Multivariable composite sensing health detection management system and intelligent terminal

Abstract

The invention relates to a multivariable composite sensing health detection management system and an intelligent terminal. The multivariable composite sensing health detection management system comprises a multi-source input module, an algorithm processing module, a power supply module and a wireless communication module. The acquisition signal output end of each independent acquisition source in the multi-source input module is connected with the acquisition signal input end of the algorithm processing module; a data communication end of the algorithm processing module is connected with a first data communication end of the wireless communication module, and a wireless communication path is established between a second data communication end of the wireless communication module and a remote terminal; and the algorithm processing module identifies the multi-source acquisition signals and local user basic information, and pushes corresponding detection results. Through the technical means of multi-sensor data fusion, intelligent algorithm recognition, remote communication transmission and the like, the comprehensiveness, accuracy and user experience of health detection are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of a health detection and management system with multi-variable composite sensing, and more particularly to a health detection and management system with multi-variable composite sensing and an intelligent terminal. Background Art

[0002] Currently, in the field of Internet of Things (IoT) smart wearable devices, such as smartwatches, smart glasses and other terminals, although health management functions such as heart rate monitoring, step counting, blood oxygen saturation (SpO2) measurement and psychological stress assessment have been widely integrated, at present, the data acquisition, signal processing, calibration and calculation methods of each sensor operate independently, resulting in a lack of a unified fusion strategy between data. Due to differences in the sampling frequencies, data formats and algorithm models of different sensors, while the device provides multi-dimensional health tracking, its overall measurement accuracy, data consistency and user experience are still limited to a certain extent. In addition, the current presentation methods of health data are relatively scattered, making it difficult for users to establish intuitive and relevant insights between multiple health parameters, restricting the ability of in-depth analysis and comprehensive evaluation of data, and thus affecting the practical application value of smart wearable devices in personalized health management and continuous user engagement. Summary of the Invention

[0003] In order to solve the problems of independent operation of data acquisition and processing methods of each sensor in existing smart wearable devices, limited measurement accuracy, insufficient data fusion ability and fragmented presentation of health information, the present application provides a health detection and management system with multi-variable composite sensing and an intelligent terminal.

[0004] A health detection and management system with multi-variable composite sensing, the health detection and management system with multi-variable composite sensing includes a multi-source input module, an algorithm processing module, a power supply module and a wireless communication module; The acquisition signal output ends of each independent acquisition source in the multi-source input module are all connected to the acquisition signal input end of the algorithm processing module, so that the algorithm processing module can obtain the acquisition signal, and the acquisition signal at least includes raw health data; The power output end of the power supply module outputs power for power supply; The data communication end of the algorithm processing module is connected to the first data communication end of the wireless communication module, and the second data communication end of the wireless communication module establishes a wireless communication path with a remote terminal; The algorithm processing module is used to identify at least the acquisition signals from multiple sources and the local user basic information through a multi-source recognition algorithm model, and push the corresponding detection results, and the detection results include health suggestions, medical suggestions and preventive suggestions.

[0005] By adopting the above technical solution, through integrating a multi-source input module, an algorithm processing module, a power supply module and a wireless communication module, the unified acquisition, processing and transmission of various sensor data are realized, thus solving the problems of independent operation of each sensor data, limited measurement accuracy, insufficient data fusion ability and fragmented health information in existing intelligent wearable devices. The system adopts a multi-source recognition algorithm model to fuse and analyze signals from different acquisition sources, and combines the user's basic information stored locally to generate personalized health detection results, including health suggestions, medical suggestions and preventive suggestions, improving the depth and accuracy of data analysis. At the same time, a data interaction channel for remote terminals is established through the wireless communication module, enabling the detection results to be transmitted to an external platform in real time, realizing remote health management and intelligent decision-making. In addition, the power supply module ensures the stability of the device during long-term use, guaranteeing that the system can continuously perform efficient data processing and health monitoring. This solution improves the comprehensiveness, accuracy and user experience of health detection through technical means such as multi-sensor data fusion, intelligent algorithm recognition and remote communication transmission.

[0006] Preferably, the multi-source input module includes: A triaxial acceleration sensor for capturing the user's motion state data and posture change data; A photoplethysmogram sensor for collecting non-invasive heart rate monitoring data and blood oxygen saturation monitoring data; A temperature sensor for monitoring environmental temperature data and skin surface temperature data.

[0007] By adopting the above technical solution, a triaxial acceleration sensor, a photoplethysmogram sensor and a temperature sensor are integrated, enabling the system to simultaneously collect the user's motion state, heart rate, blood oxygen saturation, and environmental and skin surface temperature data. The triaxial acceleration sensor can accurately capture the user's motion state and posture changes, the photoplethysmogram sensor can provide non-invasive heart rate monitoring and blood oxygen saturation detection, and the temperature sensor can simultaneously monitor the environmental temperature and the user's skin surface temperature, thus ensuring multi-dimensional coverage of health data. Through the collaborative work of these multi-source sensors, the system can provide accurate data input under different physiological states, thereby improving the accuracy and adaptability of health monitoring.

[0008] Preferably, the wireless communication module is a Bluetooth beacon module.

[0009] By adopting the above technical solution, it is possible to perform low-power and efficient wireless communication with a mobile terminal or other intelligent devices. The Bluetooth beacon module can maintain stable data transmission in the low-power mode, and at the same time has a relatively long communication range and a relatively fast data transmission rate, so as to ensure that health data can be synchronized to the remote terminal in real time and reliably. Through this wireless communication method, users can access personal health data at any time, improving the usability and convenience of the system, while ensuring the stability and security of health data transmission.

[0010] Preferably, the algorithm processing module includes: An acquisition unit, configured to collect original health data in real time, preprocess and extract key health features from the original health data, and substitute the key health features into a multi-source recognition algorithm model that has been deployed and trained in an intelligent terminal device to generate corresponding user behavior data; A customization unit, configured to call corresponding user basic information in the local database of the intelligent terminal device, and health management settings customized based on the user basic information, where the health management settings at least include health goals and warning thresholds; A screening unit, connected to the customization unit, and configured to screen normal behavior data and abnormal behavior data from the user behavior data, where the current deviation value of the abnormal behavior data generated relative to the mean value of each normal behavior data exceeds a preset deviation threshold; A first push unit, connected to the screening unit, and configured to push corresponding health suggestions and / or perform corresponding target intervention operations according to a first comparison result between the health goal and the corresponding normal behavior data; A second push unit, connected to the first push unit, and configured to push corresponding trend prediction information, and medical suggestions or prevention suggestions generated for the trend prediction information according to a second comparison result between the warning threshold and the corresponding abnormal behavior data.

[0011] By adopting the above technical solution, the algorithm processing module can obtain the original health data in real time through the acquisition unit, and preprocess and extract features from the data, so that the data has been denoised and the features have been optimized before entering the multi-source recognition algorithm model, improving the recognition accuracy of the model. The customization unit calls the basic information in the user's local database, and combines personalized health goals and warning thresholds to make health management more accurate and targeted. The screening unit can distinguish normal behavior data from abnormal behavior data to ensure that the system can accurately detect the change trend of the user's health status. The first push unit pushes health suggestions or performs target intervention operations according to the comparison result between the normal behavior data and the health goal, while the second push unit provides trend prediction information and pushes medical suggestions or preventive suggestions according to the comparison result between the abnormal behavior data and the warning threshold. This solution improves the intelligent level of health data analysis through multi-level analysis and judgment, and realizes personalized health management and early risk warning.

