Breast feeding intelligent guidance and milk component analysis system

Abstract

The invention relates to the technical field of maternal and infant health, in particular to a breast feeding intelligent guidance and milk component analysis system, which comprises a milk component analysis module for generating a milk component analysis result by collecting spectral data of a breast milk sample, the lactation posture evaluation module and the computer vision module are used for evaluating lactation posture quality and collecting milk appearance and infant sucking behavior image data respectively, the data processing module is used for integrating the data and generating a comprehensive analysis result, and the feeding scheme generation module is used for creating a personalized feeding scheme based on the result and providing a personalized feeding scheme for the infant. The remote guidance module transmits an analysis result to a professional, obtains guidance suggestions and integrates the guidance suggestions into a personalized feeding scheme, and the user interface module displays the scheme and the suggestions and receives feedback, so that rapid detection of milk components and objective evaluation of lactation postures are realized, and family daily detection and accurate guidance become possible.

Description

Technical Field

[0001] The present invention relates to the field of maternal and infant health technology, and in particular to an intelligent breastfeeding guidance and milk composition analysis system, which is applied to milk composition detection, nursing posture assessment and personalized feeding plan generation during breastfeeding. Background Art

[0002] Breastfeeding, as the most ideal form of nutrition for infants and young children, plays an irreplaceable and important role in their healthy growth. However, in practice, breastfeeding mothers generally face many challenges, such as an inability to intuitively understand milk composition, difficulty determining correct breastfeeding posture, and a lack of scientific feeding guidance.

[0003] Existing technologies primarily rely on laboratory testing equipment, which requires specialized personnel, is costly and time-consuming, making it difficult to meet daily needs. Breastfeeding posture assessment relies primarily on the experience of lactation consultants, lacking objective, quantitative standards and data support. Existing breastfeeding guidance often remains at the general recommendation stage, failing to provide personalized solutions tailored to individual differences.

[0004] Furthermore, existing technologies are fragmented, lacking a systematic solution that integrates milk composition analysis, breastfeeding posture assessment, infant sucking behavior analysis, and personalized feeding plan generation. This makes breastfeeding effectiveness evaluation difficult and leads to insufficient scientific guidance. In particular, existing technologies are unable to establish a correlation between milk composition and breastfeeding posture quality, providing data-driven guidance for scientific feeding.

[0005] Therefore, there is an urgent need to develop an intelligent system that integrates milk composition analysis, breastfeeding posture assessment, and personalized feeding plan generation to provide scientific and accurate data support and guidance for breastfeeding. Summary of the Invention

[0006] The purpose of the present invention is to overcome the shortcomings of the existing technology and provide an intelligent breastfeeding guidance and milk composition analysis system to achieve convenient detection of milk composition, objective evaluation of breastfeeding posture and automatic generation of personalized feeding plans.

[0007] The present invention proposes a breastfeeding intelligent guidance and milk composition analysis system, comprising:

[0008] Milk composition analysis module, used to collect spectral data of breast milk samples, extract spectral feature information, and generate milk composition analysis results;

[0009] a breastfeeding posture assessment module, communicatively connected to the milk composition analysis module, for collecting breastfeeding posture-related parameters and generating breastfeeding posture quality assessment data;

[0010] a computer vision module, in communication with the milk composition analysis module and the breastfeeding posture assessment module, for collecting image data of milk appearance characteristics and infant sucking behavior to generate auxiliary analysis data;

[0011] a data processing module, communicatively connected to the milk composition analysis module, the breastfeeding posture assessment module, and the computer vision module, configured to receive and process the milk composition analysis results, the breastfeeding posture quality assessment data, and the auxiliary analysis data, and generate a comprehensive analysis result;

[0012] a feeding plan generating module, communicatively connected to the data processing module, for generating a personalized feeding plan based on the comprehensive analysis results;

[0013] a remote guidance module, communicatively connected to the data processing module and the feeding plan generating module, configured to transmit the milk composition analysis results and the comprehensive analysis results to a remote professional terminal, receive professional guidance suggestions, and integrate the professional guidance suggestions into the personalized feeding plan;

[0014] A user interface module is in communication with the feeding plan generation module and the remote guidance module, and is used to present the personalized feeding plan and the professional guidance suggestions to the user and receive user feedback information.

[0015] Preferably, the milk composition analysis module includes:

[0016] A transmission spectrum acquisition unit, used to collect spectral data of breast milk samples through a dynamic thickness control mechanism consisting of a movable glass sample holder and a fixed sample chamber;

[0017] a multi-level spectral feature extraction unit, communicatively connected to the transmission spectrum acquisition unit, for performing baseline correction on the spectral data, identifying characteristic peak positions, intensities and areas, and extracting multi-dimensional feature vectors;

[0018] a multivariate correction and component prediction unit, communicatively connected to the multi-level spectral feature extraction unit, for generating milk component content data through a hierarchical prediction model based on the multi-dimensional feature vector;

[0019] an indirect component derivation unit, communicatively connected to the multivariate correction and component prediction unit, for deducing and calculating the content of milk components that are difficult to measure directly based on the milk component content data;

[0020] The dynamic closed-loop feedback unit is in communication with the indirect component derivation unit and is used to record the temporal changes in the milk component content, analyze the correlation between the milk component and the quality of the breastfeeding posture, and generate milk component optimization suggestions.

[0021] Preferably, the transmission spectrum acquisition unit comprises:

[0022] A light source array, configured to emit light covering near-infrared and mid-infrared bands;

[0023] a sample chamber for holding a breast milk sample;

[0024] A movable glass sample holder, which cooperates with the sample chamber to form a dynamic thickness control mechanism, and adjusts the sample thickness by moving up and down;

[0025] Folded optical path system to increase the interaction path between light and sample;

[0026] Temperature control system to maintain the sample at a constant temperature of 37°C;

[0027] The photodetector array is used to receive the light signal transmitted through the sample and generate spectral data.

[0028] Preferably, the multivariate correction and component prediction unit includes:

[0029] The first-level linear prediction subunit is used to make preliminary content estimates of the main milk components;

[0030] The second-level nonlinear prediction subunit is used to process the nonlinear relationship between complex components and optimize the content prediction results;

[0031] The individualized adjustment subunit is used to perform personalized adjustments to the prediction model parameters based on user historical data;

[0032] a result fusion subunit, configured to integrate the prediction results of the first-level linear prediction subunit and the second-level nonlinear prediction subunit to generate final milk component content data;

[0033] The uncertainty quantification subunit is used to generate a confidence interval for each prediction result to indicate the accuracy of the prediction.

[0034] Preferably, the breastfeeding posture assessment module includes:

[0035] A sensor array unit, including a 9-axis motion sensor, a pressure sensor, and a distance sensor, for collecting parameters related to breastfeeding posture;

[0036] a posture type identification unit, communicatively connected to the sensor array unit, for identifying a current breastfeeding posture type based on the breastfeeding posture-related parameters;

[0037] a posture quality assessment unit, communicatively connected to the sensor array unit and the posture type recognition unit, for quantitatively scoring the execution quality of the breastfeeding posture type and generating posture quality score data;

[0038] a contact area calculation unit, communicatively connected to the sensor array unit, for calculating the contact area and pressure distribution between the breast and the baby, and generating contact characteristic data;

[0039] The posture stability evaluation unit is communicatively connected to the sensor array unit and is used to analyze the frequency and amplitude of posture changes during breastfeeding and generate stability evaluation data.

[0040] Preferably, the computer vision module comprises:

[0041] An image acquisition unit, used to collect image data of milk appearance characteristics and infant sucking behavior;

[0042] a milk appearance analysis unit, communicatively connected to the image acquisition unit, for analyzing the color, turbidity, and uniformity of the milk and generating milk appearance characteristic data;

[0043] a sucking behavior recognition unit, communicatively connected to the image acquisition unit, for locating the infant's mouth and neck using a Hough transform algorithm, identifying the sucking position and sucking pattern, and generating sucking behavior characteristic data;

[0044] a swallowing detection unit, communicatively connected to the sucking behavior recognition unit, for recognizing and counting swallowing actions to generate swallowing frequency data;

[0045] The milk flow analysis unit is in communication with the image acquisition unit and is used to analyze the milk flow state through visual tracking technology, estimate the milk flow rate and flow rate, and generate flow characteristic data.

