Intelligent monitoring system for guiding breast feeding at home

Abstract

The invention discloses an intelligent monitoring system for guiding household breast feeding, and relates to the technical field of intelligent monitoring, and the system comprises a data acquisition module which is used for collecting the heart rate variability, the electrocardiogram signal and the mammary areola skin interstitial fluid hormone concentration of a mother in real time in the lactation process through an intelligent wearable device; the time sequence data, the sucking sound frequency spectrum and the force change time sequence data of the infant are aligned according to timestamps and then are output to the edge calculation module; by integrating advanced intelligent wearable equipment, multi-dimensional physiological data, including heart rate variability, electrocardiogram signals, mammary areola skin interstitial fluid hormone concentration, and sucking sound spectrum and strength change of an infant, of a mother and an infant in the lactation process can be synchronously collected, and the data are processed by a data preprocessing module and stored in a storage module; and key features are extracted and time-space alignment is carried out, so that a solid data basis is provided for subsequent dynamic modeling.

Description

Technical Field

[0001] The present invention relates to the field of intelligent monitoring technology, and in particular to an intelligent monitoring system for guiding breastfeeding at home. Background Art

[0002] With the development of modern medical technology, people's demand for maternal and infant health management is increasing, especially in the field of breastfeeding. Breastfeeding is the most ideal source of nutrition for infants, and its success is directly related to the growth and development of infants and the physical and mental health of mothers. However, during breastfeeding at home, mothers often find it difficult to accurately judge the baby's sucking status, their own physiological reactions, and the quality of interaction between mother and baby, which to a certain extent affects the effectiveness and sustainability of breastfeeding. Therefore, the development of an intelligent system that can monitor and guide breastfeeding at home in real time has become a research hotspot in the current field of medical technology.

[0003] The application number is: 202311118428.1, which is an intelligent monitoring system for guiding breastfeeding at home. It monitors and guides the baby's breastfeeding posture through the cooperation of the acquisition terminal, client, server and expert terminal. However, the comparison document is mainly based on video monitoring of breastfeeding posture, and its data collection dimensions do not cover the mother's heart rate variability, areola hormone concentration and infant sucking force time series data, and does not involve dynamic modeling of maternal and infant physiological synchronization and hormone trend inference. At the same time, the comparison document lacks an in-depth quantitative assessment of the mother-infant interaction status, and cannot construct a dynamic model to analyze the maternal and infant physiological synchronization and hormone change trends. There are also deficiencies in the personalization and dynamic optimization of intervention plans. It is difficult to adjust in real time according to multi-dimensional data and provide accurate personalized guidance, and no data closed loop has been formed to continuously optimize the strategy model.

[0004] Therefore, the development of an intelligent monitoring system for guiding breastfeeding at home has improved the quality of breastfeeding and provided strong support for scientific research and clinical practice of breastfeeding. Summary of the Invention

[0005] The purpose of the present invention is to make up for the shortcomings of the existing technology and provide an intelligent monitoring system for guiding breastfeeding at home. The system integrates smart wearable devices to synchronously collect multi-dimensional physiological data of mothers and babies during the breastfeeding process. After data preprocessing and dynamic modeling, the system can comprehensively reflect the mother-infant interaction status and calculate the mother-infant interaction index, which is intuitively displayed in a three-dimensional dynamic graph. When an abnormality is found, the system can generate an early warning signal in time and provide personalized guidance through multi-channel intervention.

[0006] To solve the above technical problems, the present invention provides the following technical solutions: an intelligent monitoring system for guiding breastfeeding at home, the system comprising:

[0007] Data acquisition module: This module uses smart wearable devices to collect real-time data on the mother's heart rate variability, electrocardiogram (ECG) signals, interstitial fluid hormone concentrations in the areola skin, and the infant's sucking sound spectrum and force change time series data during breastfeeding. The data is aligned by timestamp and output to the edge computing module.

[0008] Data preprocessing module: Bandpass filtering and wavelet denoising are performed on the collected raw data to extract the time-domain characteristics of HRV, the energy distribution of the sound frequency domain, and the sucking period and force peak characteristics of the pressure sensor data. The time-space alignment operator is then used to perform time-space joint mapping on the multimodal features and transmit them to the dynamic modeling module.

