Registration environment adjusting system for old people

Abstract

The invention discloses an old-age health-care environment adjusting system which specifically comprises an environment data collecting and processing unit, a physiological data monitoring module, a behavior detection module and a space-time cube analysis unit, wherein the space-time cube analysis unit is used for obtaining a space-time cube with a space-time label; the feature engineering unit is used for obtaining environment-behavior-physiological feature data; the random forest-Bayesian polynomial regression-particle swarm modeling and optimizing unit is used for establishing an interaction effect model used for representing the interaction effect between different modal data by utilizing a random forest and Bayesian polynomial algorithm and the environment-behavior-physiological feature data; performing optimization according to the interaction effect model, the set constraint condition and the optimization target to obtain an optimal environment control parameter; and adjusting the environment equipment based on the optimal environment control parameters so as to provide the optimal environment state for the user under the condition of comprehensively considering the association among the environment data, the physiology data and the behavior data.

Description

Technical Field

[0001] The present invention relates to the field of intelligent elderly care technology, and in particular to an elderly health care environment adjustment system. Background Art

[0002] The "Opinions of the General Office of the State Council on Promoting the Development of Elderly Care Services" mentions the need to promote the implementation of the "Internet + Elderly Care" initiative and drive the development of a smart and healthy elderly care industry. IoT technology, as a key enabler, is profoundly changing the service model for elderly care. For example, its application in elderly care environment regulation creates safe, comfortable, and healthy living conditions for the elderly by optimizing physical environmental factors, encompassing multi-dimensional controls such as temperature, humidity, air quality, lighting, and noise. Taking elderly care institutions as an example, IoT technology has enabled intelligent control of the home environment. Through big data analysis, it deeply mines multi-dimensional data collected by sensors to create a comfortable elderly care environment.

[0003] Existing technology, such as CN201810054283.6, discloses an indoor environmental control system and method for improving the quality of life for elderly people living at home. Leveraging the cloud computing data center of a community social service center, this system promotes a cloud computing and big data-based informatization model for elderly care services. This system includes: leveraging the cloud computing data center and module development to provide elderly people living at home with indoor power control and management modules, indoor temperature and humidity control and management modules, and indoor gas control and management modules.

[0004] CN201410502607.X discloses a health status monitoring system for the elderly and its software. The system includes an environmental monitoring device, a personal status monitoring device, an alarm system, a monitoring data storage and analysis server, and a monitoring terminal. Compared with existing home environment monitoring systems, this patent's core advantage lies in that the monitor not only includes an environmental monitor but also includes devices for monitoring personal status, such as instruments for measuring and monitoring body temperature, movement behavior, blood pressure, blood sugar, electrocardiogram, blood oxygen, heart rate, and body weight composition. These monitors measure various human parameters and send them to a server. The server analyzes this data, determines the health status, generates a report, and sends it to each terminal via a network. This allows real-time monitoring of the elderly's health status, reducing the occurrence of emergencies.

[0005] However, the above technologies do not comprehensively consider the relationship between environmental, physiological and behavioral data. Instead, they process the data individually and use the data separately for subsequent processing. They are unable to combine the data to comprehensively evaluate the user's health and wellness environment and make adjustments, resulting in the inability to provide users with the optimal environmental state. Summary of the Invention

[0006] The present invention provides a health-care environment adjustment system for the elderly to overcome the technical problem that the current indoor environment control method fails to comprehensively consider the relationship between environmental, physiological and behavioral data, but only processes the data in a single way and uses the data for subsequent processing separately. It is unable to combine the data to comprehensively evaluate the user's health-care environment and thus make adjustments, resulting in an inability to provide the user with the optimal environmental state.

[0007] In order to achieve the above object, the technical solution of the present invention is:

[0008] An elderly health care environment adjustment system, comprising:

[0009] An environmental data acquisition and processing unit, configured to acquire environmental data and process the environmental data to obtain processed environmental data;

[0010] A physiological data monitoring module is used to obtain the user's physiological data, including heart rate variability, blood oxygen saturation and body movement frequency data;

[0011] A behavior detection module is used to obtain user behavior data, including activity trajectory, sitting and lying time, and device usage frequency data;

[0012] A space-time cube analysis unit is used to segment the activity trajectory and generate an indoor stay heat map using the segmented activity trajectory and a kernel density estimation method to obtain a space-time cube with space-time labels;

[0013] The feature engineering unit is used to extract features based on the processed environmental data, user physiological data, and the spatiotemporal cube with spatiotemporal labels to obtain environmental-behavioral-physiological feature data;

[0014] A random forest-Bayesian polynomial regression-particle swarm modeling and optimization unit is used to establish an interaction effect model for characterizing the interaction between different modal data using random forest and Bayesian polynomial algorithms and environmental-behavioral-physiological characteristic data, and to optimize the interaction effect model and the set constraints and optimization objectives to obtain the optimal environmental control parameters;

[0015] A health assessment unit based on digital twins is used to simulate the optimal environmental control parameters in a virtual environment to determine whether the optimal environmental control parameters meet the set health indicators and energy consumption targets;

[0016] The environmental equipment control unit is used to adjust the environmental equipment according to the environmental control parameters that meet the set health indicators and energy consumption targets.

[0017] Furthermore, the environmental data acquisition and processing unit includes:

[0018] Environmental data acquisition module, used to obtain environmental data, including temperature, humidity, PM2.5, CO2 concentration, volatile organic compounds (VOCs), and light intensity data;

[0019] A sliding window and normality detection unit is used to obtain data from the environmental data acquisition module, perform normality detection on the obtained data, and process the obtained data according to the normality detection result and the set processing rules to obtain normalized environmental data;

[0020] The environmental outlier detection unit is used to perform an outlier elimination operation on the normalized environmental data to obtain processed environmental data.