[0046] Preferably, the data processing module includes:

[0047] a data synchronization unit, configured to receive and synchronize the milk composition analysis result, the breastfeeding posture quality assessment data, and the auxiliary analysis data;

[0048] a component-posture correlation analysis unit, communicatively connected to the data synchronization unit, for analyzing the correlation between milk components and breastfeeding posture quality, and establishing a component-posture-absorption efficiency correlation model;

[0049] A time series trend analysis unit, in communication with the data synchronization unit, for analyzing the changing trend of milk components over time and predicting the direction of future changes;

[0050] an abnormality detection unit, in communication with the data synchronization unit, for identifying milk components that deviate from a normal range and generating an abnormality marker and a risk level assessment;

[0051] The cloud data processing unit is in communication with the data synchronization unit, the component-posture association analysis unit, the time series trend analysis unit and the anomaly detection unit, and is used to summarize the analysis results and compare them with the group data to generate a comprehensive analysis result.

[0052] Preferably, the feeding plan generating module comprises:

[0053] a target setting unit, configured to receive and store a milk composition optimization target set by a user;

[0054] a multi-objective optimization unit, communicatively connected to the target setting unit, for performing a multi-objective balance calculation based on the comprehensive analysis result and the milk composition optimization target, and generating an optimization strategy;

[0055] a posture improvement suggestion unit, communicatively connected to the multi-objective optimization unit, for generating a breastfeeding posture improvement suggestion based on the optimization strategy;

[0056] a feeding frequency suggestion unit, in communication with the multi-objective optimization unit, for generating feeding frequency and time schedule suggestions based on the optimization strategy;

[0057] a sucking behavior guidance unit, communicatively connected to the multi-objective optimization unit, for generating guidance suggestions for the infant's sucking behavior based on the optimization strategy;

[0058] The program integration unit is in communication with the posture improvement suggestion unit, the feeding frequency suggestion unit and the sucking behavior guidance unit, and is used to integrate various suggestions to form a complete personalized feeding program.

[0059] Preferably, the remote guidance module includes:

[0060] A data encryption transmission unit, used for encrypting the milk component analysis results and the comprehensive analysis results, and securely transmitting them to a remote professional terminal;

[0061] a lactation consultant communication unit, communicatively connected to the data encryption transmission unit, for establishing a communication connection with the lactation consultant to receive breastfeeding posture correction suggestions;

[0062] a physician diagnosis communication unit, communicatively connected to the data encryption transmission unit, for establishing a communication connection with the physician to receive a diagnosis and advice on abnormal milk composition;

[0063] The professional advice integration unit is communicatively connected to the lactation consultant communication unit and the physician diagnosis communication unit, and is used to integrate the advice provided by professionals to form professional guidance advice.

[0064] Preferably, the user interface module includes:

[0065] A touch screen display unit, configured to present the personalized feeding plan and the professional guidance suggestions in the form of a graphical interface;

[0066] a data visualization unit, communicatively connected to the touch screen display unit, for converting milk composition data, breastfeeding posture quality data, and time series change data into intuitive charts and trend lines;

[0067] Voice announcement unit, used to provide real-time guidance and suggestions through voice;

[0068] a posture demonstration unit, communicatively connected to the touch screen unit, for demonstrating the correct breastfeeding posture through three-dimensional animation;

[0069] A user feedback collection unit, connected to the touch screen display unit for collecting user operation feedback and execution effect feedback;

[0070] A feedback transmission unit is communicatively connected to the user feedback collection unit and is used to transmit user feedback information to the data processing module and the feeding plan generation module for plan optimization.

[0071] The beneficial effects of the present invention include:

[0072] 1. It enables convenient detection of breast milk components, requiring only a small amount of sample (0.2-0.5 ml) to complete the analysis, shortening the detection time to less than 1 minute, making routine home testing possible;

[0073] 2. Through multi-sensor fusion technology, objective quantitative assessment of breastfeeding posture quality is achieved, transforming traditional posture guidance that relies on subjective experience into precise guidance based on data;

[0074] 3. A correlation model between milk composition and breastfeeding posture quality was established, confirming the scientific principle that "higher breastfeeding posture leads to higher nutrient absorption efficiency," providing data support for optimizing feeding plans.

[0075] 4. A closed-loop feedback mechanism of testing-guidance-verification-optimization has been implemented, enabling continuous optimization of feeding plans. After an average of three rounds of optimization, the absorption efficiency of key nutrients can be increased by 40-60%;

[0076] 5. Through the remote guidance module, remote participation of professionals is achieved, providing timely and professional guidance and suggestions for abnormal situations, thereby improving the safety and effectiveness of breastfeeding. BRIEF DESCRIPTION OF THE DRAWINGS

[0077] Figure 1 This is a diagram showing the overall architecture of the intelligent breastfeeding guidance and milk composition analysis system of the present invention;

[0078] Figure 2 This is a schematic structural diagram of the milk component analysis module of the present invention;

[0079] Figure 3 Schematic diagram of the structure of the breastfeeding posture assessment module of the present invention;

[0080] Figure 4 Schematic diagram of the structure of the computer vision module of the present invention;

[0081] Figure 5 Schematic diagram of the structure of the data processing module of the present invention;

[0082] Figure 6 A schematic diagram of the structure of a feeding plan generation module of the present invention;

[0083] Figure 7 Schematic diagram of the structure of the remote guidance module of the present invention;

[0084] Figure 8 is a structural diagram of the user interface module of the present invention;

[0085] Figure 9 This is a working principle diagram of the transmission spectrum acquisition unit of the present invention;

[0086] Figure 10 is a flowchart of the multivariate calibration and component prediction unit of the present invention;

[0087] Figure 11 is a flow chart of component-posture correlation analysis of the present invention;

[0088] Figure 12 A flow chart is generated for the personalized feeding regimen of the present invention. DETAILED DESCRIPTION

[0089] Please refer to the attached Figure 1-12 The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood by those skilled in the art that the protection scope of the present invention is not limited to the following embodiments.

[0090] See also Figure 1 The intelligent breastfeeding guidance and milk composition analysis system provided by the present invention includes a milk composition analysis module 1, a breastfeeding posture assessment module 2, a computer vision module 3, a data processing module 4, a feeding plan generation module 5, a remote guidance module 6 and a user interface module 7.

[0091] The milk composition analysis module 1 collects spectral data of breast milk samples, extracts spectral feature information, and generates milk composition analysis results. The breastfeeding posture assessment module 2 collects parameters related to breastfeeding posture and generates breastfeeding posture quality assessment data. The computer vision module 3 collects image data of milk appearance characteristics and infant sucking behavior to generate auxiliary analysis data. The data processing module 4 receives and processes the data of the above modules to generate comprehensive analysis results. The feeding plan generation module 5 generates a personalized feeding plan based on the comprehensive analysis results. The remote guidance module 6 transmits relevant data to the remote professional terminal, receives professional guidance suggestions, and integrates the professional guidance suggestions into the personalized feeding plan. The user interface module 7 is responsible for presenting personalized feeding plans and professional guidance suggestions to the user and receiving user feedback information.

[0092] In a preferred embodiment of the present invention, each module is connected via wired or wireless communication. Preferably, wireless communication uses Bluetooth 5.0 or Wi-Fi 6 technology to ensure stable and efficient data transmission; wired connection uses USB 3.1 interface to ensure fast transmission of large amounts of data.

[0093] See also Figure 2 The milk component analysis module 1 includes a transmission spectrum acquisition unit 11, a multi-level spectrum feature extraction unit 12, a multivariate correction and component prediction unit 13, an indirect component derivation unit 14 and a dynamic closed-loop feedback unit 15.

[0094] The transmission spectrum acquisition unit 11 collects spectral data from breast milk samples using a dynamic thickness control mechanism consisting of a movable glass sample holder and a fixed sample chamber. The multi-level spectral feature extraction unit 12 performs baseline correction on the spectral data, identifies the position, intensity, and area of ​​characteristic peaks, and extracts multidimensional feature vectors. The multivariate correction and component prediction unit 13 generates milk component content data based on the multidimensional feature vectors through a hierarchical prediction model. The indirect component derivation unit 14 derives and calculates the content of milk components that are difficult to measure directly based on the milk component content data. The dynamic closed-loop feedback unit 15 records the time series changes in milk component content, analyzes the correlation between milk composition and breastfeeding posture quality, and generates milk component optimization recommendations.

