threshold related to brittleness is set to encourage users to effectively break food into smaller particles through sufficient chewing, which is beneficial to digestion.

[0074] S3, real-time monitoring of chewing action, comparing data and outputting guiding feedback, realizing dynamic monitoring and intelligent feedback of user chewing process.

[0075] Real-time monitoring of user chewing action to obtain actual chewing data is through a collection unit worn on the user's head and face, which real-time collects the user's electromyography signal and vibration signal, and processes them to obtain actual chewing frequency and actual chewing quality parameters.

[0076] The vibration sensor in the collection unit (usually placed near the mandible) is most sensitive to the vibration of a specific frequency and amplitude generated during swallowing action. By analyzing the pattern of the vibration signal, swallowing events can be accurately identified. The time period between one or more swallowing events is a meal cycle, which is used to analyze all chewing actions contained therein.

[0077] In each meal cycle, chewing action produces a different, more continuous and regular vibration pattern from swallowing, accompanied by electromyography signals generated by the contraction of the mandibular muscle group (such as the masseter muscle). By fusing the activity pattern of the electromyography signal (such as voltage change, frequency characteristics) and the signal of the vibration sensor (such as vibration amplitude and frequency), a single chewing event can be distinguished. Each time a chewing event is identified, the counter is incremented, and the actual chewing frequency is accumulated.

[0078] For each identified chewing event, the corresponding electromyography signal is analyzed. The amplitude and duration of the electromyography signal are important indicators of the size of the bite force. For example, the peak voltage of the electromyography signal, integrated electromyography, etc. can be calculated. The present invention defines these quantified electromyography signal parameters as parameters representing the chewing force.

[0079] For the same chewing event, the corresponding vibration signal is analyzed. During the process of breaking down food from blocks to small particles, vibrations of different frequencies and amplitudes are generated. For example, the higher the degree of fragmentation, the more high-frequency components or stronger scattering effects the expected vibration signal may contain. This can be quantified by analyzing the spectral characteristics (such as power spectral density) or entropy of the vibration signal. The present invention defines these quantified vibration signal parameters as parameters representing the degree of food fragmentation.

[0080] Swallowing action is the end point of eating, and it can be more clearly analyzed to analyze a complete chewing process. Electromyography signals directly reflect the activity intensity of the masseter muscle, while vibration signals contain physical information generated when food is broken down. By fusing these two signals, the user's chewing quality can be more comprehensively and accurately evaluated.

[0081] The actual chewing data (frequency and quality parameters) collected are compared with the preset recommended chewing targets (frequency, electromyography force threshold, fragmentation threshold) to determine the effectiveness of chewing, and feedback information is generated based on this.

[0082] For each chewing event, the actual chewing quality parameters (force and fragmentation) obtained are compared with the recommended chewing quality thresholds ( and ) are compared. If the force parameter of a single chew is greater than or equal to , and the fragmentation parameter (which can be further refined into multiple dimensions, here taken as a single parameter) is greater than or equal to , then the chew is determined to be an "effective chew".

[0083] In a meal period, the number of effective chews is counted and divided by the total number of chews (actual chews) in the period to obtain the effective chew ratio.

[0084] The actual number of chews of the user is compared with the recommended number of chews . If the actual number is low, the user can be prompted to "chew a few more times".

[0085] If the force parameter of multiple chew events is identified to be lower than , the user can be prompted to "bite harder".

[0086] If the fragmentation parameter of multiple chew events is identified to be lower than , the user can be prompted to "chew the food more finely".

[0087] If the effective chew ratio in a meal period is lower than a preset threshold, the user can be prompted that "the chewing quality of this meal needs to be improved".

[0088] The above feedback information can be integrated, for example, when the user finishes chewing a meal, an overall evaluation can be given, such as "the number of chews in this meal meets the recommendation, but the effective chew ratio is low, it is suggested to pay attention to increasing the biting force and finely chewing the food next time".

[0089] The standard of "effective chew" is clearly defined, which can quantify the actual effect of chewing. By quantitatively comparing the user's performance with the recommended target, precise and personalized feedback can be achieved to guide the user to gradually improve the chewing habit.