[0012] Preferably, the key health features in the acquisition unit at least include gait features, heart rate variability features, and sleep features. The acquisition unit includes: The first determination subunit is used to determine the matching code of each piece of the original health data, and the matching code is determined based on the attribute information of the independent acquisition source; The second determination subunit is connected to the first determination subunit and is used to determine the corresponding data extraction rule according to the matching code; The first extraction subunit is connected to the second determination subunit and is used to extract the corresponding gait features from the original health data if the data extraction rule is a gait extraction rule; The second extraction subunit is connected to the first extraction unit and is used to extract the corresponding heart rate variability features from the original health data if the data extraction rule is a heart rate variability extraction rule; The third extraction subunit is connected to the second extraction unit and is used to extract the corresponding sleep features from the original health data if the data extraction rule is a sleep extraction rule.

[0013] By adopting the above technical solution, the health features extracted by the acquisition unit include gait features, heart rate variability features, and sleep features, enabling the system to comprehensively monitor the user's daily exercise, physiological state, and sleep quality. The first determination subunit and the second determination subunit match the code with the data extraction rules, enabling the data collected by different sensors to adapt to the correct feature extraction method. The first, second, and third extraction subunits separately process gait, heart rate variability, and sleep features, respectively, making the extraction of each feature more accurate and ensuring the reliability and effectiveness of the health monitoring data. Through this technical solution, the system can accurately extract different physiological features, thereby improving the accuracy of health monitoring and providing high-quality data support for subsequent health management and behavior analysis.

[0014] Preferably, the user behavior data in the acquisition unit at least includes gait behavior data, heart rate variability behavior data, and sleep behavior data. The acquisition unit further includes: A third determination subunit, configured to determine the corresponding data recognition rule based on a multi-source recognition algorithm model that has been deployed and trained in the intelligent terminal device; A first generation subunit, connected to the third determination subunit, and configured to, if the data recognition rule is a gait recognition rule, perform pattern recognition and intensity parameter conversion on the gait features based on the gait recognition rule to generate corresponding gait behavior data, where the gait behavior data includes a motion pattern and an intensity parameter corresponding to and quantifiable for the motion pattern; A second generation subunit, connected to the first generation subunit, and configured to, if the data recognition rule is a heart rate variability recognition rule, perform calculation and integration on the heart rate variability features based on the heart rate variability recognition rule to generate corresponding gait behavior data, where the gait behavior data includes heart rate interval change data and an autonomic nervous system activity level corresponding to and quantifiable for the heart rate interval change data; A third generation subunit, connected to the second generation subunit, and configured to, if the data recognition rule is a sleep recognition rule, perform stage recognition and quality assessment level conversion on the sleep features based on the sleep recognition rule to generate corresponding sleep behavior data, where the sleep behavior data includes a quantifiable sleep quality assessment level within each sleep stage.

[0015] By adopting the above technical solution, the system deploys a multi-source recognition algorithm model in the intelligent terminal device to ensure the adoption of accurate recognition rules during the data processing. The third determination subunit determines the data recognition rules according to the model structure. The first generation subunit performs pattern recognition and intensity parameter conversion on the gait features based on the gait recognition rules, making the analysis of the motion state more intuitive and quantitative. The second generation subunit calculates and integrates the heart rate variability, enabling the quantification of the activity level of the autonomic nervous system and providing a more accurate assessment of the heart health status. The third generation subunit performs stage recognition and sleep quality assessment on the sleep features, enabling the system to distinguish the sleep stages in detail and objectively evaluate the sleep quality. Through this solution, the system can effectively identify and quantify the user's motion, physiological state, and sleep quality, providing strong data support for personalized health management.

[0016] Preferably, the algorithm processing module further includes: The first generation unit is used to select a labeled data set from the local experimental source or some open source data sets, and use the labeled data set to train the pre-trained model to generate a corresponding behavior pattern model. The second generation unit is connected to the first generation unit and is used to add multiple sample sources, determine a large-scale labeled data set in each of the sample sources, and use the large-scale labeled data set to train the behavior pattern model to generate a corresponding generalization model. The third generation unit is connected to the second generation unit and is used to select one or more source domains from multiple sample sources based on data distribution similarity, determine the target domain according to the user's basic information, and adapt the generalization model using the transfer learning method to generate an accelerated adaptation model. The fourth generation unit is connected to the third generation unit and is used to establish an incremental learning mechanism in the accelerated adaptation model to generate a trained multi-source recognition algorithm model.

[0017] By adopting the above technical solution, the first generation unit uses the labeled data of the local experimental source or open source data set to train the pre-trained model, enabling it to adapt to the analysis requirements of basic health data. The second generation unit extracts a large-scale labeled data set from multiple sample sources and uses it to train the generalization model to improve the model's adaptability to different users. The third generation unit selects the best source domain data based on data distribution similarity and optimizes the generalization model using the transfer learning method to make it more adaptable to the data distribution of the target user. The fourth generation unit introduces an incremental learning mechanism into the accelerated adaptation model, enabling the system to continuously optimize the recognition ability and continuously improve the accuracy and personalization level of health monitoring as the user's behavior data accumulates. Through this solution, the system is continuously optimized during the training process to ensure its long-term high accuracy and adaptability.

[0018] Preferably, the algorithm processing module further includes: An acquisition unit, configured to acquire corresponding user basic information from the registration window filled out by the user, where the user basic information at least includes physiological information, health background information, and call corresponding historical health information based on the physiological information; A fifth generation unit, connected to the acquisition unit, and configured to generate a customized health goal according to the physiological information and health background information; A sixth generation unit, connected to the fifth generation unit, and configured to generate a customized warning threshold according to the physiological information and historical health information, and a dynamic adjustment mechanism for dynamically adjusting the warning threshold; An integration unit, connected to the sixth generation unit, and configured to integrate the health goal, the warning threshold, and the dynamic adjustment mechanism to generate a corresponding health management setting.

[0019] By adopting the above technical solution, the fifth generation unit sets a personalized health goal based on the user's physiological information and health background, improving the pertinence of health management. The sixth generation unit dynamically adjusts the warning threshold using historical health information and establishes a dynamic adjustment mechanism, enabling the warning system to continuously optimize as the user's state changes. The integration unit further integrates the health goal, the warning threshold, and the dynamic adjustment mechanism to form a systematic health management setting, making the health management process more accurate, dynamic, and personalized.

[0020] Preferably, the first push unit includes: A call subunit, configured to call the quantitative indicators in the normal behavior data; A fourth generation subunit, connected to the call subunit, and configured to match each of the quantitative indicators to the corresponding health goal, where each health goal is respectively corresponding to one or more of the quantitative indicators, and perform a weighted average on the quantitative indicators corresponding to each health goal to generate a corresponding target quantitative value; A fifth generation subunit, connected to the fourth generation subunit, and configured to compare the health goal with the corresponding target quantitative value to generate a corresponding first comparison result, and push corresponding health suggestions and / or perform corresponding target intervention operations according to the first comparison result.

