Inflation heating system linked with wireless body temperature monitoring and capable of automatically adjusting temperature setting

Abstract

The invention discloses an inflation heating system linked with wireless body temperature monitoring and capable of automatically adjusting temperature setting, and relates to the field of body temperature monitoring, which comprises the following steps: acquiring initial multi-modal data of a user by wearing a multi-modal monitoring device, acquiring historical multi-modal data, establishing a control platform, inputting a currently required target temperature in the control platform, and setting the target temperature according to the target temperature; the heating power of the inflation equipment is automatically adjusted for temperature adjustment, the multi-modal monitoring device collects the multi-modal data again after temperature adjustment is completed, and target multi-modal data is formed; calculating the temperature compensation amount of the displacement area of the flexible sensor; and constructing a displacement-humidity coupling compensation model based on the temperature compensation amount, the initial multi-modal data and humidity data in the target multi-modal data, and importing the target power into the displacement-humidity coupling compensation model for updating to obtain corrected power. The system has the advantages that body temperature maintenance in a dynamic environment is realized through multi-modal fusion and a compensation mechanism.

Description

Technical Field

[0001] The present invention relates to the field of body temperature monitoring, and particularly to an inflatable heating system that is linked to wireless body temperature monitoring and automatically adjusts temperature settings. Background Art

[0002] In existing body temperature maintenance systems, in scenarios such as medical care and outdoor operations, most use fixed temperature sensors for static monitoring. When the human body moves, the relative relationship between the sensor position and the body surface heat source changes, resulting in distortion of the temperature field monitoring (for example, the misjudgment rate of the core body temperature reaches 15% when the patient turns over); at the same time, the existing technology generally ignores the dynamic influence of humidity on heat conduction. In a high-humidity environment, the body surface evaporation heat dissipation rate increases by 30%-50%, but the traditional temperature control system does not establish a humidity compensation mechanism, resulting in a certain deviation between the set temperature and the actual perceived temperature. In addition, the temperature regulation system based on the PID algorithm has a significant response delay due to thermal inertia. When the body surface temperature rises rapidly after strenuous exercise, the set value of the overshoot of the system and the time to restore the steady state are also longer. More prominently, the existing solutions process temperature, humidity, and displacement data separately, lacking a multi-modal feature fusion mechanism. When the sensor deviates by 5 cm due to body movement, the positioning error of the heating area increases the risk of local overheating. These defects seriously restrict the clinical application and outdoor expansion of precise body temperature maintenance technology. Summary of the Invention

[0003] To solve the above technical problems, the present invention provides an inflatable heating system that is linked to wireless body temperature monitoring and automatically adjusts temperature settings. This technical solution solves the problems of multi-modal data fragmentation, insufficient environmental factor compensation, thermal inertia response delay, and static monitoring and adjustment lag proposed in the above background art.

[0004] To achieve the above objectives, the technical solution adopted by the present invention is as follows: An inflatable heating system that is linked to wireless body temperature monitoring and automatically adjusts temperature settings, comprising: A collection module: collects the initial multi-modal data of the user through a wearable multi-modal monitoring device. The multi-modal monitoring device is composed of a flexible temperature sensor, a laser scanning module, and a micro humidity sensor. The initial multi-modal data includes temperature data, scanning data, and humidity data; A preliminary temperature adjustment module: obtains historical multi-modal data to establish a control platform, inputs the current required target temperature inside the control platform, combines the body temperature data inside the initial multi-modal data, and independently adjusts the heating power of the inflatable device for temperature adjustment. At this time, the heating power is recorded as the target power. After the temperature adjustment is completed, the multi-modal monitoring device collects multi-modal data again to form target multi-modal data; Temperature analysis module: Based on the scan data in the initial multimodal data, the control platform establishes a reference model of the human body contour and projects it onto a plane to establish a coordinate system. Then, it projects the target multimodal data into the coordinate system, marks the position coordinates of the flexible sensors in both data, and projects the corresponding temperature data to form a temperature monitoring distortion map. Compensation calculation module: Based on the temperature monitoring distortion map, calculate the temperature compensation amount in the displacement area of the flexible sensor through the temperature compensation equation. Temperature correction module: Based on the temperature compensation amount, the humidity data in the initial multimodal data and the target multimodal data, construct a displacement-humidity coupling compensation model, import the target power into the displacement-humidity coupling compensation model for updating to obtain the corrected power, and the control platform controls the inflation device to correct and adjust the temperature based on the corrected power.

[0005] Preferably, by combining the body temperature data inside the initial multimodal data, autonomously adjust the heating power of the inflation device for temperature regulation. At this time, the heating power is denoted as the target power, which specifically includes: Extract the temperature data values of multiple monitoring points collected by the flexible temperature sensor in the initial multimodal data at the same time, calculate the average value of the multiple temperature data values, and denote it as the average temperature. Obtain the target temperature independently input by the user at the input end of the control platform, calculate the absolute value of the difference between the target temperature and the average temperature to obtain the temperature deviation. Through historical data fitting, extract the historical heat capacity constant, heat conversion efficiency, and environmental loss compensation coefficient. Multiply the temperature difference value by the heat capacity constant to obtain the theoretical heat energy demand, and divide the theoretical heat energy demand by the product of the heat conversion efficiency and the environmental loss compensation coefficient to obtain the heating power. The heat capacity constant is the heat storage capacity parameter of the heating system obtained by multiplying the specific heat capacity of the material by the mass of the material. The heat conversion efficiency is the efficiency parameter for converting electrical energy into heat energy, and the environmental loss compensation coefficient is the proportion value of heat loss during transmission. Adjust the power of the inflation device to the heating power through the control platform for temperature adjustment operation. At this time, the heating power is denoted as the target power.

[0006] Preferably, the marking of the position coordinates of the flexible sensors in both data and the projection of the corresponding temperature data to form a temperature monitoring distortion map specifically includes: Obtain the position coordinates of the flexible sensors in both data and connect them. Mark the midpoint of the connecting line of the two flexible sensors, read the distance from the midpoint to the flexible sensor, and use the flexible sensor coordinates as the center of the circle and the distance from the midpoint to the flexible sensor as the radius to draw a circle, denoted as the flexible sensor monitoring area. Select two coordinate points closest to and farthest from the horizontal axis on the two flexible sensor detection areas, draw a horizontal line parallel to the horizontal axis, select two coordinate points closest to and farthest from the vertical axis on the two flexible sensor detection areas, draw a vertical line perpendicular to the horizontal axis, and connect the four lines to form a rectangular area, denoted as the temperature monitoring distortion area; Map the temperature data of each point in the temperature monitoring distortion area to the corresponding coordinates, and merge the mapped layer with the detection area to form a temperature monitoring distortion map.

