Internet of things based integrated device for nursing of old people with sarcopenia and monitoring method

Abstract

This invention discloses an integrated device and monitoring method for sarcopenia care in the elderly based on the Internet of Things (IoT), belonging to the field of geriatric care. The integrated device and monitoring method for sarcopenia care in the elderly based on IoT includes a main body, a multimodal sensing unit, a cloud analysis unit, and an intervention execution unit. This invention solves the problems of discontinuous monitoring and lack of personalized intervention in existing technologies. Through the multimodal sensing unit, this invention can collect muscle micro-vibration signals, joint angles, gait phases, and plantar pressure data of the elderly. Using lightweight LSTM network analysis, it accurately outputs the sarcopenia progression risk index (SRI) and fall risk index (FRI), providing a scientific basis for nursing care. The intervention execution unit generates personalized nursing plans based on the risk indices, including resistance training, nutritional prescriptions, medication reminders, and environmental modification suggestions. Electrical stimulation intervention is performed through patches to meet the health needs of different elderly individuals, thereby improving the nursing effect.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of elderly care, in particular to an old-age sarcopenia nursing integrated device based on the Internet of Things and a monitoring method. BACKGROUND

[0002] Sarcopenia refers to the phenomenon of skeletal muscle mass reduction, muscle strength decline and physical function decline with age. It not only causes the elderly to be inconvenient and easy to fatigue, but also significantly increases the risk of falls, and further causes serious consequences such as fractures, which seriously affects the quality of life and healthy life of the elderly.

[0003] At present, the monitoring and intervention of sarcopenia mainly rely on clinical evaluation and some traditional detection means, such as bioelectrical impedance analysis (BIA), dual-energy X-ray absorptiometry (DEXA), etc. Traditional detection means can usually only be performed at specific time points, and cannot monitor the muscle state and body function changes of the elderly in real time. Moreover, the existing intervention measures are mostly general schemes, which do not fully consider individual differences, resulting in limited intervention effect. Therefore, the existing needs are not met, and the old-age sarcopenia nursing integrated device based on the Internet of Things and the monitoring method are proposed. SUMMARY

[0004] The purpose of the present application is to provide an old-age sarcopenia nursing integrated device based on the Internet of Things and a monitoring method. By collecting multi-modal data including muscle micro-vibration signals, joint angles, gait phases and plantar pressure data, after quality evaluation, processing and fusion, the data are input into a lightweight LSTM network, so as to output a sarcopenia progression risk index SRI and a fall risk index FRI, and then a personalized nursing scheme is generated according to the analysis results, including resistance training, nutrition prescription, medication reminder and environmental modification suggestion, and targeted electrical stimulation intervention, thereby solving the problems raised in the background art.

[0005] To achieve the above purpose, the present application provides the following technical scheme: an old-age sarcopenia nursing integrated device based on the Internet of Things, comprising a device main body, a handrail is arranged above the device main body, a plantar plate is arranged below the back of the device main body, patches are arranged on both sides of the device main body, and a main screen is arranged on the front of the device main body.

[0006] The old-age sarcopenia nursing integrated device based on the Internet of Things further comprises a multi-modal perception unit, a cloud analysis unit and an intervention execution unit. The multi-modal perception unit comprises a plantar distributed pressure sensor A and a joint elastic band B. The plantar distributed pressure sensor A is arranged on the plantar plate, and the joint elastic band B is arranged on the back of the device main body.

[0007] The multi-modal sensing unit is configured to obtain muscle micro-vibration signals, joint angles and gait phases through the joint elastic band B, and to obtain plantar pressure data through the plantar distributed pressure sensor A, and to obtain multi-modal data based on the obtained information and to perform data quality evaluation, processing and fusion;

[0008] The multi-modal sensing unit is configured to obtain muscle micro-vibration signals, joint angles and gait phases through the joint elastic band B, and to obtain plantar pressure data through the plantar distributed pressure sensor A, and to obtain multi-modal data based on the obtained information and to perform data quality evaluation, processing and fusion;

[0009] The cloud analysis unit is configured to receive the multi-modal data obtained by the multi-modal sensing unit for analysis, input the multi-modal data into a lightweight LSTM network, and output a sarcopenia progression risk index SRI and a fall risk index FRI.

[0010] The intervention execution unit is configured to automatically generate a personalized care plan including resistance training, nutrition prescription, medication compliance reminders and environmental modification suggestions according to the results output by the cloud analysis unit, and to display the personalized care plan through the main screen, and to perform electrical stimulation care intervention through the patch.