[0009] Dynamic Modeling Module: This module receives data from the Data Preprocessing Module, calculates the time-lag correlation between the sucking intensity change rate and the mother's HRV parameters, constructs a coupled oscillator model using the sucking rhythm and the mother's HRV dominant frequency as parameters to calculate synchronization, and uses the HRV-hormone dynamic mapping network to infer the mother's hormone trends based on the high-frequency components of HRV. The model then transmits the data to the Interaction Status Assessment Module.

[0010] Interaction Assessment Module: This module integrates sucking effectiveness, interaction time lag entropy, synchronization, and hormone inference data to calculate the mother-infant interaction index. It constructs a three-dimensional dynamic map based on time, sucking intensity, and maternal HRV characteristics. It uses color to encode synchronization, sets warning thresholds, and generates warning signals. The assessment results and warning signals are then sent to the intervention execution module and uploaded to the cloud.

[0011] Intervention execution module: Based on the evaluation results and early warning signals, the module selects corresponding intervention measures from the intervention measures library, implements multi-channel intervention through APP graphics, voice, and wearable device vibration feedback, optimizes intervention measures based on intervention effect data, and stores full-process data.

[0012] Furthermore, the smart wearable device in the data acquisition module is in the form of a wearable bra, integrating a PPG+ECG composite sensor, a high-sensitivity microphone array, a flexible pressure sensing matrix and a micro biochemical sensor.

[0013] Furthermore, the data preprocessing module performs time-space joint mapping on the multimodal data through a time-space alignment operator, and the formula is: in, It is the result of time-space joint mapping of multimodal data. P(t) is the infant sucking pressure signal, which is acquired in real time by the pressure sensing matrix of the smart wearable device. H(t) is the mother's HRV signal, which is acquired by the PPG+ECG composite sensor and obtained after preprocessing. t′ is the integral variable, τ is the dynamic time lag parameter, and the initial value is determined by calculating the preliminary correlation between the signals. w(t) is the Gaussian window function. The formula is: σ is the window width parameter, which is set according to the signal sampling rate, and t represents the current moment.

[0014] Furthermore, the dynamic modeling module quantifies the time-lag correlation between the sucking force change rate and the mother's HRV parameters through the interactive time-lag entropy index, and the formula is: Among them D TE is the interaction time-lag entropy index, which is used to quantify the time-lag correlation between the sucking force change rate and the mother's HRV parameters, s i is the infant sucking force feature vector, which is composed of the force peak value and sucking cycle features extracted from the pressure sensor data. j is the mother's HRV eigenvector, p(s i ,s j ) is the joint probability density, by s i and s j The kernel density estimation of historical data is obtained, p(s i )、p(s j ) is the marginal probability density, respectively for s i and s j The KDE estimation is performed separately, where n and m are the dimensions of the feature vector, which are determined by the number of features after signal preprocessing, and i and j are the summation indices.

[0015] Furthermore, the coupled oscillator model in the dynamic modeling module is constructed as follows: Where Ω represents the degree of maternal-infant physiological synchronization, which is used to measure the degree of synchronization between the infant's sucking rhythm and the mother's heart rate variability factor frequency. θ is a variable related to the phase trajectory of sucking and HRV, which is obtained by integrating and converting the infant's real-time sucking frequency and the mother's HRV factor frequency. s (t) is the real-time sucking frequency of the infant, which is calculated from the inverse of the sucking cycle of the pressure sensor data, and f h (t) is the dominant frequency of the mother's HRV. By performing power spectrum density analysis on the HRV signal, the center frequency of the frequency band with the highest energy proportion is extracted, and θ s (t), θ h (t) is the phase trajectory of sucking and HRV, which is converted into phase value by integrating the frequency and multiplying by 2π. It is the phase derivative, reflecting the real-time change of frequency, t′ is the integral variable, t represents the current moment, and dθ is the phase derivative, reflecting the real-time change of frequency.

[0016] Furthermore, the dynamic modeling module uses the HRV-hormone dynamic mapping network to infer the mother's hormone trend based on the HRV high-frequency component. The formula is: where C(t) is the inferred cortisol concentration, l is the index of the network layer, L is the number of network layers, and H HF (t) is the high-frequency component of the mother's HRV, which is extracted by bandpass filtering the HRV signal and calculating the power spectral density. K is the time delay order, k is the time delay index, Δt is the time step, and h k is the input layer weight, which is obtained through supervised learning training of historical HRV data and synchronously collected cortisol blood samples, H HF is the high-frequency component of the mother's HRV, which is extracted by bandpass filtering the HRV signal and calculating the power spectrum density. t represents the current moment, and w l is the weight matrix of each layer, φ l is the activation function and r is the normalization function.