[0021] By adopting the above technical solution, the invocation subunit of the first push unit extracts quantization metrics from normal behavior data, and the fourth generation subunit matches each quantization metric to the corresponding health goal, and calculates the target quantization value by using the weighted average method. The fifth generation subunit compares the calculated target quantization value with the health goal, generates the first comparison result, and accordingly pushes health suggestions or performs corresponding target intervention operations. This solution realizes the accurate monitoring and evaluation of health goals through quantitative analysis of user health data, makes health suggestions more scientific and intelligent, and improves the effectiveness of user health management.

[0022] An intelligent terminal uses a health detection and management system with multi-variable composite sensing.

[0023] In summary, the present application includes at least one of the following beneficial technical effects: By integrating a multi-source input module, an algorithm processing module, a power supply module, and a wireless communication module, the unified collection, processing, and transmission of various sensor data are realized, thereby solving the problems of independent operation of each sensor data, limited measurement accuracy, insufficient data fusion ability, and fragmentation of health information in existing intelligent wearable devices. The system adopts a multi-source recognition algorithm model to fuse and analyze the signals from different acquisition sources, and combines the user's basic information stored locally to generate personalized health detection results, including health suggestions, medical suggestions, and preventive suggestions, improving the depth and accuracy of data analysis. At the same time, a data interaction channel for the remote terminal is established through the wireless communication module, enabling the detection results to be transmitted to the external platform in real time, realizing remote health management and intelligent decision-making. In addition, the power supply module ensures the stability of the device during long-term use, ensuring that the system can continuously perform efficient data processing and health monitoring. This solution improves the comprehensiveness, accuracy, and user experience of health detection through technical means such as multi-sensor data fusion, intelligent algorithm recognition, and remote communication transmission. Description of the Drawings

[0024] Figure 1 is a flowchart of a health detection and management system with multi-variable composite sensing in an embodiment of the present application.

[0025] Figure 2 is a schematic diagram of icon extraction for gait feature extraction in the extraction unit in an embodiment of the present application.

[0026] Description of the reference numerals: 1. Multi-source input module; 2. Algorithm processing module; 3. Power supply module; 4. Wireless communication module. Detailed Description of the Embodiment

[0027] The following further describes the present application in detail with reference to the drawings.

[0028] In an embodiment, as Figure 1As shown, the present application discloses a health detection and management system with multi-variable composite sensing. A health detection and management system with multi-variable composite sensing includes a multi-source input module 1, an algorithm processing module 2, a power supply module 3, and a wireless communication module 4; The acquisition signal output ends of each independent acquisition source in the multi-source input module 1 are all connected to the acquisition signal input end of the algorithm processing module 2, so that the algorithm processing module 2 can obtain the acquisition signal, and the acquisition signal includes at least raw health data; The power output end of the power supply module 3 outputs power for power supply; The data communication end of the algorithm processing module 2 is connected to the first data communication end of the wireless communication module 4, and the second data communication end of the wireless communication module 4 establishes a wireless communication path with the remote terminal; The algorithm processing module 2 is used to identify at least the multi-source acquisition signals and the local user basic information through a multi-source recognition algorithm model, and push the corresponding detection results, and the detection results include health suggestions, medical suggestions, and prevention suggestions.

[0029] Specifically, each independent acquisition source of the multi-source input module 1 respectively acquires the physiological parameters and behavior data of the user. The acquisition signal output end of each acquisition source is connected to the acquisition signal input end of the algorithm processing module 2, so that the algorithm processing module 2 can receive the raw health data from multiple sources and process it. Each acquisition source continuously acquires data according to a preset sampling frequency to ensure the continuity and stability of the data. Among them, the physiological parameter acquisition sources include but are not limited to a heart rate sensor, a blood oxygen sensor, and a temperature sensor, which are used to obtain the heart rate change, blood oxygen saturation, and skin surface temperature of the user. The behavior data acquisition sources include an acceleration sensor and a gyroscope, which are used to detect the user's motion state, gait characteristics, and posture changes. All acquisition signals are transmitted to the algorithm processing module 2 in the original format to ensure the integrity of the data and provide a reliable input for subsequent analysis; Specifically, the power output end of the power supply module 3 continuously provides stable power support to each component of the system to ensure that each module can operate normally in different working modes. The power supply module 3 internally integrates a power management chip, which dynamically distributes electric energy according to the power consumption requirements of different modules to improve the power supply efficiency. When the system enters the low-power mode, the power supply module 3 reduces the power supply intensity to non-essential modules and only maintains the operation of the core computing unit and necessary sensors, thereby extending the battery life of the device. At the same time, the power supply module 3 has a battery management function. When it detects that the battery power is lower than a preset threshold, it triggers a low-battery warning signal and notifies the remote terminal through the wireless communication module 4 to remind the user to charge in time to ensure that the system can continuously provide health monitoring services; Specifically, the algorithm processing module 2 establishes a stable communication connection with the first data communication end of the wireless communication module 4 through the data communication end, enabling the processed health data to be efficiently transmitted. The second data communication end of the wireless communication module 4 conducts wireless data interaction with the remote terminal via Bluetooth or Wi-Fi. When the remote terminal requests to obtain the latest health data, the algorithm processing module 2 retrieves and formats the stored data, and pushes it to the remote terminal in real time through the wireless communication module 4, ensuring the timeliness and accuracy of the data. At the same time, the wireless communication module 4 supports the resume function of interrupted transmission, and can automatically resume data transmission after the communication is interrupted, preventing data loss or transmission failure and improving the reliability of communication; Specifically, the algorithm processing module 2 uses a multi-source recognition algorithm model to perform data fusion and pattern recognition on the received multi-source acquisition signals. First, it performs time alignment and format conversion on the original health data from different acquisition sources to ensure that the data from different sensors can be analyzed synchronously. Then, it applies a feature extraction algorithm to extract key health features from the original data. For example, it extracts heart rate variability parameters from the heart rate signal and calculates exercise features such as stride frequency and stride length from the gait signal. Then, the extracted features are input into the multi-source recognition algorithm model for analysis and classification. The model combines the user's basic information stored locally, automatically matches the historical health records, calculates the health status score based on the data change trend, and evaluates the user's current health condition in combination with the preset health standards. Finally, it generates a detection result, including health advice, medical advice, and preventive advice, to guide the user to adjust their lifestyle or take further health management measures; Specifically, the generation of health advice is based on the user's daily behavior data and health goals. By comparing the deviation between the user's recent exercise data, sleep quality, and physiological parameters and the established health goals, the algorithm processing module 2 identifies the deficiencies in the user's health management. For example, when it is detected that the user's average daily steps are below the recommended level, the health advice may include increasing the daily walking time or adjusting the exercise intensity. If the sleep quality is detected to have declined, it is recommended that the user optimize their work and rest schedule or adjust the sleep environment. All health advice is combined with the user's personal health condition and historical data to ensure the pertinence and executability of the advice; Specifically, the generation of medical advice depends on the long-term trend analysis of the user's health data. When the system detects that some physiological parameters continuously deviate from the normal range, or there are obvious health risk signals, the medical advice will be automatically generated based on the medical knowledge base and statistical analysis model. For example, when the heart rate variability index significantly decreases and is accompanied by fluctuations in blood oxygen levels, it may indicate that the user has a cardiovascular health risk. At this time, the medical advice may include seeking medical examination or conducting specific medical evaluations. The content of the medical advice can be combined with medical guidelines and the user's past medical history to make the advice more valuable for medical reference and, if necessary, push it to the remote terminal for further evaluation by professionals; Specifically, the generation of preventive suggestions is based on the cross-analysis of the user's health data and environmental factors. The system can combine the user's long-term health data, lifestyle habits, and external environmental information to predict possible health risks and provide targeted preventive measures. For example, when it is detected that the user has been in a high-stress state for a long time and the sleep quality has been continuously declining, the preventive suggestions may include relaxation training, reducing high-intensity work, or adjusting the work and rest structure. When the environmental temperature changes greatly, the preventive suggestions may prompt the user to pay attention to keeping warm or avoid exposure to high temperatures. All preventive suggestions are based on the system's in-depth analysis of the user's health trends, aiming to help the user take effective measures before problems occur, reduce health risks, and improve the overall health level.