[0011] Further, the multi-modal sensing unit comprises:

[0012] The data acquisition module is configured to use the joint elastic band B to realize synchronous acquisition of muscle micro-vibration signals, joint angles and gait phases at a pre-set sampling frequency, and simultaneously use the plantar distributed pressure sensor A to acquire plantar pressure data in real time.

[0013] The joint elastic band B is integrated with a three-axis accelerometer, a three-axis gyroscope, a micro myoelectric sensor array and an inertial measurement unit IMU, the muscle micro-vibration signals are acquired through the micro myoelectric sensor array, the gait phase is identified through the three-axis accelerometer and the three-axis gyroscope, and the joint angle change is monitored through the inertial measurement unit IMU.

[0014] The quality evaluation module is configured to evaluate the quality of the collected multi-modal data, including data integrity check, noise detection and evaluation and data consistency evaluation, and to comprehensively score the data quality based on the evaluation results.

[0015] The data processing module is configured to pre-process the multi-modal data that has passed the quality evaluation.

[0016] Further, the quality evaluation module specifically comprises:

[0017] Detect whether the collected data is missing, abnormal or incomplete;

[0018] Analyze the noise level in the data, evaluate the signal-to-noise ratio of the data, and determine whether the data is severely disturbed;

[0019] Check whether the data collected by different sensors is consistent in time and logic;

[0020] According to the evaluation results, the data quality is comprehensively scored, and the data quality of the current multi-modal data is judged according to the comprehensive score;

[0021] If the comprehensive score is within the preset score range, it is determined that the data quality of the current multi-modal data is high;

[0022] If the comprehensive score is not within the preset score range, it is determined that the data quality of the current multi-modal data is high, and the multi-modal data is re-collected.

[0023] Further, the data acquisition module further comprises a calibration module for periodically calibrating various sensors integrated on the plantar distributed pressure sensor A and the joint elastic band B, ensuring the accuracy and reliability of the collected data, and repeatedly testing the calibrated sensors multiple times, comparing the calibrated data with the pre-calibration data, evaluating the calibration effect, and determining whether the calibration reaches the expected target according to the comparison result. If not, recalibrate.

[0024] Further, the cloud analysis unit comprises:

[0025] The model loading module is configured to load a pre-constructed and trained lightweight LSTM network, input the extracted feature data into the lightweight LSTM network, perform forward propagation through the lightweight LSTM network, extract time series features, and finally output an intermediate feature vector;

[0026] The risk determination module is configured to determine the sarcopenia progression risk index SRI and the fall risk index FRI based on the lightweight LSTM network, specifically:

[0027] The intermediate feature vector output by the lightweight LSTM network is mapped to the sarcopenia progression risk index and the fall risk index through a specific neural network layer or regression model;

[0028] According to the determined sarcopenia progression risk index SRI and the fall risk index FRI, the risk is classified into low risk, medium risk and high risk;

[0029] The result output module is configured to generate a risk assessment report and output the result to the home screen for display.

[0030] Further, the risk determination module verifies and calibrates the determined sarcopenia progression risk index SRI and fall risk index FRI by comparing the determined risk indexes with historical data or known standards, adjusts the risk indexes according to the comparison results, detects abnormal values or unreasonable data in the determination results, and finally outputs the verified and calibrated sarcopenia progression risk index SRI and fall risk index FRI.

[0031] Further, the intervention execution unit comprises:

[0032] The scheme generation module is configured to generate a personalized care scheme according to the sarcopenia progression risk index SRI and the fall risk index FRI output by the cloud analysis unit, specifically:

[0033] In combination with the health status of the user, a suitable resistance training plan is generated, including training frequency, intensity and duration;

[0034] In combination with the nutritional needs of the user, a personalized nutritional prescription is generated, including daily intake of protein, vitamins and minerals recommendations;

[0035] According to the health status and medication needs of the user, a medication reminder plan is generated, including home environment modification and auxiliary equipment recommendations;

[0036] The scheme display module is configured to display the generated personalized care scheme to the user through the home screen and provide interactive functions, allowing the user to view detailed information, adjust the plan and set reminders;

[0037] The care intervention module is configured to pre-store electrical stimulation parameters, including intensity, frequency and duration, that are suitable for the sarcopenia progression risk index SRI and the fall risk index FRI, and control the patch to adjust to the corresponding electrical stimulation parameters based on the sarcopenia progression risk index SRI and the fall risk index FRI output by the cloud analysis unit to intervene in the care of the user, while recording the use of electrical stimulation.