[0007] Preferably, calculating the temperature compensation amount of the flexible sensor displacement area based on the temperature monitoring distortion map through the temperature compensation equation specifically includes: Obtain the coordinate positions of the flexible sensor in the initial multi-modal data and the target multi-modal data, and read the displacement vector length and the angle between the displacement vector and the coordinate axes between the two coordinate positions; According to the boundary of the rectangular area of the temperature monitoring distortion map, divide the rectangular area proportionally according to the displacement length along the displacement vector direction to form multiple temperature gradient sampling areas; Extract the maximum and minimum values of the temperature data in each sampling area, divide the difference between the maximum and minimum values by the displacement vector length to obtain the temperature gradient change value; Based on the displacement vector length, the angle between the displacement vector and the coordinate axes, the maximum and minimum values of the temperature data, calculate the temperature compensation amount through the compensation coefficient correction formula; Accumulate the compensation amounts of each sampling area to obtain the total temperature compensation amount and transmit it to the compensation database of the control platform to update the compensation parameters.

[0008] Preferably, calculating the temperature compensation amount based on the displacement vector length, the angle with the coordinate axes, the maximum and minimum values of the temperature data through the compensation coefficient correction formula specifically includes: Obtain the temperature gradient change value, the displacement vector length, and the angle between the displacement vector and the coordinate axes. Denote the displacement vector length as L and the angle between the displacement vector and the coordinate axes as θ, calculate the product of L and sinθ to obtain the actual displacement amount of the flexible sensor in the vertical monitoring direction, multiply the actual displacement amount by the temperature gradient change value to obtain the temperature deviation value caused by the displacement of the flexible sensor, denoted as the interval temperature difference; Based on historical data fitting to determine the time decay coefficient, obtain the interval time between the initial multi-modal data and the target multi-modal data, multiply the difference between the maximum and minimum values of the temperature data in each sampling area, the interval time, and the time decay coefficient to obtain the time decay term compensation value. The time decay coefficient is the temperature lag effect coefficient caused by the system thermal inertia, and the time decay term compensation value is the error caused by the control platform response delay.

[0009] Preferably, constructing a displacement-humidity coupling compensation model based on the temperature compensation amount, humidity data in the initial multimodal data, and humidity data in the target multimodal data specifically includes: The control platform aggregates the total temperature compensation amount, interval temperature difference, time decay term compensation value, displacement vector length, angle with the coordinate axis, temperature gradient change value, initial humidity value, and target humidity value to form an input database, and constructs a displacement-humidity coupling compensation model. The displacement-humidity coupling compensation model includes a data preprocessing module, a multimodal feature fusion module, and a power correction calculation module; The data preprocessing module normalizes all data in the input database, and calculates the humidity change gradient based on the humidity data in the initial multimodal data and the target multimodal data, and feeds it back to the multimodal feature fusion module; The multimodal feature fusion module adopts a cascaded neural network structure, including a temperature compensation branch, a displacement vector branch, and a humidity change branch. The output ends of each branch perform cross-modal fusion to obtain a fused feature vector and feed it back to the interior of the power correction calculation module; The power correction calculation module generates correction parameters including power values and time action intervals by performing matrix multiplication operations on the target power and the increment coefficient.

[0010] Preferably, the data preprocessing module normalizes all data in the input database, and calculating the humidity change gradient based on the humidity data in the initial multimodal data and the target multimodal data specifically includes: Obtain the initial humidity value in the initial multimodal data and the target humidity value in the target multimodal data, calculate the difference between the target humidity value and the initial humidity value, and record it as the humidity change value; Extract the timestamps of the initial multimodal data and the target multimodal data, read the time interval between the two, and divide the humidity change value by the time interval to obtain the humidity change gradient.

[0011] Preferably, the multimodal feature fusion module adopts a cascaded neural network structure, including a temperature compensation branch, a displacement vector branch, and a humidity change branch. The output ends of each branch perform cross-modal fusion to obtain a fused feature vector and feed it back to the interior of the power correction calculation module specifically includes: The multimodal feature fusion module records the total temperature compensation amount, interval temperature difference, and time decay term compensation value as temperature inputs to form a temperature compensation branch, records the displacement vector length, angle with the coordinate axis, and temperature gradient change value as displacement inputs to form a displacement vector branch, and records the initial humidity value and the target humidity value as humidity inputs to form a humidity change branch; The input layer of the temperature compensation branch receives the normalized total temperature compensation amount, and outputs a temperature feature vector through the activation function in the control platform; The input layer of the displacement vector branch receives the displacement vector length, angle and time attenuation compensation value, and extracts the displacement dynamic feature vector through time series modeling; The humidity change branch adopts a convolutional neural network structure. The input layer receives the humidity change gradient and historical humidity fluctuation parameters, captures the humidity time series correlation characteristics through the convolution kernel, and outputs the humidity feature vector. The temperature feature vector, displacement feature vector, and humidity feature vector are tensor-concatenated in the dimension direction to obtain a fused feature vector.

[0012] Preferably, the power correction calculation module generates a two-dimensional correction parameter including a power value and a time action interval by performing a matrix multiplication operation on the target power and the increment coefficient. Specifically, the two-dimensional correction parameter includes: The fused feature vector is input into the power correction calculation module, and the incremental coefficient matrix composed of the power adjustment factor and the time adjustment factor is obtained by summarizing. The incremental coefficient matrix is ​​a real number matrix with two rows and two columns. The elements in the first row are the power adjustment factors, and the elements in the second row are the time adjustment factors. The power adjustment factor is the weight coefficient of the corrected target power obtained by dividing the actual correction power in the historical data by the target power. The time adjustment factor is the actual time to reach the target temperature in the historical data divided by the theoretical time to obtain the control correction power action duration coefficient; Combine the target power and the constant 1 into a row vector with one row and two columns, perform matrix multiplication operation on the row vector and the incremental coefficient matrix, and obtain a two-dimensional correction parameter with one row and two columns, the first column is the correction power value, and the second column is the time interval of the power action; The corrected power value and time interval are fed back to the control platform to update the power control instructions of the inflatable device.