[0017] Furthermore, the hidden layer of the HRV-hormone dynamic mapping network in the bidirectional dynamic feedback modeling module adopts a fully connected neural network structure, and the activation function is one or more combinations of ReLU, Sigmoid or Tanh functions. The input layer of the network receives the time delay sequence of the HRV high-frequency component, the delay order is 5-15, and the time step is 0.5-2 seconds.

[0018] Furthermore, the interaction evaluation module calculates the mother-infant interaction index through a nonlinear fusion evaluation algorithm, and the formula is: MBII = α·SUC + β·(1-D TE )+γ·Ω+δ·(1-|C(t)-C0|), where MBII is the mother-infant interaction index, SUC is the sucking effectiveness, which is determined by combining sound spectrum analysis with pressure sensor data. When sucking is effective, SUC=1, and when it is ineffective, SUC=0. D TE is the interaction time-lag entropy, Ω is the maternal-infant physiological synchronization, C(t) is the inferred cortisol concentration, C0 is the baseline cortisol concentration, α, β, γ, and δ are dynamic weight coefficients, which are adaptively adjusted by the machine learning model, and α + β + γ + δ = 1.

[0019] Furthermore, the interactive evaluation module makes an early warning judgment, when MBII<θ1 or Ω<θ2 or D TE >θ3 triggers the warning, where θ1, θ2, θ3 are MBII, Ω, D TE The intervention threshold is set by historical data and adaptively updated based on the reinforcement learning mechanism.TE is the interaction time lag entropy index, Ω is the maternal-infant physiological synchronization, and MBII is the maternal-infant interaction index.

[0020] Furthermore, the three-dimensional dynamic graph in the interactive evaluation module uses time as the X-axis, sucking strength as the Y-axis, and mother's HRV value as the Z-axis, and adopts OpenGL technology for real-time rendering.

[0021] Compared with the existing technology, the intelligent monitoring system for guiding home breastfeeding has the following advantages:

[0022] Beneficial effects:

[0023] 1. By integrating advanced smart wearable devices, the present invention can synchronously collect multi-dimensional physiological data of mothers and babies during the breastfeeding process, including heart rate variability, electrocardiogram signals, areola skin interstitial fluid hormone concentrations, and the baby's sucking sound spectrum and force changes. These data are processed by the data preprocessing module, key features are extracted and time-space alignment is performed, providing a solid data foundation for subsequent dynamic modeling. The dynamic modeling module not only calculates the time-lag correlation between the sucking force change rate and the mother's HRV parameters, but also constructs a coupled oscillator model to quantify the physiological synchronization between mother and baby. By using the HRV-hormone dynamic mapping network to infer the mother's hormone trends, the system can more comprehensively understand the state of mother-infant interaction, providing a scientific basis for the formulation of personalized intervention plans.

[0024] 2. The present invention calculates the mother-infant interaction index by integrating sucking effectiveness, interaction time lag entropy, mother-infant physiological synchronization and hormone inference data, and displays it intuitively in the form of a three-dimensional dynamic graph, so that the mother-infant interaction status is clear at a glance. More importantly, the system sets an early warning threshold. When the mother-infant interaction index is abnormal, it can generate an early warning signal in time and implement multi-channel intervention through APP graphics, voice and wearable device vibration feedback. This intelligent intervention mechanism not only improves the timeliness and effectiveness of the intervention, but also continuously optimizes the strategy model by uploading the intervention effect data to the cloud, and realizes the dynamic adjustment of personalized intervention plans, which not only improves the user experience, but also provides strong technical support for scientific research on breastfeeding.

[0025] Other advantages, objects and features of the present invention will be described in part in the following description and, in part, will be apparent to those skilled in the art based on an examination of the following or may be learned from the practice of the invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. Those skilled in the art can also derive other drawings based on these drawings without inventive effort.