[0021] Furthermore, the sliding window and normality detection unit acquires data from the environmental data acquisition module, performs normality detection on the acquired data, and processes the acquired data according to the normality detection result and the set processing rules to obtain normalized environmental data. The specific steps include:

[0022] Establishing a dual time window for normality testing and data transformation, wherein the dual time window includes a short window and a long window;

[0023] The short window is used to obtain data from the environmental data acquisition module according to the set time and frequency;

[0024] The long window is used to obtain data from the environmental data acquisition module every 24 hours;

[0025] The data obtained in the long window are used to perform normality test on the data in the short window using the improved Shapiro–Wilk test method. The normality test formula is:

[0026]

[0027] Where W′ is the improved Shapiro-Wilk test statistic, which is used to dynamically adjust the data normality test results; W is the original Shapiro-Wilk test statistic, which reflects the degree of fit between the data and the normal distribution; ΔT 昼夜 represents the difference between the highest temperature and the lowest temperature within 24 hours obtained based on the long window, which is used to correct the test statistic to adapt to the change of ambient temperature;

[0028] Determine whether the normality test result satisfies the normal distribution. If so, all the data in the dual time window are used as normalized environmental data. Otherwise, the data in the short window are processed using Box-Cox transformation or logarithmic transformation, and the data in the long window and the processed data in the short window are used as normalized environmental data.

[0029] Furthermore, the environmental outlier detection unit performs an outlier elimination operation on the normalized environmental data, including:

[0030] 1) Construct a boxplot of normalized environmental data based on the quantile regression method, including:

[0031] Calculate the lower quartile Q1 and upper quartile Q3 of each time period, and determine the interquartile range IQR;

[0032] The dynamic upper and lower bounds are set based on the lower quartile Q1, upper quartile Q3 and interquartile range IQR, respectively:

[0033] The upper bound is expressed as:

[0034] Q3+k'×IQR

[0035] The lower bound is expressed as:

[0036] Q1-k'×IQR

[0037] In the formula, k' is the set threshold parameter;

[0038] Compare the normalized environmental data with the upper and lower bounds in turn, and mark the data that are not within the upper and lower bounds;

[0039] 2) Set temperature warning rules, including:

[0040] If the temperature difference between adjacent moments in a certain area exceeds the set difference threshold, the temperature of that period is marked as candidate abnormal data;

[0041] 3) Use the normalized environmental data to construct an isolation forest model and calculate the anomaly score of each type of data in the normalized environmental data. The calculation formula for the anomaly score is as follows:

[0042] Abnormal score = 0.7 × IF (X 正态化环境数据 )+0.3×LSTM(X 行为序列 )

[0043] Where IF represents isolation forest anomaly detection, and LSTM represents temporal pattern analysis of the corresponding behavior sequence;

[0044] 4) Calculate the average anomaly score value st based on the anomaly score of each type of data in the normalized environmental data, st∈(0,1];

[0045] 5) Based on the marking results of the box plot and temperature warning rules, set the rejection threshold of the corresponding type of data in the normalized environmental data, and reject abnormal temperature, humidity, PM2.5, CO2 concentration, volatile organic compounds (VOCs), and light intensity data according to the rejection threshold.

[0046] Furthermore, the spatiotemporal cube analysis unit segments the activity trajectories and uses the segmented activity trajectories and kernel density estimation method to generate indoor stay heat maps. The specific steps for obtaining the behavior analysis results with spatiotemporal labels include:

[0047] Calculate the speed of two adjacent points in the activity trajectory;

[0048] According to the speed of two adjacent points in the activity trajectory, the speed mutation point is identified according to the set speed threshold;

[0049] Segment the activity trajectory according to the speed mutation point to obtain the segmented activity trajectory;

[0050] Map the segmented activity trajectory to the indoor floor plan to obtain the mapped coordinate points;

[0051] The mapped coordinate points are calculated using a kernel density estimation method to generate a stay heat map reflecting the user's stay or high-frequency activity areas. The stay heat map is a space-time cube with space-time labels.

[0052] Furthermore, the feature engineering unit performs feature extraction based on the processed environmental data, user physiological data, and the spatiotemporal cube with spatiotemporal labels to obtain environmental-behavioral-physiological feature data including time domain features, frequency domain features, spatial features, and health indicators. The specific steps include:

[0053] 1) Extract time domain features:

[0054] Calculate the mean and variance of each type of data in the processed environmental data respectively;

[0055] Calculate the mean and variance of the user's physiological data respectively;

[0056] 2) Extract frequency domain features:

[0057] Perform wavelet transform on the processed environmental data and user physiological data, and calculate the energy distribution of the transformed data in each frequency band;

[0058] Energy entropy is extracted based on the energy distribution of the transformed data in each frequency band, and the energy entropy is used as a frequency domain feature;

[0059] 3) Extract spatial features:

[0060] Mapping the temperature field of each room or functional area in the user's environment onto the space-time cube with the space-time label;

[0061] Obtaining the temperature difference between adjacent rooms or functional areas and calculating the average value of the temperature difference in the time dimension to obtain the temperature gradient between different rooms or functional areas, and using the temperature gradient as a spatial feature;

[0062] 4) Calculate health indicators, including:

[0063] Calculate the temperature fluctuation rate using the following formula:

[0064]

[0065] Where, max(T 1h ) is the maximum temperature per unit time, min(T 1h ) is the minimum temperature per unit time; mean(T 1h ) is the average temperature per unit time;

[0066] Calculate the cumulative exposure to PM2.5 using the following formula:

[0067]

[0068] Where n is the total sampling time.