[0030] Preferably, the multi-source input module 1 includes: A triaxial acceleration sensor for capturing the user's motion state data and attitude change data; A photoplethysmogram sensor for collecting non-invasive heart rate monitoring data and blood oxygen saturation monitoring data; A temperature sensor for monitoring environmental temperature data and skin surface temperature data.

[0031] In this embodiment, the signal output end of the triaxial acceleration sensor is connected to the acquisition signal input end of the algorithm processing module 2 to ensure that the user's motion state data and attitude change data can be transmitted to the algorithm processing module 2 in real time. The triaxial acceleration sensor continuously acquires acceleration change information in a three-dimensional space during the device wearing process. By measuring the acceleration values of the X-axis, Y-axis, and Z-axis, the user's gait, motion trajectory, and body attitude changes are calculated. The acquired raw acceleration data is denoised through a built-in filtering algorithm and fused with gyroscope data for calculation to improve the accuracy of motion mode recognition. The processed motion data is transmitted to the algorithm processing module 2 in a standardized format, providing basic data for gait analysis, motion monitoring, and abnormal behavior detection; the signal output end of the photoplethysmogram sensor is connected to the acquisition signal input end of the algorithm processing module 2, enabling it to obtain the user's pulse waveform data in real time and perform processing. The photoplethysmogram sensor irradiates the user's skin with a specific wavelength of light beam and measures the scattering and absorption changes of light in blood vessels, thereby calculating the heart rate and blood oxygen saturation. The photodetector inside the sensor continuously acquires the reflected signal and converts the analog signal into a digital signal through an analog-to-digital conversion circuit. The digitized pulse waveform data is filtered, normalized, and transmitted to the algorithm processing module 2 after removing ambient light interference. The algorithm processing module 2 calculates the heart rate mean and variability parameters by combining multiple pulse waveform cycles, and estimates the blood oxygen concentration at the same time, ensuring that the obtained physiological indicators have high accuracy and stability, and finally used for health status assessment and physiological load analysis; the signal output end of the temperature sensor is connected to the acquisition signal input end of the algorithm processing module 2 to monitor the ambient temperature data and skin surface temperature data in real time and provide the temperature change trend. The temperature sensor uses a high-precision thermistor or an infrared temperature measurement element, which can detect the temperature fluctuations of the external environment and measure the temperature state of the user's skin surface. The ambient temperature data is used to evaluate the impact of the external environment on the user's physiological state. For example, a cold environment may cause peripheral vasoconstriction and affect blood circulation, while the skin temperature data can be used to infer the user's body temperature change trend and perform joint analysis with data such as heart rate and blood oxygen to determine whether the user is in an abnormal physiological state. The data of the temperature sensor is transmitted to the algorithm processing module 2 after analog-to-digital conversion. The algorithm processing module 2 calculates the temperature change rate based on the temperature time series data and combines the user's historical body temperature data and motion state to perform health status trend assessment and risk prediction.

[0032] Preferably, the wireless communication module 4 is a Bluetooth beacon module. The use of a Bluetooth beacon module in the wireless communication module 4 can achieve stable data transmission in a low-power mode, ensure that health data can be efficiently synchronized to the remote terminal, and at the same time has a relatively long communication range and a fast transmission rate, improving the battery life of the device and the user experience.

[0033] Preferably, the algorithm processing module 2 includes: An acquisition unit, configured to collect raw health data in real time, preprocess and extract key health features from the raw health data, and substitute the key health features into a multi-source recognition algorithm model that has been deployed and trained in the intelligent terminal device to generate corresponding user behavior data; Specifically, the acquisition unit receives the raw health data from the multi-source input module 1 in real time and continuously collects data at a set sampling frequency. The received raw health data includes but is not limited to triaxial acceleration signals, photoplethysmogram signals, and temperature signals. The acquisition unit first preprocesses the data to remove noise, correct outliers, and enhance effective features. For the acceleration signal, high-pass filtering is used to eliminate low-frequency drift, and at the same time, Kalman filtering is used to improve the smoothness of the signal. For the photoplethysmogram signal, band-pass filtering is used to remove high-frequency interference, and at the same time, peak detection is performed to extract pulse wave period information. For the temperature signal, moving average filtering is applied to smooth data fluctuations. After preprocessing, the acquisition unit extracts key health features from the raw health data, including gait features, heart rate variability features, and temperature fluctuation features. The extracted feature data is normalized and substituted into the multi-source recognition algorithm model deployed in the intelligent terminal device, and user behavior data is generated through model calculation to ensure the accuracy and real-time nature of the user behavior data and provide a reliable data basis for subsequent data analysis and health management.

[0034] As Figure 2 shown, it shows the data analysis in the gait feature extraction process, mainly focusing on different state intervals during walking, and analyzing the characteristics of gait changes by marking different regions (such as "walking & low speed"). The blue curve in the figure represents the gait data changing with time, which may be the changes in features such as walking speed, step length, and gait stability. As time goes by, the curve generally shows an upward trend, which may indicate that the walking speed is gradually increasing or the gait features are changing. The red box indicates that the gait remains in the "walking & low speed" state between 0 and 500 time units; the dark blue box indicates that it is still in the "walking & low speed" state between 500 and 1000 time units, and the green box indicates the interference section when the gait features are present between 1000 and 1500 time units.

[0035] A customization unit, configured to call corresponding user basic information and health management settings customized based on the user basic information in the local database of the intelligent terminal device. The health management settings at least include health goals and warning thresholds; Specifically, the customization unit calls the user's basic information in the local database of the smart terminal device and dynamically generates a health management setting based on the user's basic information. The user's basic information includes parameters such as age, gender, weight, exercise habits, and past health records. The customization unit matches the corresponding health management template according to the user's basic information and personalizes the parameters in the template to ensure that the health management setting meets the individual needs of the user. The health management setting at least includes a health goal and a warning threshold. The health goal is used to set the user's daily health management tasks, such as the daily step goal, exercise duration, sleep quality requirements, etc. The warning threshold is used to set the safety range of key health indicators, such as the upper limit of heart rate and the lower limit of blood oxygen saturation. After generating the health management setting, the customization unit stores it in the local database and provides a reference for the subsequent health monitoring process to support personalized health assessment and intervention.