[0090] Estimation model of physical properties: a simplified linear model is used to estimate the hardness (H) and the fiber degree (F), the input of which is the extracted image features:

[0091] Hardness ;

[0092] Texture roughness feature value, extracted by image texture analysis algorithm (such as ASM or Contrast feature of GLCM), the larger the value, the rougher the texture;

[0093] For the color depth feature value, it can be represented by the brightness of the average pixel value, and the larger the value, the darker the color;

[0094] And For the preset weight coefficient, it can be =0.5, =0.3; For the bias term, it can be =2, the rougher and darker the food, the higher the hardness.

[0095] Fiber degree ;

[0096] For the fiber density feature value, it is represented by the number of edge points in the unit area after edge detection; For the preset weight coefficient, it can be =0.7; For the bias term, it can be =1; the higher the fiber density of the food, the higher the fiber degree.

[0097] Other physical properties (R, C) can also be similarly established to estimate the model based on image features.

[0098] Calculation of user personalized correction coefficient ( ):

[0099] If the user's health goal is "muscle gain", the goal is mapped to . The user's age is 30 years old, which is mapped to =1.0. The user's historical eating habit data shows that he chews an average of 15 times per mouthful, which is a high value, and we want to encourage him to maintain it, which is mapped to =1.1, and the weight is set to , ,

[0100] Then =(0.2*1.0+0.4*1.2+0.4*1.1)=1.12.

[0101] Detailed process of comparing actual chewing data with recommended chewing target:

[0102] The user is eating an apple, and the image analysis module analyzes the apple image to estimate the hardness H=7, the fiber degree F=2, the brittleness C=8, and the volume V=50.

[0103] The user (middle-aged, health goal "improve digestion", historical habit about 10 times per mouthful) has a of 1.05.

[0104] Generate recommendation target:

[0105] Wherein, =5, =1, =1. =(14)*ln(51)*1.05≈14*3.93*1.05≈57.99≈58 times.

[0106] Recommended muscle power threshold ;

[0107] Wherein =5, =3, R=3, =44 (unit: µV, muscle signal amplitude)

[0108] Recommended fragmentation threshold: ;

[0109] λ=2, =16 (unit: vibration signal characteristic value)

[0110] Recommendation target: recommended chewing times ≈58 times, muscle power threshold =44 µV, fragmentation threshold =16.

[0111] In the process of real-time monitoring and data collection, the user starts to chew,

[0112] Feeding cycle 1: swallow 1 time, including 12 times of chewing.

[0113] Chewing 1: muscle signal peak value 50 µV (> 44 µV), vibration signal fragmentation characteristic value 18 (> 16). Effective chewing.

[0114] Chewing 2: muscle signal peak value 30 µV (< 44 µV), vibration signal fragmentation characteristic value 12 (< 16), ineffective chewing.

[0115] ... Analyze 12 times of chewing one by one, after the analysis is completed, 7 times of effective chewing in 12 times of chewing.

[0116] Comparison and feedback: the actual chewing times 12 times are less than the recommended 58 times; the effective chewing ratio 7 / 12 ≈ 58.3%.

[0117] Feedback: "You chewed the food in this mouth less, please try to chew a few more times,"

[0118] "The effective chewing ratio of this meal is not optimal, please pay attention to increase the biting force and fully break the food next time."

[0119] A meal monitoring system, including the following modules, which work together to achieve intelligent monitoring and guidance of the chewing process of a meal.

[0120] Image analysis module: receives and processes the food image input by the user, identifies visual features in the food through built-in computer vision algorithms (such as texture analysis, color analysis, edge detection, etc.), and drives the estimation model to output the physical properties of the food, such as hardness (H), fiber (F), crispness (C), toughness (R), and estimated food volume (V).

[0121] It can be integrated into a user terminal (such as a mobile phone app) or a separate image acquisition device. Its core is a trained image feature extractor and a physical property estimation model based on these features.