[0069] Furthermore, the random forest-Bayesian polynomial regression unit obtains the interaction effect model based on the environment-behavior-physiological characteristic data and using the random forest and Bayesian polynomial algorithms, including the following specific steps:

[0070] The random forest algorithm was used to assess health risks and rank the importance of environmental, behavioral, and physiological characteristics, and several characteristics from the importance ranking results were used as key characteristics affecting health.

[0071] Based on the key features, a second-order polynomial regression model is constructed to quantify the interaction effects between environmental features and between them and behavioral and physiological features, forming an interaction effect model to obtain the comprehensive health-comfort index y, which is expressed as:

[0072] y=β0+∑β i x i +∑β ij x i x j +ε

[0073] Among them, x i is the environmental characteristic parameter, β ij Characterize the synergistic effect of environment-behavior-physiological characteristics; β i is the i-th key feature obtained based on the random forest algorithm, β0 is the intercept, which represents the baseline health risk when all key features are equal to the reference value, x j is the eigenvalue of the jth key feature, and ε is the random error term.

[0074] Furthermore, in the process of optimizing according to the interaction effect model, the set constraints and the optimization target, the particle swarm optimization unit

[0075] The constraints set include:

[0076] The maximum indoor temperature does not exceed the set temperature threshold;

[0077] The CO2 concentration does not exceed the set concentration threshold;

[0078] PM2.5 does not exceed the set concentration threshold;

[0079] VOCs do not exceed the set concentration threshold;

[0080] The maximum device power does not exceed the set threshold;

[0081] The penalty function is introduced to deal with the constraints and the objective function formula is obtained, which is expressed as:

[0082] f′(x)=f(x)+λ∑max(0,g i (x)) 2

[0083] Among them, f(x) is the original objective function; λ is the penalty function coefficient; g i (x) is the i-th constraint condition, when g i When (x)>0, it means that the constraint condition is not satisfied; max(0,g i (x)) 2 For penalty items;

[0084] Set the optimization goal to:

[0085] Maximize comprehensive health-comfort indicators and minimize equipment energy consumption;

[0086] The inertia weight adaptive adjustment formula in the optimization process is set as:

[0087] ω(t)=ω 终值 (ω 初值 -ω 终值 )×e -kt

[0088] Where t is the number of iterations or time steps, ω(t) is the inertia weight at the tth iteration; ω 初值 is the initial higher inertia weight; ω 终值 is the lower inertia weight in the later period; k is the attenuation factor.

[0089] Beneficial effects: The present invention obtains environmental data through an environmental data acquisition and processing unit, and processes the environmental data to obtain processed environmental data; obtains user physiological data through a physiological data monitoring module, and the physiological data includes: heart rate variability, blood oxygen saturation and body movement frequency data; obtains user behavior data through a behavior detection module, and the user behavior data includes: activity trajectory, sitting and lying time and device use frequency data; the present invention realizes a data closed loop by unified processing of environmental, physiological and behavioral data, is not limited to the temporal or spatial analysis of a single data source, and can quantify the interaction between different modal data. The activity trajectory is segmented by the space-time cube analysis unit, and the segmented activity trajectory and the kernel density estimation method are used to generate an indoor stay heat map to obtain a space-time cube with space-time labels; the feature engineering unit is used to extract features based on the processed environmental data, user physiological data and the space-time cube with space-time labels to obtain environmental-behavioral-physiological feature data; the random forest-Bayesian polynomial regression unit is used to obtain an interaction effect model based on the environmental-behavioral-physiological feature data and the random forest and Bayesian polynomial algorithms; the particle swarm optimization unit is used to search for the optimal environment control parameters based on the interaction effect model and the set constraints and optimization objectives, and the environmental equipment is adjusted based on the optimal environment control parameters to provide the user with the optimal environment state while comprehensively considering the relationship between the environment, physiological and behavioral data. BRIEF DESCRIPTION OF THE DRAWINGS

[0090] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following is a brief introduction to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0091] Figure 1 This is a structural diagram of an elderly health care environment adjustment system in the present invention. DETAILED DESCRIPTION

[0092] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0093] This embodiment provides a system for regulating the elderly health care environment. Figure 1 Shown, including:

[0094] An environmental data acquisition and processing unit, configured to acquire environmental data and process the environmental data to obtain processed environmental data;

[0095] In a specific embodiment, the environmental data acquisition and processing unit includes:

[0096] Environmental data acquisition module, used to obtain environmental data, including temperature, humidity, PM2.5, CO2 concentration, volatile organic compounds (VOCs), and light intensity data;

[0097] Specifically, in this embodiment, temperature and humidity sensors, smoke sensors, light sensors and other sensors are set at multiple locations indoors to collect data such as temperature, humidity, PM2.5, and light intensity at a frequency of 1 Hz.

[0098] The sliding window and normality detection unit is used to obtain data from the environmental data acquisition module, perform normality detection on the obtained data, process the obtained data according to the normality detection results and the set processing rules, and obtain normalized environmental data; at the same time, multi-level anomaly detection is used for the normalized environmental data to ensure the quality of the input data.

[0099] The environmental outlier detection unit is used to perform an outlier elimination operation on the normalized environmental data to obtain processed environmental data.