[0036] The screening unit is used to screen out normal behavior data and abnormal behavior data from the user behavior data. Among them, the current deviation value of the abnormal behavior data generated relative to the mean value of each normal behavior data exceeds the preset deviation threshold. Specifically, the screening unit screens normal behavior data and abnormal behavior data from the user behavior data. First, it obtains the historical statistical characteristics of the user behavior data and calculates the deviation value of the current behavior data relative to the historical mean. When the current deviation value exceeds the preset deviation threshold, the behavior data is determined as abnormal behavior data. The screening unit adopts a time window mechanism during the data screening process to avoid misjudgment caused by short-term fluctuations. For example, for gait data, the screening unit calculates the mean step frequency in the past week and compares the current step frequency with the mean. If the change range of the current step frequency exceeds the set threshold, the data is marked as abnormal behavior data. Similarly, for heart rate variability data, the screening unit calculates the standard deviation range of heart rate variability and marks it as abnormal behavior data when the current heart rate variability exceeds this range. The screening unit ensures the accuracy of health assessment through the above screening method and provides more reliable data support for health management.

[0037] The first push unit is used to push the corresponding health advice and / or execute the corresponding target intervention operation according to the first comparison result between the health goal and the corresponding normal behavior data. Specifically, the first push unit pushes corresponding health suggestions or performs target intervention operations based on the first comparison result between the health target and the normal behavior data. The first push unit first obtains the health target of the current user from the health management settings and calculates the achievement rate of the normal behavior data relative to the health target. For example, when the health target is set to 10,000 steps per day and the current behavior data shows that the user's steps are only 6,000 steps, the first push unit pushes a health suggestion to increase the walking time according to the insufficient achievement rate. If the normal behavior data has reached or exceeded the health target, positive feedback information is pushed to enhance the user's motivation for health management. In some cases, if the health target is not achieved and the system detects that the user has been sitting still for a long time, the first push unit can trigger a target intervention operation, such as guiding the user to do appropriate exercise through vibration reminder or voice prompt, to ensure the timeliness and effectiveness of the health management suggestions.

[0038] The second push unit is used to push corresponding trend prediction information and medical suggestions or preventive suggestions generated for the trend prediction information according to the second comparison result between the warning threshold and the corresponding abnormal behavior data.

[0039] Specifically, the second push unit pushes corresponding trend prediction information according to the second comparison result between the warning threshold and the abnormal behavior data, and generates medical suggestions or preventive suggestions for the trend prediction information. The second push unit first obtains the warning thresholds of various health indicators from the health management settings and calculates the deviation degree of the abnormal behavior data relative to the warning threshold. For example, when the heart rate variability is lower than the preset safe range and this trend has continued for several days, the second push unit analyzes the historical data and generates trend prediction information, indicating that the user's current cardiovascular health status may have a risk of decline. If the change trend of the abnormal behavior data exceeds the safe range, the second push unit generates corresponding medical suggestions according to the medical knowledge base, such as suggesting that the user perform a blood pressure test or consult a doctor. In addition, if the abnormal behavior data shows that the health indicator is in a critical state but has not exceeded the warning range, the second push unit can push preventive suggestions. For example, for users with poor sleep quality, it is recommended to optimize the sleep environment or adjust the work and rest time to reduce the health risk and improve the prospectiveness of health management.

[0040] Preferably, the key health features in the acquisition unit at least include gait features, heart rate variability features, and sleep features. The acquisition unit includes: The first determination subunit is used to determine the matching codes of each original health data, and the matching codes are determined based on the attribute information of the independent acquisition source; The second determination subunit is used to determine the corresponding data extraction rules according to the matching codes; The first extraction subunit is configured to extract corresponding gait features from the original health data if the data extraction rule is a gait extraction rule; The second extraction subunit is configured to extract corresponding heart rate variability features from the original health data if the data extraction rule is a heart rate variability extraction rule; The third extraction subunit is configured to extract corresponding sleep features from the original health data if the data extraction rule is a sleep extraction rule.

[0041] In this embodiment, the acquisition unit classifies and extracts features from the acquired original health data. First, the first determination subunit generates a matching code according to the attribute information of the independent acquisition source to ensure that the data collected by different sensors can accurately match the corresponding data processing rules. The matching code consists of parameters such as sensor type, data format, and measurement frequency. The second determination subunit queries the preset data extraction rules according to the matching code and selects a feature extraction algorithm applicable to this data type. When the extraction rule corresponding to the matching code is a gait extraction rule, the first extraction subunit extracts gait features from the triaxial acceleration sensor data, including walking frequency, step length, and gait stability parameters. When the extraction rule corresponding to the matching code is a heart rate variability extraction rule, the second extraction subunit calculates heart rate variability parameters from the photoplethysmogram data to measure the regulation ability of the autonomic nervous system. When the extraction rule corresponding to the matching code is a sleep extraction rule, the third extraction subunit combines heart rate data and motion state data to analyze the user's sleep stage and calculate the sleep quality score to ensure that the data collected from different sensors can be efficiently converted into key health features available for health management; For example, when the user wears a smart terminal for daily activities, the first determination subunit identifies its matching code from the data stream of the triaxial acceleration sensor. The matching code indicates that the data comes from the acceleration sensor and meets the standards for gait analysis. The second determination subunit confirms the extraction rule applicable to the gait data according to the matching code and selects the corresponding signal processing method. The first extraction subunit filters out high-frequency noise from the data through low-pass filtering according to the gait extraction rule, detects the walking cycle through the gait segmentation algorithm, calculates the walking frequency and step length, and finally extracts the gait stability feature. If the user enters a resting state, the acquisition unit extracts the changes in the heart rate interval from the photoplethysmogram data, and the second extraction subunit calculates the heart rate variability parameters. When the user enters the sleep state, the third extraction subunit combines the heart rate variability and acceleration data to identify the sleep stage and calculate the proportion of deep sleep to generate complete sleep features, thereby ensuring the accurate extraction of health data and personalized health analysis.

[0042] Preferably, the user behavior data in the acquisition unit at least includes gait behavior data, heart rate variability behavior data, and sleep behavior data. The acquisition unit further includes: A third determination subunit, configured to determine corresponding data recognition rules based on a multi-source recognition algorithm model that has been deployed in the intelligent terminal device and has been trained; A first generation subunit, configured to, if the data recognition rule is a gait recognition rule, perform pattern recognition and intensity parameter conversion on gait features based on the gait recognition rule, and generate corresponding gait behavior data, where the gait behavior data includes a motion pattern and an intensity parameter corresponding to the motion pattern and capable of being quantified; A second generation subunit, configured to, if the data recognition rule is a heart rate variability recognition rule, perform calculation and integration on heart rate variability features based on the heart rate variability recognition rule, and generate corresponding gait behavior data, where the gait behavior data includes heart rate interval change data, and an autonomic nervous system activity level corresponding to the heart rate interval change data and capable of being quantified; A third generation subunit, configured to, if the data recognition rule is a sleep recognition rule, perform stage recognition and quality assessment level conversion on sleep features based on the sleep recognition rule, and generate corresponding sleep behavior data, where the sleep behavior data includes a sleep quality assessment level capable of being quantified within each sleep stage.