[0122] Dynamic target generation module: receives the physical properties of the food output by the image analysis module, and reads the user's personalized information (age, health goals, historical eating habits data) from the user's configuration. Using a pre-set mathematical model and user-specific correction coefficients (A, B, C, D, E, F, G, H, I, J, K, L, M, N, O, P, Q, R, S, T, U, V, W, X, Y, Z), calculate the recommended number of chews (N), the recommended muscle power threshold (P), and the recommended fragmentation threshold (F) for the food and the user.

[0123] The acquisition unit is worn on the user's face to collect physiological signals in real time and non-invasively during the user's meal.

[0124] The acquisition unit includes:

[0125] The flexible base is made of medical-grade silicone or TPU and other materials, has good biocompatibility and adhesion, and is used to comfortably adhere to the user's masseter muscle area.

[0126] Flexible printed circuit board (FPC), attached to the flexible base, as a carrier for connection and data transmission, allowing bending to adapt to facial contours, and integrating miniature electronic components.

[0127] The electromyography sensor is usually a dry electrode or a microneedle electrode, which directly contacts the skin and can capture weak electrical signals generated by the masseter muscle contraction.

[0128] The vibration sensor is usually a MEMS vibration sensor, installed in a position attached to the bone (such as the mandible), used to capture mechanical vibrations generated during chewing and / or swallowing.

[0129] ​​​​Inertial Measurement Unit (IMU) includes accelerometer and gyroscope, which is used to detect head posture and movement, helping to distinguish head shaking in chewing action or assisting in judging swallowing.

[0130] System on Chip (SoC) is integrated on FPC, which is the core of the acquisition unit, including:

[0131] Analog Front End (AFE) is used to amplify, filter and impedance match the raw analog signals from the sensors.

[0132] Analog-to-Digital Converter (ADC) is used to convert the analog signals processed by AFE into digital signals for subsequent processing.

[0133] Wireless Communication Module is responsible for wireless transmission of the collected and pre-processed digital signals to the Data Processing and Evaluation Module. The microprocessor in SoC is also responsible for performing basic signal preprocessing, such as noise reduction and DC bias removal, to reduce the transmission burden and post-processing pressure.

[0134] Calibration Module is used to establish individual baseline signal model when the system is first used or when the user needs it. It guides the user to complete a series of standard actions, such as forceful bite, relaxation, mouth opening and closing, etc., and collects the electromyography and vibration signals at that time. These baseline signals are used in subsequent processing, for example, to more accurately set the "zero point" baseline of the signal, or to fine-tune the analysis standards of subsequent data according to individual differences of the user.

[0135] Data Processing and Evaluation Module: receives real-time sensor data from the acquisition unit and performs in-depth analysis. Analyze the actual chewing data (actual chewing frequency and two actual chewing quality parameters), then compare it with the recommended chewing target output by the Dynamic Target Generation Module in real time, judge the effectiveness of chewing, and calculate related indicators.

[0136] Use signal processing algorithms (such as time domain analysis, frequency domain analysis, wavelet analysis, machine learning classifier) to distinguish the signals of vibration sensors to distinguish chewing and swallowing events.

[0137] Combine electromyography sensor signals (such as signal amplitude, integrated electromyography iEMG) and vibration sensor signals (such as signal frequency components, energy or entropy) to accurately assess the force and food crushing degree of a single chewing.

[0138] Data comparison and evaluation: compare the actual chewing frequency parsed with Compare the force and crushing degree parameters of a single chewing with and Compare, and get the determination result of effective chewing, and calculate the proportion of effective chewing.

[0139] Feedback module: generate and provide intuitive and easy-to-understand guidance feedback information to the user according to the comparison results of the data processing and evaluation module.

[0140] The visual interactive interface can be integrated on the display screen of the user's smartphone App, smart watch or integrated smart cutlery.

[0141] Map the user's effective chewing action to the disintegration animation of the food image. As the number of effective chews increases, the food image gradually breaks into smaller particles. Alternatively, use a progress bar to visually display the completion of the current eating process.

[0142] When the number of chews, the force, or the degree of fragmentation is detected to be insufficient, an instant reminder is given through text, voice, or a simple icon (such as a "bite" icon to indicate insufficient force).

[0143] After each meal, provide a comprehensive chewing analysis report for this meal, including total chewing times, effective chewing times, effective chewing ratio, etc., and give improvement suggestions.