[0027] Figure 1 is a flow chart of an intelligent monitoring system for guiding breastfeeding at home;

[0028] Figure 2 A framework diagram of an intelligent monitoring system for guiding breastfeeding at home. DETAILED DESCRIPTION

[0029] In order to further illustrate the technical means and effects adopted by the present invention to achieve the predetermined purpose of the invention, the specific implementation methods, structures, features and effects of the present invention are described in detail below in conjunction with the accompanying drawings and preferred embodiments.

[0030] Example 1:

[0031] Home breastfeeding support scenario for mothers of premature infants.

[0032] A mother of a premature baby breastfeeds her low-birth-weight baby at home. The baby's sucking ability is weak, and the mother hopes to use the system to monitor the breastfeeding efficiency and adjust the feeding strategy. The intelligent monitoring system continuously collects physiological data of mother and baby, focusing on the baby's sucking strength, effective sucking duration, and the impact of the mother's hormone levels on lactation.

[0033] Data acquisition module: The smart wearable device integrates multiple types of sensors in the form of a wearable bra, such as Figure 2 As shown in the figure, in view of the weak sucking force of premature infants, the flexible pressure sensing matrix adopts a high-density sensing node design to record the time series data of force changes in real time; the high-sensitivity microphone array has the function of environmental noise reduction, accurately collects the swallowing sound and intermittent breathing sound spectrum characteristics of the baby during sucking, and distinguishes the acoustic differences between effective sucking and ineffective sucking; the PPG+ECG composite sensor synchronously monitors the mother's heart rate variability signal and electrocardiogram waveform, and the micro-biochemical sensor detects the cortisol and oxytocin hormone concentrations in the interstitial fluid of the areola skin in real time in a non-invasive way. All physiological data are marked with millisecond timestamps to ensure the time consistency of multimodal data, and then output to the edge computing module for preliminary processing, such as Figure 1 shown.

[0034] Data preprocessing module: Performs multi-level preprocessing on the collected raw data: First, bandpass filtering is used to remove high-frequency interference from environmental noise and physiological signals, and then a wavelet denoising algorithm is used to further purify the mother's HRV signal and the infant's sucking pressure data. In the feature extraction step, the RMSSD time domain features are extracted from the HRV signal, the frequency domain energy distribution pattern is analyzed from the sound data, and the parameters of the sucking cycle and peak force are identified from the pressure data. The multimodal features are then jointly mapped in time and space using a spatiotemporal alignment operator to eliminate the timing deviation of different sensor data. The formula is: in, It is the result of time-space joint mapping of multimodal data. P(t) is the infant sucking pressure signal, which is acquired in real time by the pressure sensing matrix of the smart wearable device. H(t) is the mother's HRV signal, which is acquired by the PPG+ECG composite sensor and obtained after preprocessing. t′ is the integral variable, τ is the dynamic time lag parameter, and the initial value is determined by calculating the preliminary correlation between the signals. w(t) is the Gaussian window function. The formula is: σ is the window width parameter, which is set according to the signal sampling rate. t represents the current moment to ensure the accuracy of subsequent analysis. The processed standardized feature data is transmitted to the dynamic modeling module.

[0035] Dynamic modeling module: To evaluate the correlation between premature infants' sucking and maternal physiological responses, the interaction time-lag entropy index was used to quantify the time-lag correlation between the sucking intensity change rate and maternal HRV parameters. The formula is: Among them D TE is the interaction time-lag entropy index, which is used to quantify the time-lag correlation between the sucking force change rate and the mother's HRV parameters, s i is the infant sucking force feature vector, which is composed of the force peak value and sucking cycle features extracted from the pressure sensor data. j is the mother's HRV eigenvector, p(s i ,s j ) is the joint probability density, by s i and s j The kernel density estimation of historical data is obtained, p(s i )、p(s j ) is the marginal probability density, respectively for s i and s j KDE estimation is performed separately, where n and m are the dimensions of the feature vector, which are determined by the number of features after signal preprocessing, and i and j are summation indices to analyze the dynamic response relationship between the two in the time series;