[0043] In this embodiment, after the acquisition unit obtains the original health data and extracts the key health features, the third determination subunit determines, based on the multi-source recognition algorithm model deployed in the intelligent terminal device, the data recognition rules applicable to different health features, ensuring that subsequent data processing meets the established pattern recognition requirements. When the recognition rule is a gait recognition rule, the first generation subunit performs pattern recognition on the gait features and calculates the exercise intensity parameters to generate gait behavior data, where the gait behavior data includes the user's motion patterns, such as walking, running, sitting still, etc., and the intensity parameters corresponding to the motion pattern and capable of being quantified, such as walking speed, step frequency, etc. When the recognition rule is a heart rate variability recognition rule, the second generation subunit calculates the heart rate interval change and combines the autonomic nervous system activity level to integrate the heart rate variability features into gait behavior data, ensuring that the cardiac activity state is accurately characterized. When the recognition rule is a sleep recognition rule, the third generation subunit, based on information such as heart rate and motion state, recognizes the user's sleep stages, including light sleep, deep sleep, and rapid eye movement period, and converts the sleep features into a sleep quality assessment level capable of being quantified by calculating indicators such as sleep continuity and deep sleep ratio, so as to facilitate subsequent health analysis and personalized health management; For example, when the user starts exercising, the third determination subunit identifies gait data through a multi-source recognition algorithm model and confirms that the gait characteristics are applicable to the gait recognition rule. The first generation subunit classifies the user's gait patterns and calculates gait intensity parameters such as walking frequency, step length, and walking speed based on acceleration data to generate gait behavior data. When the user is sitting still or in a non-exercising state, the system identifies heart rate variability data, and the second generation subunit calculates heart rate variability indicators, combines the heart rate interval change data to infer the activity level of the autonomic nervous system, and forms complete heart rate variability behavior data. When the user enters the sleep state, the third generation subunit judges the sleep stage and calculates indicators such as the proportion of deep sleep and sleep stability based on heart rate and acceleration signals, and finally generates sleep behavior data, which can be used for subsequent health analysis, fatigue assessment, and generation of personalized health recommendations.

[0044] Specifically, in the acquisition unit, each data extraction rule and recognition rule ensure a comprehensive assessment of the user's behavior and physiological state through processing methods for different health characteristics. During the execution of the gait extraction rule, the system first preprocesses the original three-axis acceleration data, removes high-frequency noise through low-pass filtering to ensure the smoothness and stability of the data. Then, it analyzes the waveform of the acceleration signal through a gait cycle detection algorithm to identify a complete gait cycle. Within each cycle, the walking frequency (the number of steps per unit time), step length (the length of each step), and gait cycle (the time interval from the start of one step to the start of the next step) are extracted. The gait recognition rule then matches these extracted features with a preset motion pattern model through a pattern recognition algorithm to judge the user's current motion state (such as walking, running, sitting still, etc.), and further calculates the energy consumption intensity of each pattern, such as the motion intensity parameter obtained by combining the walking frequency and step length during walking, provides the motion pattern and the corresponding quantified intensity parameter, thereby generating gait behavior data.

[0045] The execution of the heart rate variability extraction rule focuses on extracting heart rate interval data from the signals obtained by the photoplethysmogram sensor. First, the pulse waveform data collected by the sensor obtains the heartbeat cycle through the peak detection algorithm, calculates the time interval between each cycle, and generates heart rate interval data. In the heart rate variability recognition rule, by performing time-domain and frequency-domain analyses on the heart rate interval data, indicators such as standard deviation of normal-to-normal RR intervals (SDNN) and mean heart rate variability are calculated. These indicators can reflect the activity level of the user's autonomic nervous system, and thus provide an assessment of cardiac health and stress status. For example, lower heart rate variability usually means that the user is in a higher stress state or may have cardiac health problems, while higher heart rate variability usually means better cardiac health and lower stress levels. Based on these data, the heart rate variability behavior data can present a quantitative assessment level, indicating the user's cardiac health status and the health status of the autonomic nervous system.

[0046] The sleep extraction rule focuses on combining multi-source sensor data, especially heart rate data and acceleration data. Acceleration data can identify the user's motion state, and combined with heart rate data, can estimate the user's sleep quality. The sleep extraction rule first identifies the user's sleep stage through acceleration data, and divides sleep into light sleep, deep sleep, and rapid eye movement (REM) stage according to different motion states (such as turning over, stillness, etc.). At the same time, the fluctuation of heart rate data is also closely related to the sleep stage. The heart rate is relatively stable during deep sleep, while the heart rate fluctuates more during light sleep. Through the combined analysis of these data, the sleep recognition rule can accurately identify the sleep stage, and calculate the sleep quality score according to parameters such as the proportion of deep sleep, sleep onset time, and arousal frequency, forming sleep behavior data. This data can not only reflect the user's sleep quality, but also provide a basis for personalized sleep improvement suggestions by evaluating the duration of different sleep stages. Finally, through the quantifiable sleep quality assessment level, it helps users better understand their sleep status and take corresponding improvement measures.

[0047] These extraction and recognition rules work together to enable the system to comprehensively evaluate the user's exercise, cardiac health, and sleep quality, and provide accurate data support and suggestions for subsequent health management.

[0048] Preferably, the algorithm processing module 2 further includes: The first generation unit is used to select a labeled dataset from a local experimental source or a partial open-source dataset, and use the labeled dataset to train a pre-trained model to generate a corresponding behavior pattern model. The role of the first generation unit is to select a suitable labeled dataset from a local experimental source or a partial open-source dataset for further training of the pre-trained model. By using these labeled datasets, the first generation unit can provide specific and representative sample data for the pre-trained model. These data contain labels of various user behavior patterns, such as exercise, rest, sleep, etc. By learning these labeled data, the pre-trained model can optimize and adjust its internal parameters, thus generating a behavior pattern model that can accurately identify and classify user behaviors.

[0049] The second generation unit is used to add multiple sample sources, determine a large-scale labeled dataset in each sample source, and use the large-scale labeled dataset to train the behavior pattern model to generate a corresponding generalization model. The role of the second generation unit is to provide rich and diverse data support for the training of the behavior pattern model by integrating large-scale labeled datasets from multiple sample sources. First, the second generation unit collects and organizes data from multiple sample sources. These sample sources can come from different user groups, different environmental conditions, or different health states, ensuring that the data has broad representativeness and comprehensiveness. Then, by labeling these data, a large-scale labeled dataset is generated, where each sample contains calibrated health behavior data and corresponding label information. Using these labeled datasets, the second generation unit trains the behavior pattern model, thus generating a generalization model that can adapt to diverse user data and environmental conditions, improving the model's prediction ability and adaptability, so that it can still accurately identify health behavior patterns and provide effective health management suggestions when facing different types of users.

[0050] A third generation unit, which is configured to select one or more source domains from multiple sample sources based on data distribution similarity, determine a target domain according to user basic information, and adapt the generalization model by using a transfer learning method to generate an accelerated adaptation model. The role of the third generation unit is to identify and select one or more source domains from multiple sample sources through analysis based on data distribution similarity. The data features contained in these source domains have a high similarity with the health data of the target user, and determine the target domain according to the basic information of the target user, that is, the personalized health data distribution of the target user. Through the transfer learning method, the knowledge learned from the source domain is transferred to the target domain to adapt the generalization model, enabling the model to better identify and adapt to the specific health characteristics and behavior patterns of the target user. This process can accelerate the learning efficiency of the model, improve the adaptability of the model to new user data, ensure that the health management system can provide accurate health advice and predictions for different users, reduce the dependence on a large amount of new data at the same time, and improve the training speed and application effect of the model.