[0013] Compared with the prior art, the present invention has the following beneficial effects: The present invention proposes an inflatable heating system that is linked to wireless body temperature monitoring and automatically adjusts the temperature setting. A vector decomposition algorithm is used to calculate the displacement component in the vertical monitoring direction. The bilinear interpolation temperature field reconstruction technology is combined to reduce the temperature monitoring error when the sensor is offset by 5 cm, thereby improving the displacement compensation accuracy. The transformed rectangular distortion zone positioning algorithm improves the temperature monitoring response speed and realizes intelligent division of the distortion zone. The flexible temperature sensor array and the laser scanning module are coordinated to realize real-time tracking of human body displacement and dynamic modeling of the temperature field. The present invention proposes an air-filled warming system that is linked to wireless body temperature monitoring and automatically adjusts the temperature setting. A micro capacitive humidity sensor is used to calculate the humidity change gradient in real time. The influence weight of humidity on heat conduction is predicted through a convolutional neural network, so that the temperature control deviation in a high humidity environment is compressed to form a dynamic compensation of humidity. The time attenuation factor is introduced in combination with LSTM timing modeling (64 memory units) to shorten the system response delay to, reduce the overshoot, and eliminate the thermal inertia lag, thereby breaking through the limitation of single temperature feedback and constructing a multi-dimensional compensation system of displacement-humidity-thermal inertia. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 It is a schematic diagram of the process of the present invention; Figure 2 This is a flow chart of combining the body temperature data in the initial multimodal data to autonomously adjust the heating power of the inflatable device for temperature regulation in the present invention, where the heating power is recorded as the target power; Figure 3 A schematic diagram of the process of marking the position coordinates of the flexible sensor in the two data and projecting the corresponding temperature data to form a temperature monitoring distortion diagram in the present invention; Figure 4 It is a schematic diagram of a flow chart of calculating the temperature compensation amount of the displacement area of ​​the flexible sensor through a temperature compensation equation based on a temperature monitoring distortion diagram in the present invention; Figure 5 It is a flow chart of calculating the temperature compensation amount by using the compensation coefficient correction formula based on the displacement vector length, the angle with the coordinate axis, the maximum value and the minimum value of the temperature data in the present invention; Figure 6 It is a flow chart of constructing a displacement-humidity coupling compensation model based on the temperature compensation amount, the initial multimodal data and the humidity data in the target multimodal data in the present invention; Figure 7 A schematic diagram of a flow chart of a data preprocessing module in the present invention normalizing all data in an input database and calculating a humidity change gradient based on humidity data in initial multimodal data and target multimodal data; Figure 8 The multimodal feature fusion module of the present invention adopts a cascade neural network structure, including a temperature compensation branch, a displacement vector branch and a humidity change branch. The output end of each branch performs cross-modal fusion to obtain a fused feature vector and feed it back to the power correction calculation module. Figure 9 The power correction calculation module of the present invention generates a two-dimensional correction parameter including a power value and a time action interval by performing a matrix multiplication operation on the target power and the increment coefficient. DETAILED DESCRIPTION

[0015] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments in the following description are only examples, and other obvious variations can be conceived by those skilled in the art.

[0016] Referring to Figure 1 As shown, a linkage wireless body temperature monitoring and automatically adjusting temperature setting inflatable heating system includes: Acquisition module: Collect the initial multimodal data of the user through a wearable multimodal monitoring device. The multimodal monitoring device is composed of a flexible temperature sensor, a laser scanning module, and a micro humidity sensor. The initial multimodal data includes temperature data, scanning data, and humidity data; Preliminary temperature adjustment module: Obtain historical multimodal data to establish a control platform, input the current required target temperature inside the control platform, and combine the body temperature data inside the initial multimodal data to independently adjust the heating power of the inflatable device for temperature adjustment. At this time, the heating power is recorded as the target power. After the temperature adjustment is completed, the multimodal monitoring device collects multimodal data again to form target multimodal data; Temperature analysis module: Based on the scanning data in the initial multimodal data, the control platform establishes a human body contour reference model and projects it onto a plane to establish a coordinate system, and projects the target multimodal data onto the coordinate system, marks the position coordinates of the flexible sensors in the two data, and projects the corresponding temperature data to form a temperature monitoring distortion map; Compensation calculation module: Calculate the temperature compensation amount of the displacement area of the flexible sensor based on the temperature monitoring distortion map through a temperature compensation equation; Temperature correction module: Based on the temperature compensation amount, the humidity data in the initial multimodal data and the target multimodal data, construct a displacement-humidity coupling compensation model, import the target power into the displacement-humidity coupling compensation model for updating to obtain the corrected power, and the control platform controls the inflatable device to correct and adjust the temperature based on the corrected power.

[0017] Build a control platform based on historical data. The control platform processes historical multimodal data by establishing a time series prediction model based on LSTM. Its hidden layer contains 128 neuron nodes, which can effectively learn the distribution law of the temperature field. The multimodal monitoring device uses a flexible temperature sensor array embedded in an elastic substrate in a serpentine arrangement, and cooperates with the TOF ranging module of the laser scanning module to collect three-dimensional point clouds of the human body surface. The micro humidity sensor uses a capacitive measurement principle and monitors the humidity change of the microenvironment in real time at a sampling frequency of 100Hz. After collecting the temperature data of each monitoring point, the control platform calculates the heating power required to reach the target temperature according to the target temperature input by the user, and adjusts the inflatable device to the required heating power, so that the inflatable device can quickly perform temperature adjustment operations. After it completes the temperature adjustment operation, the multimodal monitoring device will collect the user's temperature again to check whether the preliminary temperature adjustment meets the requirements. If it meets the requirements, no processing will be done. If the preliminary temperature adjustment does not meet the requirements, it will enter the temperature correction module through the calculation of the temperature analysis module and the compensation calculation module, so that the temperature correction module adjusts and corrects the temperature.

[0018] The key of the temperature analysis module lies in the spatial analysis of the distortion map. The control platform establishes a two-dimensional polar coordinate system and converts the body surface contour obtained by laser scanning into a polar radius parametric equation. When the displacement of the flexible sensor is detected, the Sobel operator is used for edge detection, and the sampling area is divided at intervals of 0.5cm along the displacement direction. The temperature gradient of the sampling area is calculated by the central difference method, and a time decay factor is introduced to calculate the compensation amount. Based on the calculation results, a displacement-humidity coupling model is constructed through a multi-channel neural network architecture to correct and adjust the power of the inflatable device.