[0095] In a preferred embodiment of the present invention, see Figure 9 The transmission spectrum acquisition unit 11 includes a light source array 111 , a sample chamber 112 , a movable glass sample holder 113 , a folded optical path system 114 , a temperature control system 115 and a photodetector array 116 .

[0096] The light source array 111 includes multiple light sources covering the 800-4000 cm⁻¹ band, preferably including near-infrared LEDs and mid-infrared quantum cascade lasers. Light sources of different wavelengths are sequentially activated through timing control. The sample chamber 112 is a fixed structure used to accommodate breast milk samples. The movable glass sample holder 113 is vertically movable and, in conjunction with the sample chamber 112, forms a dynamic thickness control mechanism. The movement accuracy is preferably 0.01 mm, ensuring precise and controllable sample thickness.

[0097] Folded optical path system 114 utilizes a multi-faceted elliptical reflector design, enabling multiple reflections within a confined space, effectively extending the interaction path between light and the sample. In one embodiment of the present invention, folded optical path system 114 compresses an effective optical path of 30 cm into a physical space of 10 cm, significantly improving signal strength and detection sensitivity.

[0098] Temperature control system 115 maintains the sample at a constant temperature, preferably 37°C, to simulate the human body's environmental conditions. Temperature control accuracy is ±0.1°C, ensuring consistent and comparable test results. Photodetector array 116, a cooled microphotodiode array, receives light signals transmitted through the sample and generates spectral data.

[0099] The multi-level spectral feature extraction unit 12 performs a series of processing on the raw spectral data, including baseline correction, characteristic peak identification and feature extraction. The baseline correction uses an improved piecewise polynomial fitting algorithm to automatically identify non-characteristic areas and establish a baseline model. The main steps of the algorithm include:

[0100] 1. Divide the spectral data into multiple intervals according to wavenumber, with each interval length being 50 cm-1;

[0101] 2. Apply minimum value screening to each interval to identify possible baseline points;

[0102] 3. Use cubic spline interpolation to connect these baseline points to form a preliminary baseline curve;

[0103] 4. Adjust the baseline point position through iterative optimization algorithm to minimize the difference between the baseline and the actual spectrum in the non-feature area;

[0104] 5. Correct the raw spectra using the final baseline.

[0105] Characteristic peak identification uses a dynamic window local extremum detection method. The core of this method is to automatically adjust the window size according to the spectral curvature. For narrow peaks, a small window (about 5-10 data points) is used to improve the position accuracy, while for broad peaks, a large window (about 20-50 data points) is used to improve the accuracy of area calculation.

[0106] For the separation of overlapping peaks, a mixed Gaussian model is used for peak shape decomposition. The model can be expressed as:

[0107]

[0108] Where S(v) is the spectral signal intensity at wave number v, in units of absorbance; A i is the amplitude of the i-th Gaussian peak, indicating the peak intensity; v i is the central wave number of the i-th peak, in cm-1; σ i is the standard deviation of the ith peak, which is related to the peak width and is expressed in cm-1; B(v) is the baseline function, representing the non-characteristic background; n is the number of peaks, which is usually set to 3-8 peaks for the main components in breast milk. In actual breast milk analysis, the v of the fat peak is i Usually located around 1740cm-1, σ i About 15-25cm-1; the protein peak v i Usually located around 1650 and 1550 cm-1, σ i About 20-30cm-1.

[0109] In a preferred embodiment of the present invention, characteristic peak libraries are established for common components in breast milk, such as fat, protein and lactose, including fat characteristic peaks (near 1740 cm-1), protein characteristic peaks (near 1650 and 1550 cm-1) and lactose characteristic peaks (near 1080-1030 cm-1).

[0110] The multi-dimensional feature vector finally generated by the multi-level spectral feature extraction unit 12 includes:

[0111] 1. Characteristic peak position vector, recording the precise wave number of each main component characteristic peak;

[0112] 2. Peak intensity vector, recording the height value of each characteristic peak;

[0113] 3. Peak area vector, record the integrated area of ​​each characteristic peak;

[0114] 4. Peak shape parameter vector, including half-peak width, asymmetry coefficient, etc.;

[0115] 5. Peak-to-peak relationship vector, recording the intensity ratio, area ratio, etc. of the relevant peaks.

[0116] See also Figure 10 The multivariate correction and component prediction unit 13 includes a first-level linear prediction subunit 131 , a second-level nonlinear prediction subunit 132 , an individualized adjustment subunit 133 , a result fusion subunit 134 and an uncertainty quantization subunit 135 .

[0117] The first-level linear prediction subunit 131 uses the partial least squares regression (PLSR) algorithm to make a preliminary estimate of the content of the main milk components (such as fat, main protein types, and lactose). The basic principle of the PLSR algorithm is to find the maximum correlation between the independent variable (spectral characteristics) and the dependent variable (component content) while reducing the dimension. Its model can be expressed as:

[0118] Y=X·B+E,

[0119] Here, Y is the component content matrix, with dimensions equal to the number of samples x the number of component types, in g / 100ml. X is the spectral feature matrix, with dimensions equal to the number of samples x the feature dimension, and is dimensionless. B is the regression coefficient matrix, representing the mapping between spectral features and component content, with dimensions equal to the feature dimension x the number of component types. E is the residual matrix, representing the portion that cannot be explained by the model, with the same dimensions as Y. In breast milk composition analysis, the X matrix typically contains hundreds of spectral data points, which need to be reduced to 10-30 dimensions through methods such as principal component analysis. The Y matrix typically contains 3-10 major breast milk components, such as fat (standard range 2.5-5.0g / 100ml), protein (standard range 0.9-1.3g / 100ml), and lactose (standard range 6.8-7.2g / 100ml).

[0120] The second-level nonlinear prediction subunit 132 uses the support vector regression (SVR) algorithm to process the nonlinear relationship between complex components and optimize the content prediction results. The SVR algorithm introduces a kernel function to map features into a high-dimensional space to process nonlinear relationships. Its mathematical expression is:

[0121]

[0122] Where f(x) is the prediction function, and the output is the estimated value of the content of a specific component in g / 100ml; x is the input feature vector, which comes from the spectral characteristics of breast milk; x i is the support vector, representing the feature vector in the training set; α i and is the Lagrange multiplier, obtained by optimization algorithm; K(x i ,x) is the kernel function used to calculate the inner product in the feature space; b is the bias term; and n is the number of support vectors, typically 20-30% of the training samples. In breast milk composition analysis, support vector selection focuses on samples that are difficult to accurately predict using linear models, particularly those with extreme fat and protein contents (e.g., fat <2.0g / 100ml or >5.5g / 100ml).

[0123] In a preferred embodiment of the present invention, the kernel function adopts radial basis function (RBF), which is expressed as:

[0124] K(xi , x) = exp(-γ||x i -x|| 2 ),

[0125] Among them, γ is the kernel parameter that controls the width of the function, and its unit is the inverse of the feature space. In breast milk composition analysis, the empirical value of γ is set between 0.1 and 0.5, and the specific value is adjusted according to the degree of nonlinearity of different components. For example, for fat content prediction, γ is usually set to around 0.3 because its relationship with the spectrum is relatively more nonlinear; and for lactose content prediction, γ is usually set to around 0.15 because its relationship is relatively linear. i -x|| 2 represents the eigenvector x i The square of the Euclidean distance between x and x is used to quantify the similarity between two samples in the feature space.

[0126] The individualized adjustment subunit 133 performs personalized adjustments to the prediction model parameters based on the user's historical data. This unit uses transfer learning technology to adjust the general model trained on a large sample to a personalized model that adapts to individual characteristics. The specific implementation method is domain adaptation technology, the core idea of ​​which is to minimize the distribution difference between the source domain (general model) and the target domain (individual data). The mathematical expression is:

[0127] L total =L pred +λ·L adapt ,

[0128] Among them, L total is the total loss function, the target to be minimized; L pred is the prediction loss function, usually using mean square error or absolute error to measure the accuracy of model prediction; L adapt λ is the domain adaptation loss function, which measures the difference between the source and target domain distributions, typically using the Maximum Mean Difference (MMD) metric. λ is a trade-off coefficient, controlling the balance between prediction accuracy and domain adaptation, with an empirical value of 0.3-0.7. The value of λ varies across lactation stages: in early lactation (1-14 days postpartum), individual variability is large, so λ is set higher (approximately 0.6-0.7). During the stable lactation period (after 15 days postpartum), individual variability decreases, and λ can be lowered to 0.3-0.4.