[0038] Further, the calibration module comprises:

[0039] The model training submodule is used to:

[0040] Obtain sensitive environmental factors of various sensors integrated on the plantar distributed pressure sensor A and the joint elastic band B, and collect parameters of the sensitive environmental factors of each sensor to obtain sensitive environmental parameters; wherein the various sensors integrated on the joint elastic band B include a three-axis accelerometer, a three-axis gyroscope, a micro electromyography sensor array and an inertial measurement unit IMU;

[0041] Obtaining product parameter information and working principle information of each sensor; performing core hardware parameter extraction based on the product parameter information and working principle information of each sensor to determine the core hardware parameters corresponding to each sensor;

[0042] Performing coupling analysis on the sensitive environmental parameters and core hardware parameters corresponding to each sensor to determine the coupling influence law corresponding to each sensor;

[0043] Based on the coupling influence law, a mixed feature set is constructed; the mixed feature set includes an environmental feature set, a hardware feature set, and a coupling feature set;

[0044] The sensitive environmental parameters, core hardware parameters, and mixed feature set are divided into a training data set and a validation data set according to a predetermined proportion;

[0045] Based on the training data set, a neural network model is iteratively trained to obtain a data-driven submodel;

[0046] Based on the sensor working principle, an error formula is derived to determine the quantitative relationship between embedded hardware parameters and environmental parameters, and a physical submodel is constructed;

[0047] The data-driven submodel and the physical submodel are fused to obtain a multi-dimensional environment-hardware hybrid compensation model;

[0048] The compensation value calculation sub-module is used to:

[0049] Any sensor is selected as a target sensor; a measurement value data set of the target sensor within a predetermined period is obtained; the measurement value data set is input into the multi-dimensional environment-hardware hybrid compensation model to determine the compensation value corresponding to each measurement value in the measurement value data set;

[0050] The mean value of the compensation values corresponding to each measurement value in the measurement value data set is calculated as the compensation mean value of the measurement value data set;

[0051] The calibration sub-module is used to:

[0052] Based on the compensation mean value, a compensation coefficient of the target sensor is determined;

[0053] Based on the compensation coefficient, the target sensor is calibrated;

[0054] All sensors are traversed to complete calibration of various types of sensors.

[0055] Further, the data quality is comprehensively scored, including:

[0056] Based on the data integrity evaluation result, the noise detection and evaluation result, and the data consistency evaluation result, the quality score of the collected multi-modal data is calculated;

[0057]

[0058] wherein, represents a quality score of the collected multi-modal data; represents a total number of modalities in the multi-modal data; represents a weight of the th modality data; represents a time decay factor of the th modality data; represents an integrity score of the th modality data; represents a noise score of the th modality data; represents a consistency score of the th modality data; , , respectively represent an integrity score index coefficient, a noise score index coefficient and a consistency score index coefficient of the th modality data; represents a cross-modality collaborative correction coefficient.

[0059] The monitoring method of the Internet of Things-based integrated device for nursing sarcopenia in the elderly is used for using the Internet of Things-based integrated device for nursing sarcopenia in the elderly, and comprises the following steps:

[0060] Multi-modal data containing muscle micro-vibration signals, joint angles, gait phases and plantar pressure data are synchronously collected according to a preset sampling frequency, and are uploaded to a cloud analysis unit;

[0061] After the cloud analysis unit receives the multi-modal data, quality assessment is performed, if the data quality is not up to standard, re-collection is performed, and the qualified multi-modal data is preprocessed, and key features are extracted;

[0062] The extracted feature data is input into a lightweight LSTM network, a sarcopenia progression risk index SRI and a fall risk index FRI are output, and the risks are classified;

[0063] According to the sarcopenia progression risk index SRI and the fall risk index FRI, a personalized nursing scheme containing resistance training, nutrition prescription, drug reminder and environmental modification suggestion is automatically generated, and is displayed to the user through a device main screen, and at the same time, electrical stimulation intervention is performed through a patch.

[0064] Compared with the prior art, the beneficial effects of the present application are:

[0065] This invention utilizes a multimodal sensing unit to comprehensively and accurately collect muscle micro-vibration signals, joint angles, gait phases, and plantar pressure data from elderly individuals. By employing a lightweight LSTM network for analysis, it accurately outputs the Sarcopenia Progression Risk Index (SRI) and the Fall Risk Index (FRI), providing a scientific basis for nursing care. The intervention execution unit generates personalized nursing plans based on these risk indices, including resistance training, nutritional prescriptions, medication reminders, and environmental modification suggestions. Electrical stimulation intervention via patches further meets the diverse health needs of elderly individuals, thereby enhancing nursing outcomes. Attached Figure Description

[0066] Figure 1 This is a schematic diagram of the integrated IoT-based nursing device for sarcopenia in the elderly according to the present invention.