[0036] Based on the infant's real-time sucking frequency and the mother's HRV dominant frequency, a coupled oscillator model was constructed to calculate the maternal-infant physiological synchronization. The formula is: Where Ω represents the degree of maternal-infant physiological synchronization, which is used to measure the degree of synchronization between the infant's sucking rhythm and the mother's heart rate variability factor frequency. θ is a variable related to the phase trajectory of sucking and HRV, which is obtained by integrating and converting the infant's real-time sucking frequency and the mother's HRV factor frequency. s (t) is the real-time sucking frequency of the infant, which is calculated from the inverse of the sucking cycle of the pressure sensor data, and f h (t) is the dominant frequency of the mother's HRV. By performing power spectrum density analysis on the HRV signal, the center frequency of the frequency band with the highest energy proportion is extracted, and θ s (t), θ h (t) is the phase trajectory of sucking and HRV, which is converted into phase value by integrating the frequency and multiplying by 2π. is the phase derivative, reflecting the real-time changes in frequency, t′ is the integral variable, t represents the current moment, dθ is the phase derivative, reflecting the real-time changes in frequency, capturing the coordination between the irregular sucking rhythm of premature infants and the physiological fluctuations of the mother;

[0037] Considering that mothers of premature infants often have anxiety that affects lactation, we used the HRV-hormone dynamic mapping network to infer the cortisol concentration trend based on the high-frequency components of the mother's HRV. Combined with historical hormone level data, we evaluated the potential impact of stress on lactation using the formula: where C(t) is the inferred cortisol concentration, l is the index of the network layer, L is the number of network layers, and H HF (t) is the high-frequency component of the mother's HRV, which is extracted by bandpass filtering the HRV signal and calculating the power spectral density. K is the time delay order, k is the time delay index, Δt is the time step, and h k is the input layer weight, which is obtained through supervised learning training of historical HRV data and synchronously collected cortisol blood samples, H HF is the high-frequency component of the mother's HRV, which is extracted by bandpass filtering the HRV signal and calculating the power spectrum density. t represents the current moment, and w l is the weight matrix of each layer, φ l is the activation function, r is the normalization function, and the modeling results are synchronously transmitted to the interactive evaluation module for comprehensive analysis.

[0038] Interaction evaluation module: In view of the particularity of premature infants' breastfeeding, multi-dimensional evaluation indicators are integrated: the sucking effectiveness is determined by the joint analysis of sound spectrum and pressure sensor data (for example, it is determined to be effective when the force continuously exceeds the preset threshold and is accompanied by regular swallowing sounds). Combined with the interaction time lag entropy, maternal and infant physiological synchronization and hormone inference data, a nonlinear fusion evaluation algorithm is used to calculate the mother-infant interaction index (MBII). The formula is: MBII = α·SUC + β·(1-D TE)+γ·Ω+δ·(1-|C(t)-C0|), where MBII is the mother-infant interaction index, SUC is the sucking effectiveness, which is determined by combining sound spectrum analysis with pressure sensor data. When sucking is effective, SUC=1, and when it is ineffective, SUC=0. D TE is the interaction time lag entropy, Ω is the maternal and infant physiological synchronization, C(t) is the inferred cortisol concentration, C0 is the baseline cortisol concentration, α, β, γ, δ are dynamic weight coefficients, which are adaptively adjusted by the machine learning model, and α+β+γ+δ=1. A three-dimensional dynamic map is constructed with time, sucking intensity, and maternal HRV characteristics as coordinates. The three-dimensional dynamic map has time as the X-axis, sucking intensity as the Y-axis, and maternal RMSSD value as the Z-axis. It is rendered in real time using OpenGL technology and intuitively displays synchronization changes through color gradients. The system presets multi-dimensional warning thresholds. When MBII<θ1 or Ω<θ2 or D TE >θ3 triggers the warning, where θ1, θ2, θ3 are MBII, Ω, D TE The intervention threshold is set by historical data and adaptively updated based on the reinforcement learning mechanism. TE is the interaction time lag entropy indicator, Ω represents the maternal and infant physiological synchronization, and MBII is the maternal and infant interaction index. When an abnormality is detected, an early warning signal is automatically generated, and the evaluation results and early warning information are simultaneously pushed to the intervention execution module and uploaded to the cloud storage for long-term trend analysis.