[0129] The result fusion subunit 134 integrates the prediction results of the first-level linear prediction subunit 131 and the second-level nonlinear prediction subunit 132 to generate the final milk component content data. The fusion method adopts a weighted average strategy, and the weight coefficient is dynamically adjusted according to the characteristics of each component. The mathematical expression is:

[0130] C final=w1·C linear +w2·C nonlinear ,

[0131] Among them, C f inal is the predicted value of the final ingredient content, in g / 100ml; C linear is the predicted value of the linear model (PLS R), in g / 100ml; C nonlinear is the predicted value of the nonlinear model (SVR) in g / 100 ml; w1 and w2 are weight coefficients, satisfying w1 + w2 = 1. In practice, weight coefficients are set differently for different components: w1 = 0.3 and w2 = 0.7 for fat, as the relationship between fat and the spectrum is relatively nonlinear; w1 = 0.4 and w2 = 0.6 for protein; and w1 = 0.6 and w2 = 0.4 for lactose, as the relationship between lactose and the spectrum is relatively linear.

[0132] Uncertainty quantification subunit 135 generates a confidence interval for each prediction result, indicating the accuracy of the prediction. This confidence interval calculation is based on Monte Carlo simulation, which estimates the uncertainty of the prediction result by repeatedly randomly perturbing the model parameters and observing the range of output variation. For important components such as fat and protein, a 95% confidence interval is generally required to have a relative error of no more than ±5%.

[0133] The indirect component derivation unit 14 uses milk component content data to derive and calculate the content of milk components that are difficult to measure directly. This unit establishes a derivation model based on biochemical relationships to convert directly measured components (such as free amino acids and reducing sugars) into estimates of the content of complex components (such as total protein and polysaccharide structure).

[0134] In one embodiment of the present invention, the total amount of protein is derived using amino acid profile pattern analysis combined with dynamic compensation of physiological state, and the expression is:

[0135] P total =f1(AA free )+α·f2(D,F)+β·f3(I),

[0136] Among them, P total is the estimated value of total protein, in g / 100ml; AA freeis the free amino acid content vector, containing the content of various amino acids, in mg / 100 ml; f1 is the amino acid-protein mapping function, which estimates the total protein content based on the amino acid composition pattern; D is the number of days postpartum, in days; F is the frequency of lactation, in times / day; f2 is the physiological state compensation function, which accounts for the effects of lactation stage and frequency on protein content; I is the milk image feature vector, containing parameters such as color and turbidity; f3 is the image feature compensation function; α and β are weighting coefficients, with empirical values ​​of 0.2-0.4 and 0.1-0.3, respectively. In practice, colostrum protein content is high in the early postpartum period (days 1-5) (approximately 1.5-2.0 g / 100 ml), then gradually decreases to mature milk levels (approximately 0.9-1.3 g / 100 ml). Function f2(D, F) will automatically adjust the expected protein content range according to the number of days after delivery D: when D < 5 days, the f2 contribution value is approximately 0.3-0.7g / 100ml; when D > 14 days, the f2 contribution value is approximately 0-0.2g / 100ml.

[0137] Similarly, the derived expression for the total amount of polysaccharides is:

[0138] S total =g1(S red )+γ·g2(D,F)+δ·g3(I),

[0139] Among them, s total is the estimated value of total polysaccharide content, in g / 100ml; S red is the reducing sugar content, in g / 100ml; g1 is the reducing sugar-polysaccharide mapping function; g2 and g3 are the physiological state compensation function and image feature compensation function, respectively; γ and δ are weight coefficients, with empirical values ​​of 0.3-0.5 and 0.1-0.2, respectively. In breast milk, lactose is the main carbohydrate, with a content usually between 6.8-7.2g / 100ml. Function g1(S red ) converts the measured reducing sugar content to an estimate of total lactose using a conversion factor of approximately 1.05-1.15, since galactose and glucose, produced by lactose hydrolysis, are both reducing sugars. Function g2(D,F) accounts for the effect of lactation stage on lactose content: lactose levels are slightly lower in colostrum (approximately 5.5-6.5 g / 100 ml), while mature milk has a more stable lactose content (approximately 6.8-7.2 g / 100 ml).

[0140] The dynamic closed-loop feedback unit 15 records the temporal changes in milk composition, analyzes the correlation between milk composition and breastfeeding posture quality, and generates recommendations for optimizing milk composition. This unit uses time series analysis technology to establish a composition variation model. Combined with breastfeeding posture quality data, it constructs a three-dimensional correlation model between composition, posture, and absorption efficiency.

[0141] The core of this correlation model is to quantify the influence coefficient of different breastfeeding posture qualities on the absorption efficiency of each component, which can be expressed as a matrix form:

[0142] E=Q·M,

[0143] Where E is the absorption efficiency matrix with the dimension of c×p, c is the number of components, p is the number of posture types, and the matrix element E ij It represents the absorption efficiency of the j-th posture for the i-th component, with a value range of 0-1; Q is the posture quality matrix with a dimension of p×q, q is the number of posture quality parameters, and the matrix element Q jk Indicates the score of the kth quality parameter of the jth posture, usually 0-10 points; M is the mapping matrix with dimension q×c, and the matrix element M ki This represents the coefficient of influence of the kth posture quality parameter on the absorption efficiency of the ith component. In actual breastfeeding analyses, c is typically 3-10 (including key nutrients such as fat, protein, and lactose), p is 5-8 (including common breastfeeding posture types such as the cradle, cross-cradle, and rugby position), and q is 10-15 (including posture quality assessment parameters such as back support, breast position, and head and neck alignment).

[0144] For example, in a practical case, for the absorption of fat components, the influence coefficient of the deep sucking parameter M ki The influence coefficients for head and neck linear parameters are approximately 0.15-0.20, the influence coefficients for head and neck linear parameters are approximately 0.10-0.15, and the influence coefficients for body alignment parameters are approximately 0.08-0.12. This means that when the deep suck score increases from 5 to 9, fat absorption efficiency can increase by approximately 0.15 × (9-5) = 0.6, or 60%. Through this quantitative analysis, the system can prioritize and recommend improvements to the posture parameters with the highest influence coefficients for different mother-infant pairs, thereby most effectively improving nutrient absorption.

[0145] See also Figure 3 The breastfeeding posture evaluation module 2 includes a sensor array unit 21 , a posture type recognition unit 22 , a posture quality evaluation unit 23 , a contact area calculation unit 24 and a posture stability evaluation unit 25 .

[0146] The sensor array unit 21 includes a 9-axis motion sensor, a pressure sensor, and a distance sensor to collect parameters related to breastfeeding posture. The posture type recognition unit 22 identifies the current breastfeeding posture type based on the breastfeeding posture parameters. The posture quality assessment unit 23 quantifies the execution quality of the breastfeeding posture type and generates posture quality score data. The contact area calculation unit 24 calculates the contact area and pressure distribution between the breast and the baby to generate contact feature data. The posture stability assessment unit 25 analyzes the frequency and amplitude of posture changes during breastfeeding to generate stability assessment data.

[0147] In a preferred embodiment of the present invention, the sensor array unit 21 includes a 9-axis motion sensor array distributed on the breast, back, and abdomen of a nursing mother. Each sensor integrates a 3-axis accelerometer, a 3-axis gyroscope, and a 3-axis magnetometer, with a sampling frequency of 50Hz to ensure that subtle posture changes are captured. In addition, it also includes a thin film pressure sensor array with a distribution density of 4 sensing points per square centimeter and a sensitivity of 5-500g / cm 2 and an infrared distance sensor with a measuring range of 5-50cm and an accuracy of ±1mm.

[0148] The posture recognition unit 22 uses a hybrid classification algorithm combining a decision tree and a support vector machine (SVM) to analyze sensor data in real time and identify the current breastfeeding posture. This unit can identify eight common breastfeeding postures: cradle, cross-cradle, rugby ball, side-lying, supine, upright, back-leaning, and twin-feet. In practice, posture recognition accuracy exceeds 95%.