[0067] Figure 2 This is a flowchart of the monitoring method for the IoT-based integrated nursing device for sarcopenia in the elderly according to the present invention;

[0068] Figure 3 This is a schematic diagram of the appearance of the IoT-based integrated nursing device for sarcopenia in the elderly according to the present invention;

[0069] Figure 4 This is a schematic diagram of the back of the IoT-based integrated nursing device for sarcopenia in the elderly according to the present invention.

[0070] In the diagram: 1. Main body of the equipment; 11. Handrail; 2. Foot plate; 3. Patch; 4. Main screen. Detailed Implementation

[0071] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0072] To address the limitations of existing testing methods, which typically only allow for monitoring at specific time points and cannot provide real-time observation of muscle condition and functional changes in older adults, and the fact that current interventions are often generic and fail to adequately consider individual differences, resulting in limited intervention effectiveness, please refer to [link to relevant documentation]. Figures 1-4 This embodiment provides the following technical solution:

[0073] The IoT-based integrated care device for sarcopenia in the elderly includes a main body 1, which serves as the core framework of the entire device, supporting and integrating all other functional components to ensure the stability and integrity of the device. A handrail 11 is located on the top of the main body 1 to provide support and assistance to the elderly, helping them maintain balance and stability while using the device. A footplate 2 is located on the lower back of the main body 1 to support the elderly's feet and integrates a distributed foot pressure sensor A to collect foot pressure data. Patches 3 are located on both sides of the main body 1 for electrical stimulation nursing intervention. A main screen 4 is located on the top front of the main body 1 to display personalized care plans, risk assessment reports, and provide user interaction functions.

[0074] The IoT-based integrated device for sarcopenia care in the elderly also includes a multimodal sensing unit, a cloud analysis unit, and an intervention execution unit. The multimodal sensing unit includes a plantar distributed pressure sensor A and a joint elastic bandage B. The plantar distributed pressure sensor A is installed on the plantar plate 2, and the joint elastic bandage B is installed on the back of the main body 1 of the device.

[0075] The multimodal sensing unit is configured to acquire muscle micro-vibration signals, joint angles and gait phases through joint elastic band B, acquire plantar pressure data through plantar distributed pressure sensor A, obtain multimodal data based on the acquired information, and perform data quality assessment, processing and fusion.

[0076] Among them, the various types of sensors integrated on the plantar distributed pressure sensor A and the joint elastic bandage B are calibrated, and the data quality is comprehensively scored based on the data quality assessment results. The score reflects the quality status of multimodal data in different dimensions.

[0077] The cloud-based analysis unit is configured to receive and analyze multimodal data acquired by the multimodal perception unit, input the multimodal data into a lightweight LSTM network, and output the sarcopenia progression risk index (SRI) and fall risk index (FRI).

[0078] The intervention execution unit is configured to automatically generate a personalized care plan that includes resistance training, nutritional prescriptions, medication adherence reminders and environmental modification suggestions based on the results output by the cloud analysis unit, and display it through the main screen 4. At the same time, it performs electrical stimulation care intervention through the patch 3.

[0079] The technical effects of the above-mentioned solution are as follows: Through the multimodal sensing unit, muscle micro-vibration signals, joint angles, gait phases, and plantar pressure data can be collected simultaneously to achieve comprehensive monitoring of the elderly's physical condition. The cloud analysis unit analyzes the multimodal data through a lightweight LSTM network and accurately outputs the sarcopenia progression risk index (SRI) and fall risk index (FRI) to provide a scientific basis for nursing care. The intervention execution unit automatically generates personalized nursing plans based on the risk index, including resistance training, nutritional prescriptions, medication reminders, and environmental modification suggestions. Through electrical stimulation nursing intervention, it meets the health needs of different elderly people and improves the pertinence and effectiveness of nursing care.

[0080] In summary, the IoT-based integrated nursing device for sarcopenia in the elderly is highly integrated, easy to operate, and displays information and provides interactive functions through the main screen, making it convenient for both the elderly and caregivers to use. At the same time, the cloud analysis unit enables remote monitoring and analysis, improving nursing efficiency and reducing labor costs.

[0081] The multimodal sensing unit includes:

[0082] The data acquisition module is configured to simultaneously acquire muscle micro-vibration signals, joint angles, and gait phases using a joint elastic bandage B at a pre-set sampling frequency. At the same time, it uses a plantar distributed pressure sensor A to acquire plantar pressure data in real time.