[0051] A fourth generation unit, which is configured to establish an incremental learning mechanism in the accelerated adaptation model to generate a trained multi-source recognition algorithm model. The role of the fourth generation unit is to enable the multi-source recognition algorithm model to be continuously updated and optimized to adapt to changes in user behavior data and health status by introducing an incremental learning mechanism in the accelerated adaptation model. Specifically, the incremental learning mechanism allows the model, when receiving new health data, not to retrain the entire model, but to make adaptive adjustments based on the existing knowledge and new data, so as to continuously improve the recognition accuracy and generalization ability of new data without forgetting past experience. For example, when the user's exercise pattern changes or the health status changes, this mechanism can make fine-tuning on the existing training results, enabling the model to reflect the new behavior characteristics of the user in real time, ensuring the accuracy and personalization of long-term health management. In addition, the incremental learning mechanism can effectively reduce the computational burden of model training and improve the system response speed and real-time performance.

[0052] Preferably, the algorithm processing module 2 further includes: An acquisition unit for acquiring corresponding basic user information based on the registration window filled out by the user. The basic user information includes at least physiological information, health background information, and calling corresponding historical health information based on the physiological information. The role of the acquisition unit is to obtain the basic information of the user through the registration window filled out by the user when the user first uses the device or registers. Specifically, it includes the user's physiological information (such as age, gender, weight, height, etc.) and health background information (such as past medical history, family medical history, allergy history, etc.). These information provide basic data for personalized health management. In addition, the acquisition unit will also call the user's historical health information, such as past physical examination reports, chronic disease records, medical history, etc., from the local database or cloud system based on the physiological information provided by the user. These historical health data can provide more accurate references for subsequent health assessment and risk prediction, thus supporting the setting of subsequent health goals, the adjustment of warning thresholds, and the push of personalized health suggestions.

[0053] A fifth generation unit for generating customized health goals according to the physiological information and health background information. The role of the fifth generation unit is to formulate personalized health goals based on the user's physiological information and health background information. This unit first extracts key physiological information from the user's basic health data, such as age, weight, blood pressure, heart rate, historical health records, etc., and at the same time combines the user's health background information, such as past medical history, lifestyle and exercise level, etc., to comprehensively analyze the user's health status. Based on this information, the system will refer to relevant health guidelines, medical standards and individual needs to generate health goals suitable for the user, such as setting reasonable weight control goals, daily step goals, exercise intensity goals, diet control goals, and goals for preventing specific diseases. Through customized health goals, the system can help users set practical health improvement plans, promote their better implementation of health management in daily life, and ensure the achievability and effectiveness of health goals.

[0054] The sixth generation unit is used to generate customized warning thresholds and a dynamic adjustment mechanism for dynamically adjusting the warning thresholds according to physiological information and historical health information; the sixth generation unit generates personalized warning thresholds by analyzing the user's physiological information (such as age, gender, weight, etc.) and historical health information (such as previous health records, disease history, etc.), and these thresholds are used to monitor key health indicators such as heart rate, blood oxygen saturation, blood pressure, etc. When these health indicators exceed the set safe range, the system can issue an alarm in a timely manner to remind the user to take corresponding health management measures. To ensure the flexibility and adaptability of the warning mechanism, the sixth generation unit also includes a dynamic adjustment mechanism, which automatically adjusts the warning thresholds according to changes in the user's health status and fluctuations in real-time data, ensuring that the warning system can be optimized as the user's health status changes. For example, when the user is exercising, the normal range of heart rate may be different. Through the dynamic adjustment mechanism, the system can adjust the heart rate warning threshold in real-time according to the exercise intensity, thus avoiding unnecessary false alarms. At the same time, when the user's health status deteriorates, the warning is triggered in advance to provide more personalized and accurate health monitoring.

[0055] An integration unit is used to integrate health goals, warning thresholds, and a dynamic adjustment mechanism to generate corresponding health management settings.

[0056] Preferably, the first push unit includes: A calling subunit is used to call the quantitative indicators in the normal behavior data; the role of the calling subunit is to extract and call the quantitative indicators from the normal behavior data for subsequent analysis and evaluation of the health status. The normal behavior data includes the user's physiological and behavioral data such as exercise, sleep, and heart rate. Through the calling subunit, the system can identify key quantitative indicators from these data, such as step frequency, step length, heart rate variability, sleep quality score, etc. These quantitative indicators provide basic data support for subsequent health goal comparison and behavior evaluation. The calling subunit extracts the quantitative indicators related to the goal according to the preset rules and the user's health goals, and provides them to the subsequent analysis module or push module to monitor in real-time whether the user has achieved the health goals or whether health intervention measures need to be taken, thus ensuring the personalization and accuracy of health management.

[0057] The fourth generation subunit is used to match each quantization index to the corresponding health goal. Each health goal is respectively corresponding to one or more quantization indexes, and the quantization indexes corresponding to each health goal are weighted and averaged to generate the corresponding target quantization value. The role of the fourth generation subunit is to match each quantization index extracted from the user behavior data with the health goals set by the user. According to the specific requirements of each health goal, one or more relevant quantization indexes are selected for correspondence. For example, the step goal may correspond to the daily steps and exercise intensity, and the heart rate goal may correspond to the heart rate variability and the heart rate during exercise, etc. This subunit comprehensively evaluates the quantization indexes of each health goal through the method of weighted average, and the weights are adjusted according to the importance of each quantization index to ensure that the evaluation of each health goal is more scientific and accurate. Finally, the fourth generation subunit generates the target quantization value, which can reflect the user's performance in the process of achieving the health goal and provide data support for subsequent health suggestions, intervention measures and goal adjustment.

[0058] The fifth generation subunit is used to compare the health goal with the corresponding target quantization value to generate the corresponding first comparison result, and push the corresponding health suggestions and / or execute the corresponding target intervention operations according to the first comparison result. The role of the fifth generation subunit is to compare and analyze the user's health goal with the actually achieved target quantization value. The health goals include set goals such as daily steps, exercise duration, sleep quality, etc., while the target quantization value is the actually measured health data, such as the actual steps, exercise time or sleep time, etc. This subunit generates the first comparison result by calculating the difference between the target quantization value and the set health goal. If the target quantization value fails to reach the set health goal, the system will push the corresponding health suggestions according to the difference, such as increasing the amount of exercise or adjusting the work and rest time, etc.; if the health goal has been achieved or exceeded the expectation, the system will push positive feedback or encouragement information to motivate the user to maintain good health habits. In addition, according to the comparison result, if the user fails to complete the health goal, the system can execute the corresponding intervention operations, such as reminding the user to increase activities or reminding the user to get up and move through the intelligent reminder device, so as to realize personalized health management and intervention.

[0059] Specifically, assume that the user's health goals include 10,000 steps per day and 60 minutes of exercise per day. The calling subunit first extracts quantitative metrics from the normal behavior data, such as the user's actual number of steps and exercise duration on that day. The fourth generation subunit matches these quantitative metrics with the set health goals. The step goal corresponds to the step quantity metric, and the exercise goal corresponds to the exercise duration metric. Then, according to the importance weights of each health goal (for example, steps account for 70% and exercise duration accounts for 30%), the various quantitative metrics are weighted and averaged to generate a target quantitative value. If the user's actual number of steps is 8,000 steps and the exercise duration is 50 minutes, the generated target quantitative values are 9,000 steps and 55 minutes. Next, the fifth generation subunit compares the health goals (10,000 steps and 60 minutes) with the target quantitative values, calculates the difference, and generates a first comparison result. Based on this result, if the goal is not achieved, the system will push health suggestions, such as increasing walking time or adjusting exercise intensity; if the goal is achieved, the system will push positive feedback, such as encouraging to continue maintaining good health habits. If necessary, the system can also perform intervention operations, such as reminding the user to do additional exercise.