[0149] The posture quality evaluation unit 23 quantitatively scores the execution quality of the breastfeeding posture based on multi-dimensional parameters. The scoring indicators include:

[0150] 1. Back support: assesses the fit between the mother's back and the support surface, with a maximum score of 10 points;

[0151] 2. Breast position: assess whether the breast is properly positioned relative to the baby's mouth, with a maximum score of 10 points;

[0152] 3. Head and neck linearity: assess whether the baby's head and neck are in a straight line, with a maximum score of 10 points;

[0153] 4. Body alignment: assesses whether the baby's body is aligned with the mother's body, with a maximum score of 10 points;

[0154] 5. Postural comfort: Assess the mother's overall comfort level, with a maximum score of 10.

[0155] The comprehensive score adopts the weighted average method. The weight of each indicator is set according to the degree of influence on absorption efficiency. Generally, the weight of back support is 0.15, the weight of breast position is 0.25, the weight of head and neck linearity is 0.2, the weight of body alignment is 0.2, and the weight of posture comfort is 0.2.

[0156] The contact area calculation unit 24 calculates the effective contact area and pressure distribution between the breast and the baby's mouth based on the pressure sensor array data. In a preferred embodiment of the present invention, the ideal contact area is about 4-6 cm 2 , the pressure is evenly distributed, and the central pressure value is about 100-200g / cm 2 The contact area is too small (<3cm2 ) or too large (>7cm 2 ), or uneven pressure distribution, will lead to reduced sucking efficiency and affect nutrient absorption.

[0157] The posture stability assessment unit 25 evaluates posture stability during breastfeeding by analyzing fluctuations in the 9-axis motion sensor data. This unit calculates the standard deviation of acceleration and angular velocity for each axis to generate a stability score. In one embodiment of the present invention, a posture change frequency of less than 0.5 times / minute and an amplitude change (acceleration standard deviation) of less than 0.1g is rated as "high stability." A frequency between 0.5 and 2 times / minute or an amplitude change between 0.1 and 0.3g is rated as "medium stability." A frequency greater than 2 times / minute or an amplitude change greater than 0.3g is rated as "low stability."

[0158] See also Figure 4 The computer vision module 3 includes an image acquisition unit 31 , a milk appearance analysis unit 32 , a sucking behavior recognition unit 33 , a swallowing detection unit 34 and a milk flow analysis unit 35 .

[0159] The image acquisition unit 31 collects image data of milk appearance characteristics and infant sucking behavior. The milk appearance analysis unit 32 analyzes milk color, turbidity, and uniformity to generate milk appearance characteristic data. The sucking behavior recognition unit 33 uses the Hough transform algorithm to locate the infant's mouth and neck, identify the sucking position and sucking pattern, and generate sucking behavior characteristic data. The swallowing detection unit 34 identifies and counts swallowing movements to generate swallowing frequency data. The milk flow analysis unit 35 uses visual tracking technology to analyze milk flow conditions, estimate milk flow rate and volume, and generate flow characteristic data.

[0160] In a preferred embodiment of the present invention, the image acquisition unit 31 includes a high-definition digital camera with a resolution of 1920×1080 pixels and a frame rate of 30 fps. It is equipped with autofocus and low-light enhancement features to ensure clear image data in various environmental conditions. Furthermore, an infrared night vision function is optionally available to support nighttime breastfeeding monitoring.

[0161] The milk appearance analysis unit 32 analyzes the collected milk images, extracting color, turbidity, and uniformity features. Color analysis uses the RGB and HSV color spaces, turbidity analysis is based on image contrast and scattering characteristics, and uniformity analysis is achieved by calculating texture features. These characteristic parameters can be used to preliminarily determine the milk's fat content (higher turbidity generally indicates higher fat content) and uniformity (which affects nutrient distribution).

[0162] The sucking behavior recognition unit 33 uses the Hough transform algorithm to locate the outline of the baby's mouth and neck. In the present invention, the improved Hough transform algorithm used takes into account the relative position relationship between facial feature points, improving the recognition accuracy under partial occlusion. Its mathematical expression is simplified to:

[0163]

[0164] Among them, H(x c ,y c , r) is the cumulative matrix, indicating that (x c ,y c ) as the center and the possibility of a circle with a radius of r. The larger the value, the more likely the position is to be a circular contour; (x c ,y c ) is the coordinate of the circle center, in pixels; r is the circle radius, in pixels; I is the edge image, which represents the binary image after edge detection; w(x, y) is the position weight function, which assigns different weights according to the facial feature areas, such as the eyes, nose, and mouth areas have higher weights (0.8-1.0), while the edge areas have lower weights (0.3-0.5); δ is the impulse function, when it satisfies The value is 1 when the condition is met, otherwise it is 0. In practical applications, the algorithm parameter settings are related to the age of the infant: for newborns, the search range of r is usually set to 15-25 pixels; for infants over 3 months old, the search range of r is expanded to 20-35 pixels.

[0165] Based on the positioning results, the unit is able to identify four basic sucking patterns:

[0166] 1. Shallow sucking: The baby only holds the nipple, with a small sucking amplitude and low efficiency;

[0167] 2. Deep sucking: The baby holds the areola in his mouth, sucking with a large amplitude and high efficiency;

[0168] 3. Sucking and licking: alternating sucking and licking, common in newborns;

[0169] 4. Rolling suction type: a vacuum is formed in the front of the mouth to improve the sucking efficiency.

[0170] The swallowing detection unit 34 identifies swallowing movements by analyzing the infant's neck movement patterns. This unit uses optical flow combined with a deep learning model to identify a typical swallowing sequence: slight jaw lift, contraction of the upper neck muscles, and subsequent movement of the Adam's apple. In a preferred embodiment of the present invention, this unit achieves swallowing detection accuracy exceeding 90%, and can count the number and frequency of swallows in real time, serving as an important indicator of milk intake.

[0171] The milk flow analysis unit 35 uses fluid dynamics analysis techniques from computer vision to observe the flow of milk during breastfeeding. By analyzing subtle changes in the breast surface and traces of fluid around the baby's mouth, it estimates the milk flow rate and volume. In practice, this unit categorizes flow rates into three levels: high (>5 ml / min), medium (2-5 ml / min), and low (<2 ml / min), providing an objective basis for evaluating breastfeeding efficiency.

[0172] See also Figure 5 The data processing module 4 includes a data synchronization unit 41, a component-posture association analysis unit 42, a time series trend analysis unit 43, an anomaly detection unit 44 and a cloud data processing unit 45.

[0173] The data synchronization unit 41 receives and synchronizes milk composition analysis results, breastfeeding posture quality assessment data, and auxiliary analysis data. The component-posture correlation analysis unit 42 analyzes the correlation between milk composition and breastfeeding posture quality, establishing a component-posture-absorption efficiency correlation model. The temporal trend analysis unit 43 analyzes the changing trends of milk composition over time and predicts future trends. The anomaly detection unit 44 identifies milk composition that deviates from the normal range, generates an anomaly flag, and generates a risk level assessment. The cloud-based data processing unit 45 summarizes the analysis results and compares them with group data to generate a comprehensive analysis result.

[0174] In a preferred embodiment of the present invention, the data synchronization unit 41 employs a timestamp alignment strategy to ensure that data from different modules are correctly linked in chronological order. Data synchronization employs a double-buffering mechanism to address the issue of inconsistent data collection frequencies across modules: milk composition analysis data is typically collected intermittently (before and after each feeding), while breastfeeding posture data and computer vision data are collected continuously (5-50 Hz). By setting a sliding time window (preferably 5-30 seconds), the continuous data is aggregated into feature vectors and paired with the discrete milk composition data.

[0175] The component-posture correlation analysis unit 42 is one of the core innovations of the present invention, which realizes the quantitative analysis of the relationship between milk components and the quality of breastfeeding posture. Figure 11 This unit uses a multivariate analysis method to establish an association model. First, a correlation analysis is performed between the breastfeeding posture quality parameter matrix Q (dimensions p×q, where p is the number of posture types and q is the number of quality parameters) and the milk component absorption efficiency matrix E (dimensions c×p, where c is the number of components and p is the number of posture types), to find the mapping relationship M (dimensions q×c).