[0083] Among them, the joint elastic band B integrates a triaxial accelerometer, a triaxial gyroscope, a miniature electromyography sensor array and an inertial measurement unit (IMU). The miniature electromyography sensor array acquires muscle micro-vibration signals, the triaxial accelerometer and triaxial gyroscope perform gait phase recognition, and the inertial measurement unit (IMU) monitors joint angle changes.

[0084] The quality assessment module is configured to perform quality assessment on the collected multimodal data, including data integrity check, noise detection and assessment, and data consistency assessment, and to give a comprehensive score for data quality based on the assessment results;

[0085] The data processing module is configured to preprocess the multimodal data used for quality assessment, including:

[0086] Data cleaning removes noise, deletes outliers, and fills in missing values.

[0087] By synchronizing data, data collected by different sensors are aligned in time to ensure data synchronization.

[0088] Data normalization is used to normalize data collected by different sensors, eliminating differences in data units and magnitudes.

[0089] The feature extraction module is configured to extract key feature information from preprocessed multimodal data, including extracting frequency and time domain features from muscle micro-vibration signals, extracting rate of change features from joint angles, extracting pressure distribution features from plantar pressure, and extracting dynamic change features from gait phase.

[0090] The data fusion module is configured to select appropriate fusion algorithms, including but not limited to weighted average, Bayesian fusion, and Kalman filter fusion, based on application requirements and data characteristics after feature extraction, to fuse multimodal data and form comprehensive perceptual information.

[0091] The technical effects of the above solution are as follows: The joint elastic bandage B integrates multiple sensors, enabling simultaneous acquisition of muscle micro-vibration signals, joint angles, and gait phase data. Simultaneously, the distributed plantar pressure sensor A collects plantar pressure data in real time, achieving multi-dimensional and high-precision monitoring of the elderly's physical condition. The data acquisition module, through the collaborative work of multiple sensors, ensures the comprehensiveness and accuracy of data acquisition, providing a rich data foundation for subsequent analysis. The quality assessment module performs integrity checks, noise detection and assessment, and consistency assessment on the acquired multimodal data, and comprehensively scores the data quality based on the assessment results, effectively identifying low-quality data and ensuring the reliability of subsequent analysis. The data processing module eliminates noise, outliers, and dimensional differences in the data through data cleaning, synchronization, and normalization, improving data usability and consistency, and providing high-quality data support for feature extraction and data analysis. The feature extraction module extracts key feature information from the preprocessed data, accurately reflecting the elderly's physical condition and movement characteristics. The data fusion module, based on application requirements and data characteristics, fuses multimodal data to form comprehensive perceptual information, further enhancing the analytical value and application effect of the data.

[0092] In summary, the multimodal perception unit's efficient data acquisition, processing, and fusion capabilities provide high-quality input data for the cloud-based analysis unit, making the assessment of the Sarcopenia Progression Risk Index (SRI) and Fall Risk Index (FRI) more accurate, thus providing a more reliable basis for the intervention execution unit to generate personalized care plans.

[0093] The quality assessment module, specifically:

[0094] The system checks whether the collected data is missing, contains outliers, or is incomplete.

[0095] Analyze the noise level in the data, assess the signal-to-noise ratio of the data, and determine whether the data is severely interfered with;

[0096] Check whether the data collected by different sensors are consistent in time and logic, for example, whether the joint angles match the gait phase, etc.

[0097] Based on the above evaluation results, a comprehensive score is given for the data quality, and the data quality of the current multimodal data is judged based on the comprehensive score.

[0098] If the overall score is within the preset score range, then the data quality of the current multimodal data is considered high.

[0099] If the overall score is not within the preset score range, the data quality of the current multimodal data is judged to be high or low, and multimodal data is collected again.

[0100] The technical effects of the above-mentioned technical solution are as follows: the multi-dimensional inspection of the quality assessment module can automatically identify and take measures when faced with various abnormal situations in data collection, ensuring that the data entering the analysis stage is of high quality, thereby improving the performance and reliability of the entire device. High-quality data enables the cloud analysis unit to output the sarcopenia progression risk index (SRI) and fall risk index (FRI) more accurately, thereby improving the scientificity and effectiveness of personalized care plans.

[0101] The data acquisition module also includes a calibration module, which is used to periodically calibrate the various sensors integrated on the plantar distributed pressure sensor A and the joint elastic bandage B to ensure the accuracy and reliability of the collected data. The calibrated sensors are then repeatedly tested, and the calibrated data is compared with the data before calibration to evaluate the calibration effect. Based on the comparison results, it is determined whether the calibration has achieved the expected goal. If not, the calibration is repeated.