[0039] Intervention execution module: After receiving the evaluation results, the system matches personalized plans from the intervention measures library based on common problems of premature infants in breastfeeding (such as insufficient effective sucking time and elevated maternal stress hormones): If it is detected that the infant's sucking strength is continuously low, the APP will push graphic instructions for the "intermittent stimulation breastfeeding method", and the techniques will be explained in voice. At the same time, the bra vibration feedback will help the mother to grasp the timing of stimulation; if the mother's cortisol concentration is abnormally high, it will trigger relaxation training intervention, and progressive muscle relaxation audio will be played through the wearable device, and synchronous vibration will prompt breathing rhythm; during the intervention process, the improvement of the infant's sucking efficiency and the improvement of the mother's HRV parameters will be recorded in real time, the intervention measures library will be optimized, and the strategy will be dynamically adjusted according to the growth stage of the premature infant (such as switching to the regular breastfeeding assessment mode after the weight reaches the standard), providing continuous and dynamic scientific guidance for breastfeeding of premature infants.

[0040] To summarize, this embodiment is aimed at the scenario of mothers of premature babies breastfeeding at home. The system uses high-sensitivity sensors to capture the weak sucking data of premature babies and the physiological indicators of the mothers. After preprocessing and dynamic modeling, the interactive time-delay entropy and coupled oscillator model analysis are used to evaluate the mother-child interaction status, set warning thresholds based on the breastfeeding characteristics of premature babies, and the intervention module provides personalized solutions, such as feeding skill guidance and maternal relaxation training. Combined with the effect optimization strategy, a closed-loop support system is formed to scientifically solve the problems of premature baby breastfeeding efficiency and maternal lactation, and help premature babies develop healthily.

[0041] Example 2:

[0042] A new mother breastfeeding at home.

[0043] A new mother, two weeks postpartum, was breastfeeding at home. Due to her lack of experience, she was often worried about whether the baby was sucking effectively and whether her own breastfeeding status was normal. The intelligent monitoring system integrated sensors in the form of a wearable bra to collect the following data in real time during the breastfeeding process: the baby's sucking sound spectrum, the time series data of the sucking force changes, as well as the mother's heart rate variability, electrocardiogram signals and areola skin interstitial fluid hormone concentration.

[0044] Data acquisition module: The smart wearable device integrates a PPG+ECG composite sensor, a high-sensitivity microphone array, a flexible pressure sensing matrix, and a micro-biochemical sensor in the form of a wearable bra. During the breastfeeding process, it synchronously collects: the infant's sucking sound spectrum, the time series data of sucking force changes, the mother's heart rate variability, electrocardiogram signals, and areola skin interstitial fluid hormone concentration data. All data are marked with a timestamp and output to the edge computing module.

[0045] Data preprocessing module: Bandpass filtering and wavelet denoising are performed on the collected raw data to extract HRV time domain features, sound frequency domain energy distribution, sucking cycle and force peak features of pressure sensor data; and the time-space alignment operator is used to perform time-space joint mapping of multimodal features. The formula is: The processed data is transferred to the dynamic modeling module.

[0046] Dynamic modeling module: Receives data transmitted by the data preprocessing module and calculates the time-lag correlation between the sucking force change rate and the mother's HRV parameters. The formula is: A coupled oscillator model was constructed to calculate synchronization using the sucking rhythm and the dominant frequency of the mother's HRV as parameters. The formula is: Using the HRV-hormone dynamic mapping network, the maternal hormone trend is inferred based on the high-frequency components of HRV. The formula is: The analysis results are transmitted to the interaction status evaluation module.

[0047] Interaction Assessment Module: Integrates sucking effectiveness (determined by sound spectrum and pressure sensor data), interaction time lag entropy, synchronization, and hormone inference data to calculate the mother-infant interaction index (MBII). The formula is: MBII = α·SUC + β·(1-D TE)+γ·Ω+δ·(1-|C(t)-C0|); a three-dimensional dynamic map was constructed with time, sucking intensity, and mother's HRV characteristics as coordinates, synchronization was coded with colors, and warning thresholds were set to generate warning signals. The evaluation results and warning signals were pushed to the intervention execution module and uploaded to the cloud.

[0048] Intervention execution module: Based on the evaluation results and early warning signals, select corresponding intervention measures from the intervention measures library (such as APP graphic guidance on breastfeeding posture adjustment, voice prompts on relaxation techniques, or wearable device vibration feedback reminders) and implement multi-channel intervention; optimize intervention measures based on intervention effect data, and store full process data in the cloud.