[0060] The principle and function of the second push unit are the same as those of the first push unit, so they will not be elaborated here.

[0061] An intelligent terminal uses a health detection and management system with multi-variable composite sensing.

[0062] The above-described embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. A multivariable composite sensing health detection management system, characterized in that: The multivariable composite sensing health detection and management system comprises a multi-source input module (1), an algorithm processing module (2), a power supply module (3) and a wireless communication module (4); The acquisition signal output end of each independent acquisition source in the multi-source input module (1) is connected to the acquisition signal input end of the algorithm processing module (2), so that the algorithm processing module (2) obtains the acquisition signal, and the acquisition signal at least includes the original health data; The power output end of the power supply module (3) outputs power for supplying power; The data communication end of the algorithm processing module (2) is connected to the first data communication end of the wireless communication module (4), and the second data communication end of the wireless communication module (4) establishes a wireless communication path with a remote terminal; The algorithm processing module (2) is used to identify the multi-source collected signals and local user basic information at least through a multi-source identification algorithm model, and push corresponding detection results, wherein the detection results include health advice, medical advice and prevention advice.

2. A multivariable composite sensing health detection and management system according to claim 1, characterized in that: The multi-source input module (1) comprises: Three-axis acceleration sensor, used to capture the user's motion status data and posture change data; Photoplethysmography sensor for collecting non-invasive heart rate monitoring data and blood oxygen saturation monitoring data; Temperature sensor, used to monitor ambient temperature data and skin surface temperature data.

3. The multivariable composite sensing health monitoring and management system according to claim 1 is characterized in that: The wireless communication module (4) is a Bluetooth beacon module.

4. The multivariable composite sensing health detection and management system according to claim 1, characterized in that: The algorithm processing module (2) comprises: A collection unit, used to collect raw health data in real time, pre-process and extract key health features in the raw health data, substitute the key health features into the multi-source recognition algorithm model that has been deployed in the smart terminal device and has been trained, and generate corresponding user behavior data; A customization unit, configured to call corresponding user basic information in a local database of the intelligent terminal device, and health management settings customized based on the user basic information, wherein the health management settings include at least health goals and warning thresholds; a screening unit, configured to screen out normal behavior data and abnormal behavior data from the user behavior data, wherein a current deviation value generated by the abnormal behavior data relative to an average value of each of the normal behavior data exceeds a preset deviation threshold; A first push unit, configured to push corresponding health advice and / or perform corresponding target intervention operations according to a first comparison result between the health target and the corresponding normal behavior data; The second push unit is used to push corresponding trend prediction information and medical advice or prevention advice generated for the trend prediction information according to a second comparison result between the warning threshold and the corresponding abnormal behavior data.

5. A multivariable composite sensing health detection and management system according to claim 4, characterized in that: The key health characteristics in the acquisition unit include at least gait characteristics, heart rate variability characteristics and sleep characteristics, and the acquisition unit includes: A first determination subunit, configured to determine a matching code for each of the original health data, wherein the matching code is determined based on the attribute information of the independent collection source; A second determination subunit, configured to determine a corresponding data extraction rule according to the matching code; a first extraction subunit, configured to extract corresponding gait features from the original health data if the data extraction rule is a gait extraction rule; a second extraction subunit, configured to extract corresponding heart rate variability features from the original health data if the data extraction rule is a heart rate variability extraction rule; The third extraction subunit is configured to extract corresponding sleep features from the original health data if the data extraction rule is a sleep extraction rule.

6. A multivariable composite sensing health detection and management system according to claim 5, characterized in that: The user behavior data in the collection unit at least includes gait behavior data, heart rate variability behavior data and sleep behavior data, and the collection unit further includes: A third determination subunit is used to determine a corresponding data recognition rule based on a multi-source recognition algorithm model that has been deployed in the intelligent terminal device and has been trained; A first generating subunit is used for, if the data recognition rule is a gait recognition rule, performing pattern recognition and intensity parameter conversion on the gait feature based on the gait recognition rule to generate corresponding gait behavior data, wherein the gait behavior data includes a motion pattern and an intensity parameter corresponding to the motion pattern and quantified; a second generating subunit, configured to, if the data identification rule is a heart rate variability identification rule, calculate and integrate the heart rate variability characteristics based on the heart rate variability identification rule to generate corresponding gait behavior data, wherein the gait behavior data includes heart rate interval variation data and an autonomic nervous system activity level corresponding to the heart rate interval variation data and quantifiable; The third generating subunit is used for, if the data identification rule is a sleep identification rule, performing stage identification and quality assessment level conversion on the sleep characteristics based on the sleep identification rule to generate corresponding sleep behavior data, wherein the sleep behavior data includes quantifiable sleep quality assessment levels in each sleep stage.

7. The multivariable composite sensing health monitoring and management system according to claim 2 is characterized in that: The algorithm processing module (2) also includes: A first generating unit is used to select a labeled data set from a local experimental source or a part of an open source data set, train a pre-trained model using the labeled data set, and generate a corresponding behavior pattern model; A second generating unit is used to add multiple sample sources, determine a large-scale annotated data set from each of the sample sources, train the behavior pattern model using the large-scale annotated data set, and generate a corresponding generalized model; A third generation unit is used to select one or more source domains from multiple sample sources based on data distribution similarity, determine a target domain based on user basic information, adapt the generalization model using a transfer learning method, and generate an accelerated adaptation model; The fourth generating unit is used to establish an incremental learning mechanism in the accelerated adaptation model to generate a trained multi-source recognition algorithm model.

8. The multivariable composite sensing health detection and management system according to claim 2, characterized in that: The algorithm processing module (2) also includes: An acquisition unit, configured to acquire corresponding basic user information based on a registration window filled out by a user, wherein the basic user information includes at least physiological information, health background information, and historical health information corresponding to the physiological information; a fifth generating unit, configured to generate a customized health goal according to the physiological information and the health background information; a sixth generating unit, configured to generate a customized warning threshold value according to the physiological information and the historical health information, and a dynamic adjustment mechanism for dynamically adjusting the warning threshold value; An integration unit is used to integrate the health goal, the warning threshold and the dynamic adjustment mechanism to generate corresponding health management settings.

9. The multivariable composite sensing health monitoring and management system according to claim 2, characterized in that: The first pushing unit includes: A calling subunit, used for calling the quantitative indicators in the normal behavior data; A fourth generating subunit is used to match each of the quantitative indicators to the corresponding health goals, each of the health goals corresponds to one or more of the quantitative indicators, and to perform weighted averaging on the quantitative indicators corresponding to each of the health goals to generate a corresponding target quantitative value; The fifth generating subunit is used to compare the health goal with the corresponding target quantified value, generate a corresponding first comparison result, and push corresponding health advice and / or perform corresponding target intervention operations according to the first comparison result.

10. An intelligent terminal, characterized in that: A health detection and management system using a multivariable composite sensor as described in any one of claims 1-9.

Images