[0176] The mapping relationship is solved by using the partial least squares method combined with regularization technology. The mathematical expression is:

[0177]

[0178] Among them, ||·|| F λ represents the Frobenius norm, which is used to calculate the size of the matrix and is defined as the square root of the sum of the squares of the matrix elements. λ is a regularization parameter that controls model complexity and prevents overfitting, with an empirical value of 0.01-0.1. The setting of λ varies for different breast milk components: smaller values ​​(approximately 0.01-0.03) are typically used for fat because of its significant correlation with postural quality; larger values ​​(approximately 0.05-0.1) are used for protein and lactose to avoid model overfitting. In practice, by minimizing the above expression, the optimal mapping matrix M can be obtained, thereby establishing a quantitative relationship between postural quality parameters and the absorption efficiency of each component.

[0179] For example, in the case of a mother who was 2 months postpartum, the system found that the mother's milk fat content was 3.2g / 100ml (within the normal range), but the baby's actual absorption efficiency was low, only 65%. Through component-posture association analysis, the system identified that the key issues were the low head and neck linear parameter score (5.5 points out of 10) and insufficient sucking depth (shallow sucking). In response to this situation, the system recommended adjusting the posture to ensure that the baby's head and neck remain in a straight line, and guiding the correct sucking technique so that the baby can suck deeply. After implementing the recommendations, the fat absorption efficiency increased to 85%, and the baby's weight gain rate improved significantly.

[0180] The Time Series Trend Analysis Unit 43 uses time series analysis techniques to track and record the changing trends of breast milk composition over time. This unit combines the Autoregressive Integrated Moving Average (ARIMA) model with seasonal decomposition methods to identify cyclical patterns and long-term trends in breast milk composition changes. In the breast milk composition analysis scenario, changes on three time scales are of primary interest:

[0181] 1. Intraday variation: Fluctuations in milk composition over a 24-hour period, such as fat content, which is usually lower in the morning and higher in the evening;

[0182] 2. Inter-lactation variation: the difference in composition between foremilk and hindmilk during the same lactation;

[0183] 3. Long-term changes: Gradual changes in milk composition as lactation progresses.

[0184] The abnormality detection unit 44 identifies milk components that deviate from the normal range based on statistical methods and expert knowledge base. This unit adopts a three-level threshold strategy:

[0185] 1. Normal range: The content of each component is within the reference range and no intervention is required;

[0186] 2. Warning range: If the ingredient content slightly deviates from the reference range, the system will give adjustment suggestions;

[0187] 3. Abnormal range: The content of the ingredient deviates seriously from the reference range, triggering the early warning mechanism. It is recommended to seek professional guidance.

[0188] Taking fat content as an example, the reference range is 2.5-5.0g / 100ml, the warning range is 1.5-2.5g / 100ml or 5.0-6.0g / 100ml, and the abnormal range is <1.5g / 100ml or >6.0g / 100ml. For special indicators such as antibiotic residues, any detection is considered abnormal and triggers a high-level warning.

[0189] The cloud data processing unit 45 uploads the local analysis results to the cloud platform for comparison and statistical analysis with anonymous group data. This unit uses regional hashing technology and differential privacy protection algorithms to ensure user data security. The main functions of cloud processing include:

[0190] 1. Group data comparison: compare individual data with group data under similar conditions (e.g., similar number of days postpartum, similar physical conditions);

[0191] 2. Model optimization: Continuously optimize prediction models and correlation models based on large-scale data;

[0192] 3. Knowledge mining: Discover new nutritional patterns and associations and update the expert knowledge base.

[0193] In practice, cloud data processing utilizes an incremental computing strategy. When new data arrives, only the affected computational results are updated, significantly improving processing efficiency. Regarding data security, in addition to basic encrypted transmission, data sharding and abnormal access detection technologies are employed to comprehensively safeguard user privacy.

[0194] See also Figure 6 The feeding plan generation module 5 includes a goal setting unit 51, a multi-goal optimization unit 52, a posture improvement suggestion unit 53, a feeding frequency suggestion unit 54, a sucking behavior guidance unit 55 and a plan integration unit 56.

[0195] The goal setting unit 51 receives and stores the user-set milk composition optimization goal. The multi-objective optimization unit 52 performs a multi-objective balance calculation based on the comprehensive analysis results and the milk composition optimization goal, generating an optimization strategy. The posture improvement suggestion unit 53 generates suggestions for improving breastfeeding posture based on the optimization strategy. The feeding frequency suggestion unit 54 generates suggestions for feeding frequency and schedule based on the optimization strategy. The sucking behavior guidance unit 55 generates guidance suggestions for infant sucking behavior based on the optimization strategy. The plan integration unit 56 integrates these suggestions into a complete, personalized feeding plan.

[0196] In a preferred embodiment of the present invention, the goal setting unit 51 provides three goal setting modes:

[0197] 1. Balanced nutrition mode: pursuit of balanced optimization of various nutrients;

[0198] 2. Focused enhancement mode: Optimize specific ingredients (such as fat or protein);

[0199] 3. Intelligent recommendation mode: The system automatically recommends the most suitable optimization target based on the baby's growth data and breast milk analysis results.

[0200] The multi-objective optimization unit 52 is the core link of the feeding plan generation, and adopts the Pareto optimization method to deal with multiple optimization objectives that may conflict with each other. Figure 12 , this method defines the objective function vector F(x)=[f1(x),f2(x),...,f n (x)], where x is the decision variable vector (including posture parameters, feeding frequency, etc.), f i (x) is the i-th optimization goal (such as the absorption efficiency of each nutrient).

[0201] When dealing with multi-objective problems, a solution x1 is said to dominate another solution x2 if and only if for all i,f i (x1)≤f i (x2) (assuming the goal is to minimize), and there is at least one j such that f j (x1)<f j (x2). The Pareto front is the set of all non-dominated solutions.

[0202] In this paper, the multi-objective optimization adopts the improved NSGA-II algorithm (Non-dominated Sorting Genetic Algorithm II), which can quickly converge to the Pareto front and maintain the diversity of solutions through the elite retention strategy and crowding distance sorting mechanism. The main steps of the algorithm include:

[0203] 1. Initialize the population, including multiple possible feeding plans;

[0204] 2. Perform non-dominated sorting on the population and classify the solutions into different levels;

[0205] 3. Calculate the congestion of each solution to ensure the diversity of solutions;

[0206] 4. Generate new populations through selection, crossover and mutation operations;

[0207] 5. Merge the parent and offspring populations and select the best solution to form a new generation;

[0208] 6. Repeat steps 2-5 until the termination condition is reached.

[0209] Finally, the unit selects an optimal equilibrium point from the Pareto frontier as the optimization strategy, and the selection criteria are based on the user's preference weight or the system's intelligent recommendation.

[0210] The posture improvement suggestion unit 53 generates specific suggestions for breastfeeding posture improvement based on the optimization strategy. This unit uses a gap analysis method to compare the current breastfeeding posture quality with the target requirement, focusing on the posture parameters that have the greatest impact on the absorption of the target component. The suggestions include posture type selection, specific posture adjustment methods, and adjustment range, presented to the user in an intuitive and easy-to-understand format.

[0211] The feeding frequency recommendation unit 54 optimizes feeding frequency and timing based on the milk composition analysis results and absorption efficiency. This unit considers three key factors:

[0212] 1. Milk secretion pattern: adjust feeding time according to the user's milk secretion peak and trough period;

[0213] 2. Daily variation pattern of ingredients: Consider the changes in the content of different ingredients throughout the day and arrange the best feeding time;

[0214] 3. Infant energy needs: Calculate appropriate feeding intervals based on the infant's growth stage and activity level.

[0215] In actual application, feeding frequency recommendations are usually divided into two periods: daytime and nighttime. The recommended interval during the day is 2-3 hours, and it can be appropriately extended to 3-4 hours at night, but always respecting the baby's hunger signals first.

[0216] The sucking behavior guidance unit 55 generates guidance suggestions for the baby's sucking behavior based on the sucking behavior analysis results of the computer vision module. This unit provides targeted solutions for different sucking problems, such as:

[0217] 1. Shallow sucking problem: It is recommended to adjust the baby's feeding position to ensure that enough of the areola is covered;

[0218] 2. Weak sucking: It is recommended to try the U-shaped breast grip method to provide additional support;

[0219] 3. Sucking-swallowing incoordination: It is recommended to use slow feeding techniques and give the baby enough breathing time.

[0220] The plan integration unit 56 integrates the suggestions from various units to form a complete personalized feeding plan. This unit adopts a hierarchical structure, dividing the suggestions into three levels: core suggestions (must be implemented), auxiliary suggestions (conditionally implemented), and reference information (for user understanding). The suggestions are sorted according to chronological order and operational logic to ensure the implementability and coherence of the plan.