[0102] The technical effects of the above technical solution are as follows:

[0103] The cloud-based analytics unit includes:

[0104] The model loading module is configured to load a pre-built and trained lightweight LSTM network, input the extracted feature data into the lightweight LSTM network, perform forward propagation through the lightweight LSTM network, extract time series features, and finally output intermediate feature vectors.

[0105] The risk assessment module is configured based on a lightweight LSTM network to determine the sarcopenia progression risk index (SRI) and fall risk index (FRI), specifically:

[0106] By using specific neural network layers or regression models, the intermediate feature vectors output by the lightweight LSTM network are mapped to the sarcopenia progression risk index and the fall risk index.

[0107] Based on the established Sarcopenia Progression Risk Index (SRI) and Fall Risk Index (FRI), the risk is classified into low risk, medium risk, and high risk.

[0108] The results output module is configured to generate a risk assessment report and output the results to the main screen 4 for display.

[0109] The risk determination module verifies and calibrates the determined sarcopenia progression risk index (SRI) and fall risk index (FRI). It compares the determined risk indices with historical data or known standards, adjusts the risk indices based on the comparison results, detects outliers or unreasonable data in the determination results, and finally outputs the verified and calibrated sarcopenia progression risk index (SRI) and fall risk index (FRI).

[0110] The technical effects of the above solution are as follows: The model loading module, by loading a pre-built and trained lightweight LSTM network, can efficiently process time series data, extract key features, and output intermediate feature vectors. The risk determination module, based on the output of the lightweight LSTM network, maps the intermediate feature vectors to the Sarcopenia Progression Risk Index (SRI) and Fall Risk Index (FRI) through specific neural network layers or regression models, achieving accurate quantitative assessment of health risks for the elderly. This quantitative assessment provides a scientific basis for subsequent personalized care plans. At the same time, the risk determination module verifies and calibrates the determined risk indices. This mechanism ensures the accuracy and reliability of the risk assessment results and avoids misjudgments caused by model bias or data anomalies. The result output module generates a risk assessment report and displays the results on the main screen 4 for easy viewing and understanding by users.

[0111] The intervention execution unit includes:

[0112] The treatment plan generation module is configured to generate personalized care plans based on the Sarcopenia Progression Risk Index (SRI) and Fall Risk Index (FRI) output by the cloud-based analytics unit. Specifically:

[0113] Based on the user's health condition, a suitable resistance training plan is generated, including training frequency, intensity, and duration;

[0114] Based on the user's nutritional needs, a personalized nutritional prescription is generated, including recommendations for daily intake of protein, vitamins, and minerals.

[0115] Based on the user's health condition and medication needs, a medication reminder plan is generated, including suggestions for home environment modifications and assistive devices;

[0116] The solution display module is configured to display the generated personalized care plan to the user through the main screen 4, and provide interactive functions, allowing users to view detailed information, adjust the plan and set reminders;

[0117] The nursing intervention module is configured to pre-store electrical stimulation parameters, including intensity, frequency, and duration, that are compatible with the Sarcopenia Progression Risk Index (SRI) and Fall Risk Index (FRI). Based on the SRI and FRI output by the cloud analysis unit, the control patch 3 is adjusted to the corresponding electrical stimulation parameters to provide nursing intervention to the user, while recording the usage of electrical stimulation.

[0118] The technical effects of the above solution are as follows: The solution generation module generates personalized care plans based on the SRI and FRI output by the cloud analysis unit. These plans cover resistance training programs, nutritional prescriptions, medication reminders, home environment modifications, and assistive device recommendations, comprehensively meeting the health needs of different elderly individuals. The solution display module displays the personalized care plans on the main screen and provides interactive functions, allowing users to view detailed information, adjust plans, and set reminders. This enables elderly individuals and their caregivers to easily understand and manage care plans, improving user participation and satisfaction. The care intervention module pre-stores electrical stimulation parameters compatible with SRI and FRI. Based on the output results of the cloud analysis unit, it automatically adjusts the electrical stimulation parameters of the patches to provide precise care interventions for users. It also records the use of electrical stimulation, providing data support for subsequent care effect evaluation.

[0119] Specifically, the calibration module includes:

[0120] The model training submodule is used for:

[0121] The sensitive environmental factors of the plantar distributed pressure sensor A and the various sensors integrated on the joint elastic bandage B are acquired, and the parameters of the sensitive environmental factors of each sensor are collected to obtain the sensitive environmental parameters; wherein, the various sensors integrated on the joint elastic bandage B include a triaxial accelerometer, a triaxial gyroscope, a miniature electromyography sensor array and an inertial measurement unit (IMU);

[0122] Obtain product parameter information and working principle information of each sensor; extract core hardware parameters based on the product parameter information and working principle information of each sensor, and determine the core hardware parameters corresponding to each sensor.