[0100] A physiological data monitoring module is used to obtain the user's physiological data, including heart rate variability, blood oxygen saturation and body movement frequency data;

[0101] Specifically, in this embodiment, data is collected at a frequency of 0.1 Hz through sensors such as a photoelectric heart rate sensor (for measuring heart rate variability HRV), a blood oxygen sensor (for measuring blood oxygen saturation SpO2), an infrared body temperature sensor, and an accelerometer. Specifically, the sensors can be worn or implanted in the smart mattress.

[0102] Specifically, in this embodiment, a joint denoising algorithm is used to pre-process the original user physiological data (HRV, SpO2, etc.) sampled at a low frequency, thereby effectively reducing the motion artifact noise in the HRV data. After processing, the HRV value is increased from the original 28ms to 42ms, and the noise is reduced by about 37%, which greatly reduces the interference of noise on data analysis and improves the availability and accuracy of the data.

[0103] A behavior detection module is used to obtain user behavior data, including activity trajectory, sitting and lying time, and device usage frequency data;

[0104] Specifically, data such as activity trajectories, sitting and lying time, and frequency of use of equipment (such as wheelchairs or crutches) are collected based on UWB ultra-wideband positioning, millimeter-wave radar or six-axis inertial sensors. Such data are often event-triggered or recorded with irregular timestamps and have strong spatiotemporal correlation.

[0105] Specifically, in this embodiment, the space-time cube analysis unit uses a pre-defined algorithm to automatically segment the collected activity trajectories into three segments, corresponding to resting, walking, and toileting. Simultaneously, a kernel density estimation method is used to generate a dwell heat map of the behavioral data. The results show that the probability density of the bedside area reaches 0.72, exceeding the preset threshold, indicating that this area is a key area of concern.

[0106] A space-time cube analysis unit is used to segment the activity trajectory and generate an indoor stay heat map using the segmented activity trajectory and a kernel density estimation method to obtain a space-time cube with space-time labels;

[0107] The feature engineering unit is used to extract features based on the processed environmental data, user physiological data, and behavioral analysis results with spatiotemporal labels to obtain environmental-behavioral-physiological feature data;

[0108] A random forest-Bayesian polynomial regression-particle swarm modeling and optimization unit is used to establish an interaction effect model for characterizing the interaction between different modal data using random forest and Bayesian polynomial algorithms and environmental-behavioral-physiological characteristic data, and to optimize the interaction effect model and the set constraints and optimization objectives to obtain the optimal environmental control parameters;

[0109] Specifically, this embodiment achieves a closed-loop data model by integrating environmental, physiological, and behavioral data. This model is not limited to temporal or spatial analysis of a single data source, but can quantify the interactions between different modal data. For example, by jointly modeling high temperature and high humidity environments with physiological abnormalities (such as heart rate and SpO2 changes), it accurately reveals their multiplier effects on health indicators such as respiratory function and cardiovascular stress.

[0110] A health assessment unit based on digital twins is used to simulate the optimal environmental control parameters in a virtual environment to determine whether the optimal environmental control parameters meet the set health indicators and energy consumption targets;

[0111] The environmental equipment control unit is used to adjust the environmental equipment according to the environmental control parameters that meet the set health indicators and energy consumption targets.

[0112] In a specific embodiment, the sliding window and normality detection unit acquires data from the environmental data acquisition module, performs normality detection on the acquired data, and processes the acquired data according to the normality detection result and the set processing rules to obtain normalized environmental data. The specific steps include:

[0113] Establishing a dual time window for normality testing and data transformation, wherein the dual time window includes a short window and a long window;

[0114] The short window is used to obtain data from the environmental data acquisition module according to the set time and frequency;

[0115] Specifically, in this embodiment, a short window is set to extract 300 pieces of environmental data from the 1Hz data every 5 minutes for normality testing. At the same time, a long window distribution analysis is performed on the all-day data every day. The improved Shapiro–Wilk normality test is performed on the temperature data within 5 minutes. The test results show abnormal temperature fluctuations (the detection index W' value is 0.76). Based on the test results, the system automatically starts the Box-Cox transformation to perform distribution correction on the temperature data to eliminate the impact of data skewness on subsequent models. By preprocessing the data, the accuracy and robustness of the system can be improved.

[0116] The long window is used to obtain data from the environmental data acquisition module every 24 hours to capture diurnal variations and seasonal trends;

[0117] The data obtained in the long window are used to perform normality test on the data in the short window using the improved Shapiro–Wilk test method. The normality test formula is:

[0118]

[0119] Where W′ is the improved Shapiro-Wilk test statistic, which is used to dynamically adjust the data normality test results; W is the original Shapiro-Wilk test statistic, which reflects the degree of fit between the data and the normal distribution; ΔT 昼夜 represents the difference between the highest temperature and the lowest temperature within 24 hours obtained based on the long window, which is used to correct the test statistic to adapt to the change of ambient temperature;

[0120] Specifically, this embodiment constructs ΔT by long window data 昼夜 , assisting the short window to complete the normality test / transformation and improve the accuracy of normality judgment.

[0121] Determine whether the normality test result satisfies the normal distribution. If so, all the data in the dual time window are used as normalized environmental data. Otherwise, the data in the short window are processed using Box-Cox transformation or logarithmic transformation, and the data in the long window and the processed data in the short window are used as normalized environmental data.

[0122] Specifically, in this embodiment, when W′<0.9, the Box-Cox transformation logarithmic transformation is started to process the data in the short window.