[0049] To sum up, in this embodiment, the intelligent monitoring system collects multi-dimensional physiological data of mother and baby through various sensors integrated in the wearable bra in the scenario of new mothers breastfeeding at home. After preprocessing, it uses the interactive time-lag entropy index, coupled oscillator model and HRV-hormone dynamic mapping network for analysis, calculates the mother-baby interaction index and generates a three-dimensional map to achieve abnormal warning. The system implements intervention through multiple channels and combines the effect data optimization strategy to form a closed loop, providing scientific guidance for new mothers, solving their difficulties in judging the breastfeeding status, and improving the quality of breastfeeding.

[0050] The above description is merely a preferred embodiment of the present invention and does not constitute any form of limitation to the present invention. Although the present invention has been disclosed as above in terms of a preferred embodiment, it is not intended to limit the present invention. Any person skilled in the art can, without departing from the scope of the technical solution of the present invention, make some changes or modifications to equivalent embodiments using the technical contents disclosed above. However, any brief modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention are still within the scope of the technical solution of the present invention.

Claims

1. An intelligent monitoring system for guiding breastfeeding at home, characterized in that: The system includes: Data acquisition module: This module uses smart wearable devices to collect real-time data on the mother's heart rate variability, electrocardiogram (ECG) signals, interstitial fluid hormone concentrations in the areola skin, and the infant's sucking sound spectrum and force change time series data during breastfeeding. The data is aligned by timestamp and output to the edge computing module. Data preprocessing module: Bandpass filtering and wavelet denoising are performed on the collected raw data to extract the time-domain characteristics of HRV, the energy distribution of the sound frequency domain, and the sucking period and force peak characteristics of the pressure sensor data. The time-space alignment operator is then used to perform time-space joint mapping on the multimodal features and transmit them to the dynamic modeling module. Dynamic Modeling Module: This module receives data from the Data Preprocessing Module, calculates the time-lag correlation between the sucking intensity change rate and the mother's HRV parameters, constructs a coupled oscillator model using the sucking rhythm and the mother's HRV dominant frequency as parameters to calculate synchronization, and uses the HRV-hormone dynamic mapping network to infer the mother's hormone trends based on the high-frequency components of HRV. The model then transmits the data to the Interaction Status Assessment Module. Interaction Assessment Module: This module integrates sucking effectiveness, interaction time lag entropy, synchronization, and hormone inference data to calculate the mother-infant interaction index. It constructs a three-dimensional dynamic map based on time, sucking intensity, and maternal HRV characteristics. It uses color to encode synchronization, sets warning thresholds, and generates warning signals. The assessment results and warning signals are then sent to the intervention execution module and uploaded to the cloud. Intervention execution module: Based on the evaluation results and early warning signals, the module selects corresponding intervention measures from the intervention measures library, implements multi-channel intervention through APP graphics, voice, and wearable device vibration feedback, optimizes intervention measures based on intervention effect data, and stores full-process data.

2. The intelligent monitoring system for guiding breastfeeding at home according to claim 1, characterized in that: The smart wearable device in the data acquisition module is in the form of a wearable bra, integrating a PPG+ECG composite sensor, a high-sensitivity microphone array, a flexible pressure sensing matrix and a micro biochemical sensor.

3. The intelligent monitoring system for guiding breastfeeding at home according to claim 1, characterized in that: In the data preprocessing module, the time-space joint mapping of multimodal data is performed through the time-space alignment operator, and the formula is: in, It is the result of time-space joint mapping of multimodal data. P(t) is the infant sucking pressure signal, which is acquired in real time by the pressure sensing matrix of the smart wearable device. H(t) is the mother's HRV signal, which is acquired by the PPG+ECG composite sensor and obtained after preprocessing. t′ is the integral variable, τ is the dynamic time lag parameter, and the initial value is determined by calculating the preliminary correlation between the signals. w(t) is the Gaussian window function. The formula is: σ is the window width parameter, which is set according to the signal sampling rate, and t represents the current moment.

4. The intelligent monitoring system for guiding breastfeeding at home according to claim 1, characterized in that: The dynamic modeling module quantifies the time-lag correlation between the sucking force change rate and the mother's HRV parameters through the interactive time-lag entropy index, and the formula is: Among them D TE is the interaction time-lag entropy index, which is used to quantify the time-lag correlation between the sucking force change rate and the mother's HRV parameters, s i is the infant sucking force feature vector, which is composed of the force peak value and sucking cycle features extracted from the pressure sensor data. j is the mother's HRV eigenvector, p(s i ,s j ) is the joint probability density, by s i and s j The kernel density estimation of historical data is obtained, p(s i )、p(s j ) is the marginal probability density, respectively for s i and s j The KDE estimation is performed separately, where n and m are the dimensions of the feature vector, which are determined by the number of features after signal preprocessing, and i and j are the summation indices.