[0221] See also Figure 7 The remote guidance module 6 includes a data encryption transmission unit 61, a lactation consultant communication unit 62, a physician diagnosis communication unit 63 and a professional advice integration unit 64.

[0222] The data encryption and transmission unit 61 encrypts the milk composition analysis and comprehensive analysis results and securely transmits them to a remote professional terminal. The lactation consultant communication unit 62 establishes a communication connection with the lactation consultant to receive advice on correcting breastfeeding posture. The physician diagnosis communication unit 63 establishes a communication connection with the physician to receive diagnoses and recommendations on abnormal milk composition. The professional advice integration unit 64 integrates the advice provided by professionals into professional guidance.

[0223] In a preferred embodiment of the present invention, the data encryption transmission unit 61 employs end-to-end encryption technology to ensure the security of sensitive data during transmission. Specifically, it utilizes the AES-256 encryption algorithm and the RSA asymmetric key exchange mechanism, meeting the high standards for healthcare data protection. Furthermore, this unit implements data desensitization, removing or obfuscating user identification information before transmission, retaining only the data necessary for analysis.

[0224] The lactation consultant communication unit 62 establishes a communication bridge between the user and the professional lactation consultant. This unit supports three communication modes:

[0225] 1. Asynchronous consultation: Users submit questions and consultants respond within 24 hours;

[0226] 2. Appointment consultation: Users can schedule a specific time for a real-time video consultation;

[0227] 3. Emergency consultation: When encountering urgent problems, quickly connect to the online on-duty consultant.

[0228] The unit also provides intelligent screening functions to match the most suitable professional consultants according to the user's specific problems and needs, such as lactation experts, posture guidance experts or newborn lactation experts.

[0229] The physician diagnostic communication unit 63 is specifically designed for professional diagnosis of abnormal breast milk composition. When the abnormality detection unit detects a significant deviation from normal breast milk composition or detects potentially harmful substances (such as antibiotic residues or heavy metals), it automatically sends a diagnostic request to a specialist physician. The physician can remotely review the detailed analysis data and, combined with supplemental information provided by the user (such as medication status and dietary habits), provide a professional diagnosis and treatment recommendations.

[0230] The professional advice integration unit 64 is responsible for integrating advice from various professionals and transforming it into user-interpretable and actionable guidance. This unit uses semantic analysis technology to extract key information from professional advice, structure it, and integrate it with the system-generated feeding plan. During this integration process, the unit prioritizes physician recommendations, followed by lactation consultant recommendations, and finally system-generated recommendations. This ensures that professional medical advice is prioritized when potential health risks arise.

[0231] See also Figure 8 The user interface module 7 includes a touch screen unit 71 , a data visualization unit 72 , a voice broadcast unit 73 , a posture demonstration unit 74 , a user feedback collection unit 75 and a feedback transmission unit 76 .

[0232] The touchscreen display unit 71 presents personalized feeding plans and professional guidance suggestions in the form of a graphical interface. The data visualization unit 72 converts milk composition data, breastfeeding posture quality data, and time-series change data into intuitive charts and trend lines. The voice broadcast unit 73 provides real-time guidance and suggestions via voice. The posture demonstration unit 74 demonstrates the correct breastfeeding posture through three-dimensional animation. The user feedback collection unit 75 collects user feedback on operation and execution results. The feedback transmission unit 76 transmits user feedback information to the data processing module and feeding plan generation module for plan optimization.

[0233] In a preferred embodiment of the present invention, the touchscreen display unit 71 utilizes a 7-10-inch high-definition touchscreen with a resolution of at least 1920×1080 pixels, supporting multi-touch and gesture operation. The interface design adheres to ergonomic principles, taking into account the single-handed operation of breastfeeding mothers. Large buttons, a simple layout, and an intuitive color coding system ensure convenient operation. The interface primarily includes four functional areas: milk analysis, breastfeeding guidance, history recording, and professional consultation.

[0234] The Data Visualization Unit 72 converts complex analytical data into intuitive and easy-to-understand visualizations. This unit supports a variety of visualization charts, including:

[0235] 1. Ingredient content radar chart: intuitively displays the relative content of each nutrient;

[0236] 2. Time series trend line chart: shows the changing trend of key components over time;

[0237] 3. Posture quality heat map: Displays the scores of each posture parameter in the form of a heat map;

[0238] 4. Comparison bar chart: Compare user data with reference values ​​or historical data.

[0239] When displaying data, the unit automatically filters the most critical information to avoid information overload. At the same time, it uses eye-catching colors and marks to attract users' attention to abnormal data.

[0240] The voice announcement unit 73 uses natural speech synthesis technology to provide real-time voice guidance and feedback. This unit supports multiple languages ​​and dialects, allowing users to select their preferred language. Voice content includes operational guidance, data interpretation, suggestions, and reminders. This voice guidance feature is particularly useful during breastfeeding, when hands are limited in operating the screen. Voice announcements use a graded volume control strategy, with important reminders at a higher volume and general information at a lower volume to avoid disturbing the baby's rest.

[0241] The Posture Demonstration Unit 74 uses 3D animation technology to visually display recommended breastfeeding postures and adjustment methods. This unit includes standard demonstration animations for all common breastfeeding postures and can generate personalized adjustment guidance animations based on the user's specific situation and system recommendations. The animation supports functions such as multi-angle viewing, zooming in on specific areas, and slow playback, helping users to accurately understand the details of correct posture. In addition, this unit provides a comparison function, displaying the user's current posture (captured by the computer vision module) side by side with the standard posture, intuitively highlighting areas that need improvement.

[0242] The user feedback collection unit 75 is responsible for collecting user operation feedback and execution effect feedback, and is an important data source for continuous optimization of the system. The unit provides multiple feedback channels, including rating scales (1-5 stars), multiple-choice questions (multiple choices), short text input, and voice recording. The feedback content covers aspects such as system usage experience, difficulty of executing suggestions, and effect evaluation. To increase user participation, the unit adopts a lightweight design, with a single feedback operation not exceeding 30 seconds, and provides appropriate incentive mechanisms, such as unlocking additional functions or obtaining points by completing feedback.

[0243] Feedback transmission unit 76 securely transmits user feedback information to the data processing module and feeding plan generation module for continuous optimization of system performance and personalized solutions. This unit adopts an incremental transmission strategy, transmitting only new or changed feedback data to reduce network load. Data transmission prioritizes Wi-Fi networks, automatically switching to mobile data networks when Wi-Fi is unavailable, ensuring that feedback information reaches the backend system in a timely manner. Furthermore, this unit implements an offline caching mechanism, temporarily storing feedback data when the network is unavailable and automatically synchronizing it upon network recovery to ensure data integrity.

[0244] In summary, the intelligent breastfeeding guidance and milk composition analysis system provided by the present invention realizes all-round intelligent guidance of the breastfeeding process through the collaborative work of multiple modules such as milk composition analysis, breastfeeding posture assessment, and computer vision analysis. Based on the convenient detection of milk composition, the system combines the objective assessment of breastfeeding posture and the analysis of infant sucking behavior to establish a correlation model between milk composition and breastfeeding posture quality, providing data support for scientific feeding. Through a closed-loop feedback mechanism, the system can continuously optimize the feeding plan, improve the efficiency of nutrient absorption, and promote the healthy growth of infants and young children. At the same time, the system also supports remote professional guidance, provides timely and professional intervention suggestions for abnormal situations, and improves the safety and effectiveness of breastfeeding.

[0245] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention are intended to be included within the scope of protection of the present invention.

Claims

1. Breastfeeding intelligent guidance and milk composition analysis system, characterized by: include: Milk composition analysis module, used to collect spectral data of breast milk samples, extract spectral feature information, and generate milk composition analysis results; a breastfeeding posture assessment module, communicatively connected to the milk composition analysis module, for collecting breastfeeding posture-related parameters and generating breastfeeding posture quality assessment data; a computer vision module, in communication with the milk composition analysis module and the breastfeeding posture assessment module, for collecting image data of milk appearance characteristics and infant sucking behavior to generate auxiliary analysis data; a data processing module, communicatively connected to the milk composition analysis module, the breastfeeding posture assessment module, and the computer vision module, configured to receive and process the milk composition analysis results, the breastfeeding posture quality assessment data, and the auxiliary analysis data, and generate a comprehensive analysis result; a feeding plan generating module, communicatively connected to the data processing module, for generating a personalized feeding plan based on the comprehensive analysis results; a remote guidance module, communicatively connected to the data processing module and the feeding plan generating module, configured to transmit the milk composition analysis results and the comprehensive analysis results to a remote professional terminal, receive professional guidance suggestions, and integrate the professional guidance suggestions into the personalized feeding plan; A user interface module is in communication with the feeding plan generation module and the remote guidance module, and is used to present the personalized feeding plan and the professional guidance suggestions to the user and receive user feedback information.