[0019] Refer to Figure 2 As shown, combining the body temperature data inside the initial multimodal data, autonomously adjusting the heating power of the inflatable device for temperature regulation. At this time, the heating power is recorded as the target power, which specifically includes: Extract the temperature data values of multiple monitoring points collected by the flexible temperature sensor in the initial multimodal data at the same time, calculate the average value of the multiple temperature data values, and record it as the average temperature; Obtain the target temperature independently input by the user at the input end of the control platform, calculate the absolute value of the difference between the target temperature and the average temperature, and obtain the temperature deviation; Through historical data fitting, extract the historical heat capacity constant, heat conversion efficiency, and environmental loss compensation coefficient. Multiply the temperature difference value by the heat capacity constant to obtain the theoretical heat energy demand. Divide the theoretical heat energy demand by the product of the heat conversion efficiency and the environmental loss compensation coefficient to obtain the heating power. The heat capacity constant is the heat storage capacity parameter of the heating system obtained by multiplying the specific heat capacity of the material by the mass of the material. The heat conversion efficiency is the efficiency parameter for converting electrical energy into heat energy. The environmental loss compensation coefficient is the proportion value of heat loss during the transmission process; The control platform adjusts the power of the inflatable device to the heating power for temperature adjustment operation. At this time, the heating power is recorded as the target power.

[0020] The sensors are distributed on the silicone substrate in a 5×5 matrix form, with an adjacent node spacing of 2.5 cm. Data synchronization acquisition is achieved through the I2C bus. The weighted average algorithm built into the control platform adopts a regional grading strategy: the sensors within 3 cm of the human core area (chest and abdomen) are classified as type A monitoring points and given a weight coefficient of 0.7; the sensors in the edge area are used as type B monitoring points and given a weight of 0.3. This can control the deviation between the calculated average temperature and the core body temperature within ±0.15°C. The regional grading weighted algorithm reduces the error in the characterization of the average temperature and can still maintain a certain measurement accuracy under the local temperature fluctuations caused by human movement.

[0021] The control platform establishes a training set containing multiple groups of historical operation data. Using the temperature difference as the independent variable and the actual energy consumption as the dependent variable, a linear regression model is constructed to calculate the target power. Among them, the heat capacity constant is obtained by measuring the specific heat capacity curve of the phase change material in the 25 - 40°C range with a DSC differential scanning calorimeter and combining the mass of 500 g of phase change microcapsules in the inflatable device. The heat conversion efficiency is calibrated by measuring the resistance of the heating wire using the four-wire method and calculating it in combination with real-time voltage monitoring. The environmental loss compensation coefficient is finally determined by measuring the heat dissipation loss in the range of wind speed from 0 to 2 m / s by building an infrared thermal imaging experimental platform. After introducing the real-time wind speed sensing data into the environmental loss compensation coefficient, the steady-state control accuracy of the system in the forced convection environment is improved.

[0022] Refer to Figure 3 As shown, the position coordinates of the flexible sensors in the two data are marked, and the corresponding temperature data are projected to form a temperature monitoring distortion map, which specifically includes: Connect the position coordinates of the flexible sensors in the two data, mark the midpoint of the connection line of the two flexible sensors, read the distance from the midpoint to the flexible sensor, and draw a circle with the flexible sensor coordinates as the center and the distance from the midpoint to the flexible sensor as the radius, which is recorded as the flexible sensor monitoring area; Select the two coordinate points closest to and farthest from the horizontal axis on the two flexible sensor detection areas, draw horizontal lines parallel to the horizontal axis, select the two coordinate points closest to and farthest from the vertical axis on the two flexible sensor detection areas, draw vertical lines perpendicular to the horizontal axis, and connect the four lines to form a rectangular area, which is recorded as the temperature monitoring distortion area; Map the temperature data of each point in the temperature monitoring distortion area to the corresponding coordinates, and merge the mapping layer with the detection area to form a temperature monitoring distortion map.

[0023] The control platform is configured with a laser scanning module with an accuracy of 0.05 mm. The TOF sensor carried by it collects three-dimensional coordinate data at a frequency of 1 kHz, and a Cartesian coordinate system is established by least squares fitting. During the detection of the displacement of the flexible sensor, an improved Hough transform algorithm is used to perform linear fitting on the coordinate points (x1, y1) and (x2, y2) obtained within two monitoring periods, and the displacement vector length is calculated through the formula and the arctangent function is used to calculate the displacement direction angle. The monitoring area is constructed using a circle drawing algorithm. The dynamic coordinate system establishment method based on the displacement vector improves the positioning accuracy of the temperature distortion area, reduces the error compared with the fixed grid method, and at the same time, the polar coordinate boundary scanning algorithm shortens the generation time of the distortion area, meeting the real-time monitoring requirements. With the midpoint coordinate as the center of the circle and the radius . For the determination of the boundary of the temperature distortion area, the system uses a polar coordinate scanning method along the monitoring ring, and searches for extreme points in the directions of θ = 0° and θ = 90° respectively: the coordinate points of the maximum and minimum points in the horizontal axis direction and the coordinate points of the maximum and minimum points in the vertical axis direction are selected through the quicksort algorithm. The algorithm is used for line segment clipping to form a rectangular distortion area with a side length of 2r × 2r. The temperature data mapping uses bilinear interpolation technology to convert discrete sensor data into a temperature field distribution map, and its mathematical expression is , where are the temperature values of adjacent sensor nodes. Under the condition that the sensor spacing is 5 cm, the bilinear interpolation model reduces the temperature field reconstruction error and improves the accuracy compared with the traditional nearest neighbor interpolation.