[0123] In a specific embodiment, the environmental outlier detection unit performs an outlier removal operation on the normalized environmental data, including:

[0124] 1) Construct a boxplot of normalized environmental data based on the quantile regression method, including:

[0125] Calculate the lower quartile Q1 and upper quartile Q3 of each time period, and determine the interquartile range IQR;

[0126] The dynamic upper and lower bounds are set based on the lower quartile Q1, upper quartile Q3 and interquartile range IQR, respectively:

[0127] The upper bound is expressed as:

[0128] Q3+k'×IQR

[0129] The lower bound is expressed as:

[0130] Q1-k'×IQR

[0131] In the formula, k' is the set threshold parameter;

[0132] Compare the normalized environmental data with the upper and lower bounds in turn, and mark the data that are not within the upper and lower bounds;

[0133] 2) Set temperature warning rules, including:

[0134] If the temperature difference between adjacent moments in a certain area exceeds the set difference threshold, the temperature of that period is marked as candidate abnormal data;

[0135] In this embodiment, the difference threshold is set to 3°C.

[0136] 3) Use the normalized environmental data to construct an isolation forest model and calculate the anomaly score of each type of data in the normalized environmental data. The calculation formula for the anomaly score is as follows:

[0137] Abnormal score = 0.7 × IF (X 正态化环境数据 )+0.3×LSTM(X 行为序列 )

[0138] Where IF represents isolation forest anomaly detection, and LSTM represents temporal pattern analysis of the corresponding behavior sequence;

[0139] 4) Calculate the average anomaly score value st based on the anomaly score of each type of data in the normalized environmental data, st∈(0,1];

[0140] 5) Based on the marking results of the box plot and temperature warning rules, set the rejection threshold of the corresponding type of data in the normalized environmental data, and reject abnormal temperature, humidity, PM2.5, CO2 concentration, volatile organic compounds (VOCs), and light intensity data according to the rejection threshold.

[0141] Specifically, for data that has been marked by temperature warning rules or box plots, the threshold can be lowered to increase the confidence level of elimination and achieve rule-model fusion.

[0142] Specifically, this embodiment uses an improved sliding window to perform normality test, a joint denoising algorithm, and an outlier detection method that combines dynamic box plots / isolation forests to automatically identify hidden nonlinear and dynamically changing features in the data, avoiding the limitations of traditional methods that rely solely on manually defined features.

[0143] In a specific embodiment, the spatiotemporal cube analysis unit segments the activity trajectory, and uses the segmented activity trajectory and kernel density estimation method to generate an indoor stay heat map. The specific steps of obtaining the behavior analysis results with spatiotemporal labels include:

[0144] Calculate the speed of two adjacent points in the activity trajectory;

[0145] According to the speed of two adjacent points in the activity trajectory, the speed mutation point is identified according to the set speed threshold; specifically, when the speed difference between two adjacent points exceeds the preset speed threshold, the two points can be considered as speed mutation points.

[0146] Segment the activity trajectory according to the speed mutation point to obtain the segmented activity trajectory;

[0147] Map the segmented activity trajectory to the indoor floor plan to obtain the mapped coordinate points;

[0148] The mapped coordinates are calculated using kernel density estimation (KDE) to generate a dwell heat map reflecting where users stay or frequently move. This map is a spatiotemporal cube with temporal and spatial labels, which helps identify areas where seniors frequently move and where potential safety risks may exist. The kernel density estimation method for generating dwell heat maps is well known in the art and will not be detailed here.

[0149] Specifically, this embodiment organically combines event triggering characteristics (such as instrument use and hotspots) with environmental regulation through the space-time cube analysis unit, providing more comprehensive data support for subsequent health risk assessment.

[0150] In a specific embodiment, the feature engineering unit performs feature extraction based on the processed environmental data, user physiological data, and a spatiotemporal cube with spatiotemporal labels to obtain environmental-behavioral-physiological feature data including time domain features, frequency domain features, spatial features, and health indicators. The specific steps include:

[0151] 1) Extract time domain features:

[0152] Calculate the mean and variance of each type of data in the processed environmental data respectively;

[0153] Calculate the mean and variance of the user's physiological data respectively;

[0154] 2) Extract frequency domain features:

[0155] Perform wavelet transform on the processed environmental data and user physiological data, and calculate the energy distribution of the transformed data in each frequency band;

[0156] Energy entropy is extracted based on the energy distribution of the transformed data in each frequency band, and the energy entropy is used as a frequency domain feature;

[0157] 3) Extract spatial features:

[0158] Mapping the temperature field of each room or functional area in the user's environment onto the space-time cube with the space-time label;

[0159] Obtaining the temperature difference between adjacent rooms or functional areas and calculating the average value of the temperature difference in the time dimension to obtain the temperature gradient between different rooms or functional areas, and using the temperature gradient as a spatial feature;

[0160] 4) Calculate health indicators, including:

[0161] Calculate the temperature fluctuation rate using the following formula:

[0162]

[0163] Where, max(T 1h ) is the maximum temperature per unit time, min(T 1h ) is the minimum temperature per unit time; mean(T 1h ) is the average temperature per unit time;

[0164] Calculate the cumulative exposure to PM2.5 using the following formula:

[0165]

[0166] Where n is the total sampling time.

[0167] In a specific embodiment, the random forest-Bayesian polynomial regression unit obtains the interaction effect model based on the environment-behavior-physiological characteristic data and using the random forest and Bayesian polynomial algorithms, including the following specific steps:

[0168] The random forest algorithm was used to assess health risks and rank the importance of environmental, behavioral, and physiological characteristics, and several characteristics from the importance ranking results were used as key characteristics affecting health.

[0169] Specifically, key features are screened out by ranking the importance of environmental-behavioral-physiological features using the random forest algorithm. These features cover multiple aspects, including but not limited to time domain features, frequency domain features, spatial features, and health indicator-related features.