[0123] A coupling analysis was performed on the sensitive environmental parameters and core hardware parameters corresponding to each sensor to determine the coupling influence law for each sensor.

[0124] A hybrid feature set is constructed based on the aforementioned coupling effect law; the hybrid feature set includes an environmental feature set, a hardware feature set, and a coupling feature set.

[0125] Sensitive environmental parameters, core hardware parameters, and hybrid feature sets are divided into training datasets and validation datasets according to a preset ratio.

[0126] The neural network model is iteratively trained based on the training dataset to obtain a data-driven sub-model.

[0127] Based on the working principle of the sensor, the error formula is derived, the quantitative relationship between the embedded hardware parameters and the environmental parameters is determined, and a physical sub-model is constructed.

[0128] By fusing the data-driven sub-model with the physical sub-model, a multi-dimensional environment-hardware hybrid compensation model is obtained.

[0129] The compensation value calculation submodule is used for:

[0130] Select any sensor as the target sensor; acquire the measurement value dataset of the target sensor within a preset period; input the measurement value dataset into a multi-dimensional environment-hardware hybrid compensation model to determine the compensation value corresponding to each measurement value in the measurement value dataset;

[0131] Calculate the mean of the compensation values ​​corresponding to each measurement value in the measurement value dataset, and use it as the compensation mean of the measurement value dataset;

[0132] The calibration submodule is used for:

[0133] The compensation coefficient of the target sensor is determined based on the compensation mean.

[0134] The target sensor is calibrated based on the compensation coefficient;

[0135] It traverses all sensors and completes the calibration of various types of sensors.

[0136] In this embodiment, the sensitive environmental parameters include temperature, humidity, air pressure; magnetic field strength; vibration frequency and amplitude; and skin condition parameters.

[0137] In this embodiment, the core hardware parameters include:

[0138] The core hardware parameters of the pressure sensor include: diaphragm material, thickness, and effective pressure-bearing area; data acquisition method: production records + laser thickness gauge.

[0139] The core hardware parameters of the triaxial accelerometer include: measurement range, sensitive axis alignment error, core material, and zero-drift temperature coefficient; the acquisition method is: factory calibration report + precision turntable test.

[0140] The core hardware parameters of a three-axis gyroscope include: range, zero-bias stability, core material, and magnetic shielding coefficient; the acquisition method is: impedance analyzer + constant temperature chamber testing.

[0141] The core hardware parameters of the miniature electromyography sensor array include: electrode material, electrode diameter, spacing, and preamplifier gain; the acquisition method is: microscopic measurement + circuit parameter tester.

[0142] The core hardware parameters of the IMU sensor include: internal accelerometer / gyroscope model, mounting axis orthogonality error, and time synchronization error; acquisition method: three-dimensional coordinate measuring machine + synchronization signal generator.

[0143] In this embodiment, the coupling influence law corresponding to each sensor is determined, including:

[0144] Pressure sensor: Diaphragm material (hardware) determines the temperature coefficient of elastic modulus; silicon diaphragms are more sensitive to temperature than ceramic diaphragms; diaphragm thickness (hardware) amplifies the effect of temperature.

[0145] Triaxial accelerometer: The stiffness of the core material (hardware) changes with temperature; quartz cores are more stable than silicon cores; the measurement range (hardware) is coupled with the vibration amplitude (environment).

[0146] Three-axis gyroscope: The permeability of the core material (hardware) is affected by the magnetic field (environment); zero bias stability (hardware) is synergistic with temperature (environment).

[0147] Miniature electromyography sensor array: electrode material (hardware) coupled with skin humidity (environment).

[0148] IMU sensor: The mounting axis orthogonality error (hardware) is coupled with the vibration frequency (environment).

[0149] In this embodiment, a hybrid feature set is constructed based on the coupling effect law, including:

[0150] Based on the above coupling rules, a hybrid feature set containing the following dimensions is constructed:

[0151] Environmental characteristics: temperature, humidity, magnetic field strength, vibration amplitude, skin moisture;

[0152] Hardware characteristics: diaphragm thickness (pressure), core material (accelerometer), magnetic core material (gyroscope), electrode spacing (electromyography), installation error (IMU);

[0153] Coupling characteristics: temperature-diaphragm thickness interaction (pressure), magnetic field-core material interaction (gyroscope), vibration-installation error interaction (IMU), etc.