[0024] Referring to Figure 4 shown, the calculation of the temperature compensation amount for the displacement area of the flexible sensor through the temperature compensation equation based on the temperature monitoring distortion map specifically includes: Obtain the coordinate positions of the flexible sensor in the initial multimodal data and the target multimodal data, and read the displacement vector length and the angle between the displacement vector and the coordinate axis between the two coordinate positions; According to the boundary of the rectangular area of the temperature monitoring distortion map, divide the rectangular area proportionally according to the displacement length along the displacement vector direction to form multiple temperature gradient sampling areas; Extract the maximum and minimum values of the temperature data in each sampling area, and divide the difference between the maximum and minimum values by the displacement vector length to obtain the temperature gradient change value; Based on the displacement vector length, the angle between the displacement vector and the coordinate axis, the maximum and minimum values of the temperature data, calculate the temperature compensation amount through the compensation coefficient correction formula; Accumulate the compensation amounts of each sampling area to obtain the total temperature compensation amount and transmit it to the compensation database of the control platform to update the compensation parameters.

[0025] The temperature compensation amount calculation module adopts a method combining spatial thermodynamic analysis and dynamic partition sampling. The laser displacement sensor array with an in-built gyroscope can detect the three-dimensional displacement of the flexible sensor in real time. Cooperating with the processor to solve the displacement vector parameters, the vector synthesis formula is used: displacement length , included angle to calculate the displacement vector, where are the coordinate offsets respectively. At the same time, the system adopts the triangulation algorithm to divide the rectangular distortion area into N equally spaced units (N = 5×L, L is the displacement length) along the displacement vector direction. The size of each unit is accurate to 1 mm² to ensure the accuracy of the temperature field spatial resolution.

[0026] A local coordinate system is established along the displacement direction in the temperature gradient sampling area. The sliding window filtering technology (window width 3×3 pixels) is used to extract the temperature extreme values in each unit. The included angle between the displacement vector and the coordinate axis compensates for the angular deviation between the displacement direction and the main heat conduction direction. The dynamic partition algorithm improves the calculation accuracy of the temperature gradient, reduces the error compared with the fixed grid method. At the same time, the sliding window filtering technology effectively suppresses the sensor noise and improves the signal-to-noise ratio of the temperature extreme value detection.

[0027] Referring to Figure 5 as shown, calculating the temperature compensation amount based on the displacement vector length, the included angle with the coordinate axis, the maximum and minimum values of the temperature data through the compensation coefficient correction formula specifically includes: Obtain the temperature gradient change value, the displacement vector length, and the included angle between the displacement vector and the coordinate axis. Denote the displacement vector length as L and the included angle between the displacement vector and the coordinate axis as θ. Calculate the product of L and sinθ to obtain the actual displacement amount of the flexible sensor in the vertical monitoring direction. Multiply the actual displacement amount by the temperature gradient change value to obtain the temperature deviation value caused by the displacement of the flexible sensor, denoted as the interval temperature difference; Based on historical data fitting to determine the time decay coefficient, obtain the interval time between the initial multimodal data and the target multimodal data. Multiply the difference between the maximum and minimum values of the temperature data in each sampling area, the interval time, and the time decay coefficient to obtain the time decay term compensation value. The time decay coefficient is the temperature lag effect coefficient caused by the thermal inertia of the system, and the time decay term compensation value is the error caused by the response delay of the control platform.

[0028] The digital inclinometer sensor and the laser ranging module with a sampling rate of 1000 Hz obtain the three-dimensional displacement parameters in real time through the quaternion attitude solution algorithm. Decompose the total displacement L into the monitoring direction component and the vertical direction component , where The angular measurement error is controlled within ±0.5°. The calculation of the temperature gradient change value adopts the improved central difference method, and its mathematical expression is , where n = 5 is the number of sampling areas divided along the displacement direction.

[0029] The determination of the time decay coefficient is completed by the step response experiment method: Apply a 10 °C step temperature change in a constant temperature environment, and record the time required for the system to reach 63.2% of the steady state as the thermal time constant. The experimentally measured τ = 8.3 ± 0.5 s. The calculation formula for the time decay term compensation value is , where is the extreme value of the temperature difference in the sampling area, is the data acquisition interval, is the dynamic decay factor. The actual displacement compensation amount is calculated using the heat flux density integral formula: , where is the air thermal conductivity, is the effective contact area of the sensor (r = 5 mm).

[0030] Referring to Figure 6 as shown, constructing a displacement-humidity coupling compensation model based on the humidity data in the temperature compensation amount, initial multimodal data, and target multimodal data specifically includes: The control platform aggregates the total temperature compensation amount, interval temperature difference, time decay term compensation value, displacement vector length, angle with the coordinate axis, temperature gradient change value, initial humidity value, and target humidity value to form an input database, and constructs a displacement-humidity coupling compensation model. The displacement-humidity coupling compensation model includes a data preprocessing module, a multimodal feature fusion module, and a power correction calculation module; The data preprocessing module normalizes all data in the input database, and calculates the humidity change gradient based on the humidity data in the initial multimodal data and target multimodal data, and feeds it back to the multimodal feature fusion module; The multimodal feature fusion module adopts a cascaded neural network structure, including a temperature compensation branch, a displacement vector branch, and a humidity change branch. The output ends of each branch perform cross-modal fusion to obtain a fused feature vector and feed it back to the interior of the power correction calculation module; The power correction calculation module generates correction parameters including power values and time action intervals by performing matrix multiplication operations on the target power and the increment coefficient.

[0031] The control platform aggregates parameters such as the total temperature compensation amount, interval temperature difference, and time decay term compensation value to form an input database, and performs feature fusion using a three-layer cascaded neural network architecture. The data preprocessing module normalizes the input parameters by the standardization method, and the humidity change gradient calculation uses the differential method: Extract the initial humidity value H0 and the target humidity value H1, and combine the time stamp difference , the instantaneous humidity change rate ΔH / Δt is calculated. The multi-modal feature fusion module adopts a heterogeneous neural network architecture. The temperature compensation branch is configured with a three-layer fully connected network with ReLU activation function. The displacement vector branch uses an LSTM network to capture the temporal displacement features. The humidity change branch extracts the humidity temporal correlation characteristics through a one-dimensional convolution kernel (width 5, stride 1). The three-way feature vectors are cross-modally fused through dimension expansion in the splicing layer to form a 128-dimensional feature vector.