[0170] Based on the key features, a second-order polynomial regression model is constructed to quantify the interaction effects between environmental features and between them and behavioral and physiological features, forming an interaction effect model to obtain the comprehensive health-comfort index y, which is expressed as:

[0171] y=β0+∑β i x i +∑β ij x i x j +ε

[0172] Among them, x i is the environmental characteristic parameter, β ij Characterizes the synergistic effects of environment, behavior, and physiological characteristics (such as the multiplicative effect of high temperature and humidity on cardiovascular stress); β i is the i-th key feature obtained based on the random forest algorithm, β0 is the intercept, which represents the baseline health risk when all key features are equal to the reference value, x j is the eigenvalue of the jth key feature, and ε is the random error term. In a specific embodiment, the particle swarm optimization unit performs optimization according to the interaction effect model and the set constraints and optimization objectives.

[0173] The constraints set include:

[0174] The maximum indoor temperature does not exceed the set temperature threshold;

[0175] The CO2 concentration does not exceed the set concentration threshold;

[0176] PM2.5 does not exceed the set concentration threshold;

[0177] VOCs do not exceed the set concentration threshold;

[0178] The maximum device power does not exceed the set threshold;

[0179] Specifically, this embodiment sets the maximum indoor temperature to no more than 28°C.

[0180] Introducing the penalty function to deal with the constraints, we get the objective function formula, which is expressed as:

[0181] f′(x)=f(x)+λ∑max(0,g i (x)) 2

[0182] Where f(x) is the original objective function; λ is the penalty function coefficient, which is used to adjust the penalty intensity of constraint violation; g i (x) is the i-th constraint condition, when g i When (x)>0, it means the constraint is violated; max(0,g i (x)) 2 As the penalty term, the part that violates the constraint is quadratically weighted to increase the constraint penalty;

[0183] Set the optimization goal to:

[0184] Maximize comprehensive health-comfort indicators and minimize equipment energy consumption;

[0185] The inertia weight adaptive adjustment formula in the optimization process is set as:

[0186] ω(t)=ω 终值 +(ω 初值 ω 终值 )×e -kt

[0187] Where t is the number of iterations or time steps, ω(t) is the inertia weight at the tth iteration; ω 初值 is the initial higher inertia weight (used to enhance global search); ω 终值 is the lower inertia weight in the later stage (used to improve local convergence accuracy); k is the attenuation factor (controls the attenuation speed of the inertia weight from the initial to the later stage).

[0188] Specifically, this embodiment dynamically adjusts the inertia weight according to the optimization process, and sets a high weight (ω 初始 =0.9) to enhance global search, and set low weight (ω 终值 =0.4) to improve local convergence.

[0189] Specifically, in this embodiment, the health assessment unit of the digital twin is used to construct the room structure, equipment model and sensor layout in the virtual environment, load the elderly behavior prediction model (short-term behavior prediction based on LSTM, predicting the movement status in the next 30 minutes), and receive the optimal environmental control parameters output by the particle swarm optimization unit to simulate the indoor layout, elderly behavior patterns, etc.; calculate the impact of indoor temperature, humidity, air quality and other responses on health indicators and energy consumption, and judge whether the preset health, comfort and energy consumption goals are met; if the goals are met, the optimal environmental control parameters are sent to the on-site equipment (air conditioning, fresh air system, humidifier / dehumidifier, etc.) for execution; if not, the simulation deviation results are fed back to the particle swarm optimization unit for iteration or adjustment until the goals are met or the maximum number of iterations is reached.

[0190] Specifically, digital twin simulation technology is used to pre-verify the effectiveness of optimal environmental control parameters in a virtual environment, ensuring that health, comfort and energy consumption are optimized after instructions are issued on-site, significantly improving the intelligent management level of the overall system.

[0191] The data collected in this embodiment are shown in Table 1:

[0192] Table 1

[0193]

[0194]

[0195] Environmental data:

[0196] Acquisition parameters: temperature 25.6°C (standard deviation 1.2°C), PM2.5 concentration 45 μg / m 3 ;

[0197] Data characteristics: High-frequency time series data (sampling frequency 1 Hz), the temperature gradient in the bedroom area is 0.8°C / m.

[0198] Collection location: Mainly collected from multiple nodes in the bedroom to ensure comprehensive monitoring of indoor temperature distribution.

[0199] Physiological data:

[0200] Acquisition parameters: heart rate variability (HRV) was 28ms (50ms lower than the normal level), blood oxygen saturation (SpO2) was 97%, and body movement frequency was 3 times / hour;

[0201] Data characteristics: Low-frequency data (sampling frequency is about 0.1 Hz), but there is artifact noise caused by motion;

[0202] Collection location: Monitoring through smart mattresses and wearable devices.

[0203] Behavioral data:

[0204] Collection parameters: The activity trajectory length of the elderly is about 80m / hour, the sitting and lying time accounts for 65%, and the frequency of device use is 0;

[0205] Data characteristics: The data is formed into a spatiotemporal trajectory through the indoor positioning system. The number of trajectory segments is 3, and the generated stay heat map shows that the probability density of the bedside area is 0.72;

[0206] Collection location: Indoor positioning devices (such as UWB positioning systems or pressure pads) are installed in bedrooms and related activity areas.

[0207] Specifically, environmental, behavioral, and physiological characteristic data are used as inputs to the random forest algorithm for health risk assessment. Several features in the importance ranking results are used as key features affecting health. The Bayesian polynomial algorithm is used to conduct a comprehensive analysis of the data to output a comprehensive health-comfort index.