5. The intelligent monitoring system for guiding breastfeeding at home according to claim 1, characterized in that: The coupled oscillator model in the dynamic modeling module is constructed as follows: Where Ω represents the degree of maternal-infant physiological synchronization, which is used to measure the degree of synchronization between the infant's sucking rhythm and the mother's heart rate variability factor frequency. θ is a variable related to the phase trajectory of sucking and HRV, which is obtained by integrating and converting the infant's real-time sucking frequency and the mother's HRV factor frequency. s (t) is the real-time sucking frequency of the infant, which is calculated from the inverse of the sucking cycle of the pressure sensor data, and f h (t) is the dominant frequency of the mother's HRV. By performing power spectrum density analysis on the HRV signal, the center frequency of the frequency band with the highest energy proportion is extracted, and θ s (t), θ h (t) is the phase trajectory of sucking and HRV, which is converted into phase value by integrating the frequency and multiplying by 2π. It is the phase derivative, reflecting the real-time change of frequency, t′ is the integral variable, t represents the current moment, and dθ is the phase derivative, reflecting the real-time change of frequency.

6. The intelligent monitoring system for guiding breastfeeding at home according to claim 1, characterized in that: The dynamic modeling module uses the HRV-hormone dynamic mapping network to infer the mother's hormone trend based on the HRV high-frequency component. The formula is: where C(t) is the inferred cortisol concentration, l is the index of the network layer, L is the number of network layers, and H HF (t) is the high-frequency component of the mother's HRV, which is extracted by bandpass filtering the HRV signal and calculating the power spectral density. K is the time delay order, k is the time delay index, Δt is the time step, and h k is the input layer weight, which is obtained through supervised learning training of historical HRV data and synchronously collected cortisol blood samples, H HF is the high-frequency component of the mother's HRV, which is extracted by bandpass filtering the HRV signal and calculating the power spectrum density. t represents the current moment, and w l is the weight matrix of each layer, φ l is the activation function and r is the normalization function.

7. The intelligent monitoring system for guiding breastfeeding at home according to claim 6, characterized in that: The hidden layer of the HRV-hormone dynamic mapping network in the bidirectional dynamic feedback modeling module adopts a fully connected neural network structure, and the activation function is one or more combinations of ReLU, Sigmoid or Tanh functions. The input layer of the network receives the time delay sequence of the HRV high-frequency component, with a delay order of 5-15 and a time step of 0.5-2 seconds.

8. The intelligent monitoring system for guiding breastfeeding at home according to claim 6, characterized in that: The interaction evaluation module calculates the mother-infant interaction index through a nonlinear fusion evaluation algorithm. The formula is: MBII = α·SUC + β·(1-D TE )+γ·Ω+δ·(1-|C(t)-C0|), where MBII is the mother-infant interaction index, SUC is the sucking effectiveness, which is determined by combining sound spectrum analysis with pressure sensor data. When sucking is effective, SUC=1, and when it is ineffective, SUC=0. D TE is the interaction time-lag entropy, Ω is the maternal-infant physiological synchronization, C(t) is the inferred cortisol concentration, C0 is the baseline cortisol concentration, α, β, γ, and δ are dynamic weight coefficients, which are adaptively adjusted by the machine learning model, and α + β + γ + δ = 1.

9. The intelligent monitoring system for guiding breastfeeding at home according to claim 8, characterized in that: In the interactive evaluation module, the early warning judgment is made when MBII<θ1 or Ω<θ2 or D TE >θ3 triggers the warning, where θ1, θ2, θ3 are MBII, Ω, D TE The intervention threshold is set by historical data and adaptively updated based on the reinforcement learning mechanism. TE is the interaction time lag entropy index, Ω is the maternal-infant physiological synchronization, and MBII is the maternal-infant interaction index.

10. The intelligent monitoring system for guiding breastfeeding at home according to claim 1, characterized in that: The three-dimensional dynamic graph in the interactive evaluation module uses time as the X-axis, sucking strength as the Y-axis, and mother's HRV value as the Z-axis, and adopts OpenGL technology for real-time rendering.

Images