[0032] The power correction calculation module adopts a dynamic matrix adjustment mechanism. The incremental coefficient matrix is obtained through offline training: a supervised learning model is constructed based on historical operation data, with the ratio of the actual correction power to the target power as the power adjustment factor and the ratio of the actual response time to the theoretical time as the time adjustment factor, and the least squares method is used to optimize the matrix parameters. During implementation, the target power P and the unit constant 1 form a row vector [P 1], which is multiplied by the trained 2×2 incremental matrix, and the corrected power P' = aP + b and the time action interval T' = cP + d are output, where a, b, c, and d are matrix elements, thereby improving the compensation accuracy and shortening the response time. Through the coupling calculation of the humidity change rate and the displacement vector, the thermal inertia lag effect is effectively overcome. Its dynamic matrix adjustment mechanism enables the power control to have self-adaptability and still maintain a certain temperature control accuracy in complex body movement scenarios.

[0033] Refer to Figure 7 As shown, the data preprocessing module normalizes all data in the input database and calculates the humidity change gradient based on the humidity data in the initial multi-modal data and the target multi-modal data, specifically including: Obtain the initial humidity value in the initial multi-modal data and the target humidity value in the target multi-modal data, calculate the difference between the target humidity value and the initial humidity value, and record it as the humidity change value; Extract the timestamps of the initial multi-modal data and the target multi-modal data, read the time interval between the two, and divide the humidity change value by the time interval to obtain the humidity change gradient.

[0034] In the data preprocessing stage, the dynamic humidity gradient calculation and multi-modal normalization fusion technology are adopted to achieve precise standardization of compensation parameters. Through a micro-capacitive humidity sensor array with a sampling rate of 100 Hz, the humidity detection resolution reaches 0.1%RH, and a GPS synchronous clock module is used to obtain timestamp data with an accuracy of 0.1 ms. During its use, first, the humidity value H0 at the initial time t0 and the humidity value H1 at the target time t1 are synchronously read through the I2C bus, and the absolute humidity difference ΔH = H1 - H0 is calculated using double-precision floating-point arithmetic. The time interval is calculated as ΔT = (t1 - t0) / 1000 to convert milliseconds to seconds. When ΔT < 50 ms, the previous valid gradient value is automatically enabled. The humidity change gradient is calculated using the differential algorithm: ∇H = ΔH / ΔT, and the calculation result is stored as a 32-bit floating-point number. The instantaneous fluctuation noise is eliminated through sliding window filtering (window width of 5 sampling points). The data normalization process constructs a feature matrix for heterogeneous data such as temperature compensation amounts and displacement vectors, and performs three-sigma truncation processing through the formula X' = (X - μ) / (3σ), where μ and σ are the rolling mean and standard deviation from the historical database respectively. Thus, the accuracy of characterizing humidity changes is improved through real-time differential calculation, and the synchronization error of timestamps is reduced.

[0035] Refer to Figure 8 As shown, the multi-modal feature fusion module adopts a cascaded neural network structure, including a temperature compensation branch, a displacement vector branch, and a humidity change branch. The output ends of each branch perform cross-modal fusion to obtain a fused feature vector, which is fed back to the power correction calculation module. Specifically, it includes: The multi-modal feature fusion module records the total temperature compensation amount, the interval temperature difference, and the time decay term compensation value as temperature inputs to form a temperature compensation branch, records the displacement vector length, the angle with the coordinate axis, and the temperature gradient change value as displacement inputs to form a displacement vector branch, and records the initial humidity value and the target humidity value as humidity inputs to form a humidity change branch; The input layer of the temperature compensation branch receives the normalized total temperature compensation amount, and outputs a temperature feature vector through the activation function in the control platform; The input layer of the displacement vector branch receives the displacement vector length, the angle, and the time decay term compensation value, and extracts the displacement dynamic feature vector through time series modeling; The humidity change branch adopts a convolutional neural network structure. The input layer receives the humidity change gradient and historical humidity fluctuation parameters, captures the humidity time series correlation characteristics through the convolutional kernel, and outputs a humidity feature vector; The temperature feature vector, the displacement feature vector, and the humidity feature vector are tensor-concatenated in the dimension direction to obtain a fused feature vector.

[0036] The temperature compensation branch adopts a three-layer fully connected network structure. The input layer receives the total temperature compensation amount, interval temperature difference, and time decay term that have been normalized. The hidden layer is configured with 128 neuron nodes, and an activation function is used to prevent gradient vanishing, outputting a 32-dimensional temperature feature vector. The displacement vector branch deploys a bidirectional LSTM network. The input layer receives the displacement length, included angle, and time decay term, sets 64 memory units, captures the temporal characteristics of the displacement trajectory through a time step unfolding structure, and the output end uses a max pooling layer to extract a 24-dimensional displacement dynamic feature vector. The humidity change branch constructs a one-dimensional convolutional neural network. The input layer receives the humidity gradient ∇H and a temporal matrix containing the humidity fluctuations of the previous 5 cycles, configures 3 convolutional layers, expands the receptive field through dilated convolution, and finally outputs a 16-dimensional humidity feature vector. The feature fusion layer uses a tensor splicing technique to splice the temperature, displacement, and humidity feature vectors after zero-padding alignment in the second dimension to form a 72-dimensional fused feature vector. During actual processing, the temperature branch uses a batch normalization layer to accelerate convergence, the displacement branch introduces a dropout layer to prevent overfitting, and the humidity branch uses causal convolution to ensure temporal causality. Thus, the displacement tracking error is reduced and the humidity response delay is shortened through the hybrid architecture. The feature dimension expansion also solves the problem of dimension mismatch during heterogeneous data fusion, improving the inference speed of the neural network.

[0037] Refer to Figure 9 As shown, the power correction calculation module generates a two-dimensional correction parameter including a power value and a time action interval by performing a matrix multiplication operation on the target power and the increment coefficient, specifically including: Input the fused feature vector into the power correction calculation module, and summarize to obtain an increment coefficient matrix composed of a power adjustment factor and a time adjustment factor. The increment coefficient matrix is a real matrix with two rows and two columns. The elements in the first row are the power adjustment factors, and the elements in the second row are the time adjustment factors. The power adjustment factor is the weight coefficient of the corrected target power obtained by dividing the actual corrected power in the historical data by the target power. The time adjustment factor is the coefficient of the duration of the control corrected power action obtained by dividing the actual time to reach the target temperature in the historical data by the theoretical time; Combine the target power and the constant 1 into a row vector with two columns, perform a matrix multiplication operation on the row vector and the increment coefficient matrix, and obtain a two-dimensional correction parameter with two columns. The first column is the corrected power value, and the second column is the time interval of the power action; Feed back the corrected power value and the time interval to the control platform to update the power control instruction of the inflation device.