2. The breastfeeding intelligent guidance and milk composition analysis system according to claim 1, characterized in that: The milk component analysis module includes: A transmission spectrum acquisition unit, used to collect spectral data of breast milk samples through a dynamic thickness control mechanism consisting of a movable glass sample holder and a fixed sample chamber; a multi-level spectral feature extraction unit, communicatively connected to the transmission spectrum acquisition unit, for performing baseline correction on the spectral data, identifying characteristic peak positions, intensities and areas, and extracting multi-dimensional feature vectors; a multivariate correction and component prediction unit, communicatively connected to the multi-level spectral feature extraction unit, for generating milk component content data through a hierarchical prediction model based on the multi-dimensional feature vector; an indirect component derivation unit, communicatively connected to the multivariate correction and component prediction unit, for deducing and calculating the content of milk components that are difficult to measure directly based on the milk component content data; The dynamic closed-loop feedback unit is in communication with the indirect component derivation unit and is used to record the temporal changes in the milk component content, analyze the correlation between the milk component and the quality of the breastfeeding posture, and generate milk component optimization suggestions.

3. The breastfeeding intelligent guidance and milk composition analysis system according to claim 2, characterized in that: The transmission spectrum acquisition unit comprises: A light source array, configured to emit light covering near-infrared and mid-infrared bands; a sample chamber for holding a breast milk sample; A movable glass sample holder, which cooperates with the sample chamber to form a dynamic thickness control mechanism, and adjusts the sample thickness by moving up and down; Folded optical path system to increase the interaction path between light and sample; Temperature control system to maintain the sample at a constant temperature of 37°C; The photodetector array is used to receive the light signal transmitted through the sample and generate spectral data.

4. The breastfeeding intelligent guidance and milk composition analysis system according to claim 2, characterized in that: The multivariate correction and component prediction unit includes: The first-level linear prediction subunit is used to make preliminary content estimates of the main milk components; The second-level nonlinear prediction subunit is used to process the nonlinear relationship between complex components and optimize the content prediction results; The individualized adjustment subunit is used to perform personalized adjustments to the prediction model parameters based on user historical data; a result fusion subunit, configured to integrate the prediction results of the first-level linear prediction subunit and the second-level nonlinear prediction subunit to generate final milk component content data; The uncertainty quantification subunit is used to generate a confidence interval for each prediction result to indicate the accuracy of the prediction.

5. The breastfeeding intelligent guidance and milk composition analysis system according to claim 1, characterized in that: The breastfeeding posture assessment module includes: A sensor array unit, including a 9-axis motion sensor, a pressure sensor, and a distance sensor, for collecting parameters related to breastfeeding posture; a posture type identification unit, communicatively connected to the sensor array unit, for identifying a current breastfeeding posture type based on the breastfeeding posture-related parameters; a posture quality assessment unit, communicatively connected to the sensor array unit and the posture type recognition unit, for quantitatively scoring the execution quality of the breastfeeding posture type and generating posture quality score data; a contact area calculation unit, communicatively connected to the sensor array unit, for calculating the contact area and pressure distribution between the breast and the baby, and generating contact characteristic data; The posture stability evaluation unit is communicatively connected to the sensor array unit and is used to analyze the frequency and amplitude of posture changes during breastfeeding and generate stability evaluation data.

6. The breastfeeding intelligent guidance and milk composition analysis system according to claim 1, characterized in that: The computer vision module includes: An image acquisition unit, used to collect image data of milk appearance characteristics and infant sucking behavior; a milk appearance analysis unit, communicatively connected to the image acquisition unit, for analyzing the color, turbidity, and uniformity of the milk and generating milk appearance characteristic data; a sucking behavior recognition unit, communicatively connected to the image acquisition unit, for locating the infant's mouth and neck using a Hough transform algorithm, identifying the sucking position and sucking pattern, and generating sucking behavior characteristic data; a swallowing detection unit, communicatively connected to the sucking behavior recognition unit, for recognizing and counting swallowing actions to generate swallowing frequency data; The milk flow analysis unit is in communication with the image acquisition unit and is used to analyze the milk flow state through visual tracking technology, estimate the milk flow rate and flow rate, and generate flow characteristic data.

7. The breastfeeding intelligent guidance and milk composition analysis system according to claim 1, characterized in that: The data processing module includes: a data synchronization unit, configured to receive and synchronize the milk composition analysis result, the breastfeeding posture quality assessment data, and the auxiliary analysis data; a component-posture correlation analysis unit, communicatively connected to the data synchronization unit, for analyzing the correlation between milk components and breastfeeding posture quality, and establishing a component-posture-absorption efficiency correlation model; A time series trend analysis unit, in communication with the data synchronization unit, for analyzing the changing trend of milk components over time and predicting the direction of future changes; an abnormality detection unit, in communication with the data synchronization unit, for identifying milk components that deviate from a normal range and generating an abnormality marker and a risk level assessment; The cloud data processing unit is in communication with the data synchronization unit, the component-posture association analysis unit, the time series trend analysis unit and the anomaly detection unit, and is used to summarize the analysis results and compare them with the group data to generate a comprehensive analysis result.

8. The breastfeeding intelligent guidance and milk composition analysis system according to claim 1, characterized in that: The feeding plan generation module includes: a target setting unit, configured to receive and store a milk composition optimization target set by a user; a multi-objective optimization unit, communicatively connected to the target setting unit, for performing a multi-objective balance calculation based on the comprehensive analysis result and the milk composition optimization target, and generating an optimization strategy; a posture improvement suggestion unit, communicatively connected to the multi-objective optimization unit, for generating a breastfeeding posture improvement suggestion based on the optimization strategy; a feeding frequency suggestion unit, in communication with the multi-objective optimization unit, for generating feeding frequency and time schedule suggestions based on the optimization strategy; a sucking behavior guidance unit, communicatively connected to the multi-objective optimization unit, for generating guidance suggestions for the infant's sucking behavior based on the optimization strategy; The program integration unit is in communication with the posture improvement suggestion unit, the feeding frequency suggestion unit and the sucking behavior guidance unit, and is used to integrate various suggestions to form a complete personalized feeding program.

9. The breastfeeding intelligent guidance and milk composition analysis system according to claim 1, characterized in that: The remote guidance module includes: A data encryption transmission unit, used for encrypting the milk component analysis results and the comprehensive analysis results, and securely transmitting them to a remote professional terminal; a lactation consultant communication unit, communicatively connected to the data encryption transmission unit, for establishing a communication connection with the lactation consultant to receive breastfeeding posture correction suggestions; a physician diagnosis communication unit, communicatively connected to the data encryption transmission unit, for establishing a communication connection with the physician to receive a diagnosis and advice on abnormal milk composition; The professional advice integration unit is communicatively connected to the lactation consultant communication unit and the physician diagnosis communication unit, and is used to integrate the advice provided by professionals to form professional guidance advice.

10. The intelligent breastfeeding guidance and milk composition analysis system according to claim 1, characterized in that: The user interface module includes: A touch screen display unit, configured to present the personalized feeding plan and the professional guidance suggestions in the form of a graphical interface; a data visualization unit, communicatively connected to the touch screen display unit, for converting milk composition data, breastfeeding posture quality data, and time series change data into intuitive charts and trend lines; Voice announcement unit, used to provide real-time guidance and suggestions through voice; a posture demonstration unit, communicatively connected to the touch screen unit, for demonstrating the correct breastfeeding posture through three-dimensional animation; A user feedback collection unit, connected to the touch screen display unit for collecting user operation feedback and execution effect feedback; A feedback transmission unit is communicatively connected to the user feedback collection unit and is used to transmit user feedback information to the data processing module and the feeding plan generation module for plan optimization.

Images