[0208] In this example, the calculated health risk assessment value is 0.68, exceeding the preset threshold of 0.6. When the health risk assessment value indicates a high risk (health risk assessment value > 0.6), the multi-objective dynamic optimization process is initiated. Furthermore, random forest analysis reveals that the key health-influencing features are nighttime PM2.5 concentration (Gini importance 0.35) and HRV (0.28), providing a basis for subsequent optimization and control.

[0209] Based on the dual objectives of maximizing the comprehensive health-comfort index and minimizing equipment energy consumption, the particle swarm optimization unit searches for the optimal control strategy. During the optimization process, the system automatically integrates the weights and constraints of each parameter and ultimately generates the following optimization solution:

[0210] Air conditioning set temperature: 24.0℃

[0211] Air purifier gear: 2 gears

[0212] Light intensity: 50 (unit can be set according to system)

[0213] After the optimization plan is generated, the system will immediately issue an execution instruction through the control module, and at the same time start the digital twin verification link for simulation prediction. The system sends the optimized parameter instructions to the on-site equipment, such as adjusting the air conditioning temperature to 24.0℃, setting the air purifier to 2nd gear operation, and adjusting the indoor lighting to 50 units. After the control instruction took effect, the monitoring data feedback showed that HRV recovered from the original 42ms to 52ms within 40 minutes, and SpO2 stabilized at 98%. At the same time, through monitoring of sleep data, it was found that the duration of deep sleep increased from 2.1 hours to 2.6 hours, an increase of about 23%. Environmental data collected the next day further showed that the PM2.5 concentration dropped to 32μg / m 3 , the standard deviation of temperature fluctuation was reduced to 0.5℃, indicating that the overall environmental conditions were improved.

[0214] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A system for regulating the elderly's health care environment, characterized in that: include: An environmental data acquisition and processing unit, configured to acquire environmental data and process the environmental data to obtain processed environmental data; A physiological data monitoring module is used to obtain the user's physiological data, including heart rate variability, blood oxygen saturation and body movement frequency data; A behavior detection module is used to obtain user behavior data, including activity trajectory, sitting and lying time, and device usage frequency data; A space-time cube analysis unit is used to segment the activity trajectory and generate an indoor stay heat map using the segmented activity trajectory and a kernel density estimation method to obtain a space-time cube with space-time labels; The feature engineering unit is used to extract features based on the processed environmental data, user physiological data, and the spatiotemporal cube with spatiotemporal labels to obtain environmental-behavioral-physiological feature data; A random forest-Bayesian polynomial regression-particle swarm modeling and optimization unit is used to establish an interaction effect model for characterizing the interaction between different modal data using random forest and Bayesian polynomial algorithms and environmental-behavioral-physiological characteristic data, and to optimize the interaction effect model and the set constraints and optimization objectives to obtain the optimal environmental control parameters; A health assessment unit based on digital twins is used to simulate the optimal environmental control parameters in a virtual environment to determine whether the optimal environmental control parameters meet the set health indicators and energy consumption targets; The environmental equipment control unit is used to adjust the environmental equipment according to the environmental control parameters that meet the set health indicators and energy consumption targets.

2. The elderly health care environment adjustment system according to claim 1 is characterized in that: The environmental data acquisition and processing unit includes: Environmental data acquisition module, used to obtain environmental data, including temperature, humidity, PM2.5, CO2 concentration, volatile organic compounds (VOCs), and light intensity data; A sliding window and normality detection unit is used to obtain data from the environmental data acquisition module, perform normality detection on the obtained data, and process the obtained data according to the normality detection result and the set processing rules to obtain normalized environmental data; The environmental outlier detection unit is used to perform an outlier elimination operation on the normalized environmental data to obtain processed environmental data.

3. The elderly health care environment adjustment system according to claim 2 is characterized in that: The sliding window and normality detection unit acquires data from the environmental data acquisition module, performs normality detection on the acquired data, and processes the acquired data according to the normality detection result and the set processing rules to obtain normalized environmental data. The specific steps include: Establishing a dual time window for normality testing and data transformation, wherein the dual time window includes a short window and a long window; The short window is used to obtain data from the environmental data acquisition module according to the set time and frequency; The long window is used to obtain data from the environmental data acquisition module every 24 hours; The data obtained in the long window are used to perform normality test on the data in the short window using the improved Shapiro–Wilk test method. The normality test formula is: Where W′ is the improved Shapiro-Wilk test statistic, which is used to dynamically adjust the data normality test results; W is the original Shapiro-Wilk test statistic, which reflects the degree of fit between the data and the normal distribution; ΔT 昼夜 represents the difference between the highest temperature and the lowest temperature within 24 hours obtained based on the long window, which is used to correct the test statistic to adapt to the change of ambient temperature; Determine whether the normality test result satisfies the normal distribution. If so, all the data in the dual time window are used as normalized environmental data. Otherwise, the data in the short window are processed using Box-Cox transformation or logarithmic transformation, and the data in the long window and the processed data in the short window are used as normalized environmental data.