[0038] The power correction calculation module is deployed on an FPGA hardware accelerator, configured with a double-precision floating-point arithmetic unit, and the increment coefficient matrix is dynamically updated through a supervised learning model: based on multiple groups of operation records in the historical database, matrix parameters are obtained through multiple linear regression training, where the power adjustment factor α is the power adjustment factor, p_actual and p_target are the actual power and the target power, and the time adjustment factor , during training, a regularization term is added to prevent overfitting. During processing, the fused feature vector is mapped to a 4D latent space through a fully connected layer, and after being normalized by the Sigmoid function, a 2×2 increment matrix is generated, and its mathematical expression is , where is the bias compensation term. The matrix multiplication operation is implemented using a parallel pipeline architecture. When the target power P is input, a row vector [P 1] is constructed and multiplied by the increment matrix: the corrected power , the time interval . The system is specially designed with an overflow protection mechanism. When P' exceeds the maximum power of the device, which is 2000W, it is automatically limited, and the time interval is processed by piecewise linearization. Thus, the power adjustment response speed can be improved through the matrix transformation mechanism, and the temperature overshoot in the case of sudden human movement is reduced. The design of the bias compensation term effectively solves the residual problem of the nonlinear system, improves the power compensation accuracy in a low-temperature environment, and reduces the time interval prediction error. The high-frequency matrix operation implemented by hardware acceleration enables the system to complete the closed-loop adjustment of power parameters within a 100ms cycle, with higher efficiency than the traditional software implementation method.

[0039] In summary, the advantages of the present invention are as follows: Through the displacement-humidity coupling compensation model and the dynamic power correction mechanism, accurate body temperature maintenance is achieved.

[0040] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. What is described in the above embodiments and the specification is only the principle of the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements fall within the scope of the present invention claimed. The scope of protection claimed by the present invention is defined by the appended claims and their equivalents.

Claims

1. An inflatable heating system with linked wireless body temperature monitoring and automatic temperature adjustment, characterized in that: Including: Collection module: Collects the user's initial multimodal data through a wearable multimodal monitoring device. The multimodal monitoring device consists of a flexible temperature sensor, a laser scanning module, and a micro humidity sensor. The initial multimodal data includes temperature data, scanning data, and humidity data; Preliminary temperature adjustment module: Obtains historical multimodal data to establish a control platform, inputs the currently required target temperature inside the control platform, combines with the body temperature data inside the initial multimodal data, and independently adjusts the heating power of the inflatable device for temperature adjustment. At this time, the heating power is recorded as the target power. After the temperature adjustment is completed, the multimodal monitoring device collects multimodal data again to form target multimodal data; Temperature analysis module: The control platform, based on the scanning data in the initial multimodal data, establishes a human body contour reference model and projects it onto a plane to establish a coordinate system, and projects the target multimodal data onto the coordinate system, marks the position coordinates of the flexible sensors in the two data, and projects the corresponding temperature data to form a temperature monitoring distortion map; Compensation calculation module: Calculates the temperature compensation amount of the displacement area of the flexible sensor based on the temperature monitoring distortion map through a temperature compensation equation; Temperature correction module: Based on the temperature compensation amount, the humidity data in the initial multimodal data and the target multimodal data, constructs a displacement-humidity coupling compensation model, imports the target power into the displacement-humidity coupling compensation model for updating to obtain the corrected power, and the control platform controls the inflatable device to correct and adjust the temperature based on the corrected power.

2. The inflatable heating system for linked wireless body temperature monitoring and automatic temperature setting adjustment according to claim 1, wherein, The combination of the body temperature data inside the initial multimodal data, and independently adjusts the heating power of the inflatable device for temperature adjustment. At this time, the heating power is recorded as the target power specifically includes: Extracts the temperature data values of multiple monitoring points collected by the flexible temperature sensor in the initial multimodal data at the same time, calculates the average value of the multiple temperature data values, and records it as the average temperature; Obtains the target temperature independently input by the user at the input end of the control platform, calculates the absolute value of the difference between the target temperature and the average temperature to obtain the temperature deviation; Through historical data fitting, extracts the historical heat capacity constant, heat conversion efficiency, and environmental loss compensation coefficient, multiplies the temperature difference value by the heat capacity constant to obtain the theoretical heat energy demand, and divides the theoretical heat energy demand by the product of the heat conversion efficiency and the environmental loss compensation coefficient to obtain the heating power. The heat capacity constant is the heat storage capacity parameter of the heating system obtained by multiplying the specific heat capacity of the material by the mass of the material. The heat conversion efficiency is the efficiency parameter of converting electrical energy into heat energy, and the environmental loss compensation coefficient is the proportion value of heat loss during the transmission process; Adjusts the power of the inflatable device to the heating power through the control platform for temperature adjustment operation. At this time, the heating power is recorded as the target power.

3. The inflatable heating system for linked wireless body temperature monitoring and automatic temperature setting adjustment according to claim 2, wherein The marking of the position coordinates of the flexible sensors in the two data, and the projection of the corresponding temperature data to form a temperature monitoring distortion map specifically includes: Obtains the position coordinates of the flexible sensors in the two data and connects them, marks the midpoint of the connection line of the two flexible sensors, reads the distance from the midpoint to the flexible sensor, and makes a circle with the flexible sensor coordinates as the center and the distance from the midpoint to the flexible sensor as the radius, which is recorded as the flexible sensor monitoring area; Select two coordinate points closest to and farthest from the horizontal axis on the two flexible sensor detection areas, draw horizontal lines parallel to the horizontal axis, select two coordinate points closest to and farthest from the vertical axis on the two flexible sensor detection areas, draw vertical lines perpendicular to the horizontal axis, and connect the four lines to form a rectangular area, denoted as the temperature monitoring distortion area; Map the temperature data of each point in the temperature monitoring distortion area to the corresponding coordinates, and merge the mapped layer with the detection area to form a temperature monitoring distortion map.