[0144] The above only describes the preferred embodiments of the present application and is not intended to limit the present application. Any changes and substitutions within the scope of the technical solutions of the present application should be included in the protection scope of the present application.

Claims

1. A method of monitoring eating and chewing, characterized by, The method comprises the following steps: S1, obtaining an image of food to be eaten; S2, based on the food image, determining the physical properties of the food through an analysis model, and combining pre-set user individual information to generate a recommended chewing target containing recommended chewing times and recommended chewing quality thresholds, wherein the recommended chewing quality thresholds include muscle power thresholds and fragmentation thresholds; S3, through a collection unit worn on the head and face of the user, real-time monitoring of the user's chewing action is performed to obtain actual chewing data, including actual chewing times and actual chewing quality parameters; and The actual chewing data is compared with the recommended chewing target, and guiding feedback information is output to the user based on the comparison result.

2. The method of claim 1, wherein, The physical properties include hardness, fiber, toughness, and brittleness calculated from the food image; The user personalization information includes the user's age, health goals, historical eating habit data, and a user personalization correction factor is generated therefrom .

3. The method of claim 1, wherein the method further comprises: The recommended chewing times: ； wherein is a base number, H is a hardness, F is a fiber, V is an estimated food volume, and is a preset weight coefficient, is a user individualization correction coefficient; Myoelectric force threshold: ; wherein is ductile, and is a preset weight coefficient; recommended crush threshold: ; wherein is brittle, is a preset weight coefficient.

4. The method of claim 1, wherein the method further comprises: The step of obtaining actual chewing data through the collection unit includes: Through the vibration sensor signal in the collection unit, swallowing events are identified, and the eating period is divided accordingly; In the eating period, combined with the electromyography sensor signal and the vibration sensor signal, single chewing events are identified and counted to obtain the actual chewing times; and For each chewing event, the corresponding electromyography signal is analyzed to obtain a parameter representing the chewing force, and the corresponding vibration signal is analyzed to obtain a parameter representing the food fragmentation, which are collectively used as the actual chewing quality parameters.

5. The method of claim 1, wherein, The step of comparing the actual chewing data with the recommended chewing target includes matching the actual chewing quality parameters of single chewing with the recommended chewing quality thresholds to determine whether it is effective chewing, and calculating the effective chewing ratio in an eating period.

6. A food intake monitoring system according to any one of claims 1-5, characterized in that It comprises: An image analysis module configured to obtain and analyze the food image to determine the physical properties of the food; A dynamic target generation module configured to generate a recommended chewing target containing recommended chewing times and recommended chewing quality thresholds in combination with the physical properties and user individual information; A collection unit worn on the head and face of the user configured to real-time monitor the user's chewing action, A data processing and evaluation module configured to process the received signals to obtain actual chewing data including actual chewing times and actual chewing quality parameters, and to compare them with the recommended chewing target in real time; A feedback module configured to output guiding feedback information to the user according to the comparison result.

7. A food intake monitoring system according to claim 6, wherein The collection unit includes a flexible substrate for attaching to the user's masseter region, a flexible printed circuit board on the flexible substrate, and an electromyography sensor, a vibration sensor, an inertial measurement unit connected to the flexible printed circuit board, and a system on chip, which has an analog front end, an analog-to-digital converter and a wireless communication module integrated inside, for pre-processing the sensor signals and wireless transmission.

8. A food intake monitoring system according to claim 6, wherein The data processing and evaluation module distinguishes chewing and swallowing events by analyzing the signals of the vibration sensor, and evaluates the force of single chewing and the fragmentation of food by fusing the signals of the electromyography sensor and the vibration sensor.

9. A food intake monitoring system according to claim 6, wherein The feedback module includes a visual interactive interface configured to map the user's effective chewing action into a disassembled animation or a progress bar of the food image to provide intuitive progress feedback.

10. The eating monitoring system of claim 6, wherein, A calibration module is also included, which is configured to guide the user to complete action collection to establish a baseline model for subsequent signal processing and evaluation upon initial use or user designation.

Images