4. The elderly health care environment adjustment system according to claim 3 is characterized in that: The environmental outlier detection unit performs an outlier elimination operation on the normalized environmental data, comprising: 1) Construct a boxplot of normalized environmental data based on the quantile regression method, including: Calculate the lower quartile Q1 and upper quartile Q3 of each time period, and determine the interquartile range IQR; The dynamic upper and lower bounds are set based on the lower quartile Q1, upper quartile Q3 and interquartile range IQR, respectively: The upper bound is expressed as: Q3+k'×IQR The lower bound is expressed as: Q1-k'×IQR In the formula, k' is the set threshold parameter; Compare the normalized environmental data with the upper and lower bounds in turn, and mark the data that are not within the upper and lower bounds; 2) Set temperature warning rules, including: If the temperature difference between adjacent moments in a certain area exceeds the set difference threshold, the temperature of that period is marked as candidate abnormal data; 3) Use the normalized environmental data to construct an isolation forest model and calculate the anomaly score of each type of data in the normalized environmental data. The calculation formula for the anomaly score is as follows: Abnormal score = 0.7 × IF (X 正态化环境数据 )+0.3×LSTM(X 行为序列 ) where IF represents isolation forest anomaly detection and LSTM represents temporal pattern analysis of corresponding behavior sequences; 4) Calculate the average anomaly score value st based on the anomaly score of each type of data in the normalized environmental data, st∈(0,1]; 5) Based on the marking results of the box plot and temperature warning rules, set the rejection threshold of the corresponding type of data in the normalized environmental data, and reject abnormal temperature, humidity, PM2.5, CO2 concentration, volatile organic compounds (VOCs), and light intensity data according to the rejection threshold.

5. The elderly health care environment adjustment system according to claim 4 is characterized in that: The space-time cube analysis unit segments the activity trajectories and uses the segmented activity trajectories and kernel density estimation method to generate indoor stay heat maps. The specific steps to obtain the behavior analysis results with space-time labels include: Calculate the speed of two adjacent points in the activity trajectory; According to the speed of two adjacent points in the activity trajectory, the speed mutation point is identified according to the set speed threshold; Segment the activity trajectory according to the speed mutation point to obtain the segmented activity trajectory; Map the segmented activity trajectory to the indoor floor plan to obtain the mapped coordinate points; The mapped coordinate points are calculated using a kernel density estimation method to generate a stay heat map reflecting the user's stay or high-frequency activity areas. The stay heat map is a space-time cube with space-time labels.

6. The elderly health care environment adjustment system according to claim 5 is characterized in that: The feature engineering unit extracts features based on the processed environmental data, user physiological data, and the spatiotemporal cube with spatiotemporal labels to obtain environmental-behavioral-physiological feature data including time domain features, frequency domain features, spatial features, and health indicators. The specific steps include: 1) Extract time domain features: Calculate the mean and variance of each type of data in the processed environmental data respectively; Calculate the mean and variance of the user's physiological data respectively; 2) Extract frequency domain features: Perform wavelet transform on the processed environmental data and user physiological data, and calculate the energy distribution of the transformed data in each frequency band; Energy entropy is extracted based on the energy distribution of the transformed data in each frequency band, and the energy entropy is used as a frequency domain feature; 3) Extract spatial features: Mapping the temperature field of each room or functional area in the user's environment onto the space-time cube with the space-time label; Obtaining the temperature difference between adjacent rooms or functional areas and calculating the average value of the temperature difference in the time dimension to obtain the temperature gradient between different rooms or functional areas, and using the temperature gradient as a spatial feature; 4) Calculate health indicators, including: Calculate the temperature fluctuation rate using the following formula: Where, max(T 1h ) is the maximum temperature per unit time, min(T 1h ) is the minimum temperature per unit time; mean(T 1h ) is the average temperature per unit time; Calculate the cumulative exposure to PM2.5 using the following formula: Where n is the total sampling time.

7. The elderly health care environment adjustment system according to claim 6 is characterized in that: The random forest-Bayesian polynomial regression unit obtains the interaction effect model based on the environment-behavior-physiological characteristic data and uses the random forest and Bayesian polynomial algorithms, and the specific steps include: The random forest algorithm was used to assess health risks and rank the importance of environmental, behavioral, and physiological characteristics, and several characteristics from the importance ranking results were used as key characteristics affecting health. Based on the key features, a second-order polynomial regression model is constructed to quantify the interaction effects between environmental features and between them and behavioral and physiological features, forming an interaction effect model to obtain the comprehensive health-comfort index y, which is expressed as: y=β0+Σβ i x i +Sv ij x i x j +e Among them, x i is the environmental characteristic parameter, β ij Characterize the synergistic effect of environment-behavior-physiological characteristics; β i is the i-th key feature obtained based on the random forest algorithm, β0 is the intercept, which represents the baseline health risk when all key features are equal to the reference value, x j is the eigenvalue of the jth key feature, and ε is the random error term.

8. The elderly health care environment adjustment system according to claim 7 is characterized in that: In the process of optimizing according to the interaction effect model, the set constraints and the optimization goal, the particle swarm optimization unit The constraints set include: The maximum indoor temperature does not exceed the set temperature threshold; The CO2 concentration does not exceed the set concentration threshold; PM2.5 does not exceed the set concentration threshold; VOCs do not exceed the set concentration threshold; The maximum device power does not exceed the set threshold; The penalty function is introduced to deal with the constraints and the objective function formula is obtained, which is expressed as: f′(x)=f(x)+λΣmax(0,g i (x)) 2 Among them, f(x) is the original objective function; λ is the penalty function coefficient; g i (x) is the i-th constraint condition, when g i When (x)>0, it means that the constraint condition is not satisfied; max(0,g i (x)) 2 For penalty items; Set the optimization goal to: Maximize comprehensive health-comfort indicators and minimize equipment energy consumption; The inertia weight adaptive adjustment formula in the optimization process is set as: ω(t)=ω 终值 (oh 初值 oh 终值 ×e -kt Where t is the number of iterations or time steps, ω(t) is the inertia weight at the tth iteration; ω 初值 is the initial higher inertia weight; ω 终值 is the lower inertia weight in the later period; k is the attenuation factor.

Images