4. The inflatable warming system for linked wireless body temperature monitoring and automatic temperature setting adjustment according to claim 3, wherein The calculation of the temperature compensation amount for the displacement area of the flexible sensor based on the temperature monitoring distortion map through the temperature compensation equation specifically includes: Obtain the coordinate positions of the flexible sensor in the initial multi-modal data and the target multi-modal data, and read the length of the displacement vector and the angle between the displacement vector and the coordinate axes between the two coordinate positions; According to the boundaries of the rectangular area of the temperature monitoring distortion map, divide the rectangular area proportionally according to the displacement length along the displacement vector direction to form multiple temperature gradient sampling areas; Extract the maximum and minimum values of the temperature data in each sampling area, divide the difference between the maximum and minimum values by the length of the displacement vector to obtain the temperature gradient change value; Based on the length of the displacement vector, the angle between the displacement vector and the coordinate axes, the maximum and minimum values of the temperature data, calculate the temperature compensation amount through the compensation coefficient correction formula; Accumulate the compensation amounts of each sampling area to obtain the total temperature compensation amount and transmit it to the compensation database of the control platform to update the compensation parameters.

5. The inflatable warming system for linked wireless body temperature monitoring and automatic temperature setting adjustment according to claim 4, characterized in that The calculation of the temperature compensation amount based on the length of the displacement vector, the angle with the coordinate axes, the maximum and minimum values of the temperature data through the compensation coefficient correction formula specifically includes: Obtain the temperature gradient change value, the length of the displacement vector, and the angle between the displacement vector and the coordinate axes. Denote the length of the displacement vector as L and the angle between the displacement vector and the coordinate axes as θ. Calculate the product of L and sinθ to obtain the actual displacement amount of the flexible sensor in the vertical monitoring direction. Multiply the actual displacement amount by the temperature gradient change value to obtain the temperature deviation value caused by the displacement of the flexible sensor, denoted as the interval temperature difference; Based on the historical data fitting to determine the time decay coefficient, obtain the interval time between the initial multi-modal data and the target multi-modal data, and multiply the difference between the maximum and minimum values of the temperature data in each sampling area, the interval time, and the time decay coefficient to obtain the time decay term compensation value. The time decay coefficient is the temperature lag effect coefficient caused by the system thermal inertia, and the time decay term compensation value is the error caused by the response delay of the control platform.

6. The inflatable warming system for linked wireless body temperature monitoring and automatic temperature setting adjustment according to claim 5, characterized in that, The construction of the displacement-humidity coupling compensation model based on the temperature compensation amount, the humidity data in the initial multi-modal data and the target multi-modal data specifically includes: The control platform aggregates the total temperature compensation amount, the interval temperature difference, the time decay term compensation value, the length of the displacement vector, the angle with the coordinate axes, the temperature gradient change value, the initial humidity value, and the target humidity value to form an input database, and constructs a displacement-humidity coupling compensation model. The displacement-humidity coupling compensation model includes a data preprocessing module, a multi-modal feature fusion module, and a power correction calculation module; The data preprocessing module normalizes all the data in the input database, and calculates the humidity change gradient based on the humidity data in the initial multi-modal data and the target multi-modal data, and feeds it back to the multi-modal feature fusion module; The multi-modal feature fusion module adopts a cascaded neural network structure, including a temperature compensation branch, a displacement vector branch, and a humidity change branch. Cross-modal fusion is performed at the output ends of each branch to obtain a fused feature vector and feed it back into the power correction calculation module; The power correction calculation module generates correction parameters including power values and time action intervals by performing matrix multiplication on the target power and the increment coefficient.

7. An inflatable warming system for linked wireless body temperature monitoring and automatic temperature setting adjustment according to claim 6, characterized in that, The data preprocessing module normalizes all the data in the input database, and calculating the humidity change gradient based on the humidity data in the initial multi-modal data and the target multi-modal data specifically includes: Obtain the initial humidity value in the initial multi-modal data and the target humidity value in the target multi-modal data, calculate the difference between the target humidity value and the initial humidity value, and record it as the humidity change value; Extract the timestamps of the initial multi-modal data and the target multi-modal data, read the time interval between them, and divide the humidity change value by the time interval to obtain the humidity change gradient.

8. The inflatable heating system with linked wireless body temperature monitoring and automatic temperature adjustment according to claim 7, characterized in that: The multi-modal feature fusion module adopts a cascaded neural network structure, including a temperature compensation branch, a displacement vector branch, and a humidity change branch. Cross-modal fusion is performed at the output ends of each branch to obtain a fused feature vector and feed it back into the power correction calculation module specifically includes: The multi-modal feature fusion module records the total temperature compensation amount, the interval temperature difference, and the time decay term compensation value as the temperature input to form the temperature compensation branch, records the displacement vector length, the angle with the coordinate axis, and the temperature gradient change value as the displacement input to form the displacement vector branch, and records the initial humidity value and the target humidity value as the humidity input to form the humidity change branch; The input layer of the temperature compensation branch receives the normalized total temperature compensation amount, and outputs a temperature feature vector through the activation function in the control platform; The input layer of the displacement vector branch receives the displacement vector length, the angle, and the time decay term compensation value, and extracts the displacement dynamic feature vector through time series modeling; The humidity change branch adopts a convolutional neural network structure. The input layer receives the humidity change gradient and the historical humidity fluctuation parameters, captures the humidity time series correlation characteristics through the convolution kernel, and outputs a humidity feature vector; Tensor splice the temperature feature vector, the displacement feature vector, and the humidity feature vector in the dimension direction to obtain a fused feature vector.

9. The inflatable temperature-raising system for linked wireless body temperature monitoring and automatic temperature setting adjustment according to claim 8, characterized in that, The power correction calculation module generates two-dimensional correction parameters including power values and time action intervals by performing matrix multiplication on the target power and the increment coefficient specifically includes: Input the fused feature vector into the power correction calculation module, and summarize to obtain an incremental coefficient matrix composed of a power adjustment factor and a time adjustment factor. The incremental coefficient matrix is a real matrix with two rows and two columns. The elements in the first row are the power adjustment factors, and the elements in the second row are the time adjustment factors. The power adjustment factor is the weight coefficient for correcting the target power obtained by dividing the actual corrected power in historical data by the target power. The time adjustment factor is the coefficient for controlling the duration of the corrected power action obtained by dividing the actual time to reach the target temperature in historical data by the theoretical time. Combine the target power and the constant 1 into a row vector with two columns, and perform a matrix multiplication operation on the row vector and the incremental coefficient matrix to obtain a two-dimensional correction parameter with two columns in one row. The first column is the corrected power value, and the second column is the time interval for the power action. Feed back the corrected power value and the time interval to the control platform to update the power control instruction of the inflation device.

Images