[0154] In this embodiment, the neural network model is iteratively trained based on the training dataset to obtain a data-driven sub-model, including: iteratively training the neural network model based on the training dataset to learn nonlinear mapping relationships; for example: electromyography sensor: input "skin humidity + electrode material + spacing", output "baseline drift error"; IMU: input "vibration frequency + installation error + temperature", output "attitude calculation error".

[0155] In this embodiment, an error formula is derived based on the sensor's working principle to determine the quantization relationship between embedded hardware parameters and environmental parameters, and a physical sub-model is constructed. This includes: deriving the error formula based on the sensor's working principle and determining the quantization relationship between embedded hardware parameters and environmental parameters. Specific quantization relationships include: Pressure sensor: Error ΔP = × Temperature × Diaphragm thickness + × Humidity × Membrane material coefficient ( , (For experimental fitting coefficients); accelerometer: error = ,in This is the elastic modulus at temperature T (related to the core material). For the initial measurement of acceleration, Indicates the core's elastic modulus at the reference temperature; gyroscope: error. = × Magnetic field strength × Core permeability + × Temperature × Zero-bias coefficient; , Coupling coefficient; Electromyography sensor: Error = (1 - SNR (humidity, electrode material)) × , Represents the raw measurement value of the electromyography sensor; IMU: error. = Installation error × Vibration amplitude × Frequency coefficient (θ is the attitude angle). The physical sub-model imposes principle constraints on the output results.

[0156] The working principle and beneficial effects of the above technical solution are as follows: It acquires the sensitive environmental factors of the plantar distributed pressure sensor A and the joint elastic bandage B (which integrates multiple sensors such as a triaxial accelerometer, triaxial gyroscope, miniature electromyography sensor array, and inertial measurement unit IMU), and collects the parameters of these factors to obtain the sensitive environmental parameters. This step can collect relevant data on the impact of the sensor's working environment, obtain product parameter information and working principle information of each sensor, and extract the core hardware parameters from this information to determine the core hardware parameters corresponding to each sensor. This clarifies the parameters related to the sensor's own hardware characteristics; and it determines the sensitive environmental parameters and core hardware parameters of each sensor. The parameters are coupled to identify the patterns of their mutual influence, thus determining the coupling effect patterns. Based on this, a hybrid feature set is constructed, which includes an environmental feature set (from sensitive environmental parameters), a hardware feature set (from core hardware parameters), and a coupling feature set (reflecting the mutual influence between the two). The sensitive environmental parameters, core hardware parameters, and hybrid feature set are divided into training datasets and validation datasets according to a preset ratio. The neural network model is iteratively trained using the training dataset to obtain a data-driven sub-model. This process allows the model to learn the patterns in the data. Simultaneously, based on the sensor's working principle, the error formula is derived to determine the quantitative relationship between the embedded hardware parameters and environmental parameters, thus constructing a physical sub-model. Finally, the data-driven sub-model and the physical sub-model are fused to obtain a multi-dimensional environment-hardware hybrid compensation model. This model integrates information from both data-driven and physical principles. Any sensor is selected as the target sensor, and its measurement data dataset within a preset period is obtained. This dataset is input into the multi-dimensional environment-hardware hybrid compensation model, which determines the compensation value corresponding to each measurement value in the dataset based on previously learned information about the environment, hardware, and their coupling relationships. The mean of these compensation values ​​is then calculated, resulting in the average compensation value of the measurement dataset. The compensation coefficient of the target sensor is determined based on the obtained average compensation value, and this coefficient is used to calibrate the target sensor. By iterating through all sensors, calibration of various types of sensors is completed. This means that precise calibration can be performed on each sensor according to its specific working environment and hardware characteristics. Since this technical solution calibrates various types of sensors integrated on the plantar distributed pressure sensor A and the joint elastic bandage B (such as a triaxial accelerometer, triaxial gyroscope, miniature electromyography sensor array, and inertial measurement unit IMU), it has strong versatility. Multiple sensors can be calibrated in one system, reducing the overall system error and improving the reliability of the entire system.

[0157] Specifically, a comprehensive score is given for data quality, including:

[0158] Based on the data integrity assessment results, noise detection and assessment results, and data consistency assessment results, the quality score of the collected multimodal data is calculated.

[0159]

[0160] in, This indicates the quality score of the collected multimodal data; This represents the total number of modes in multimodal data; Indicates the first Weights of modal data; Indicates the first Time decay factor for modal data; Indicates the first Completeness score of modal data; Indicates the first Noise scoring for modal data; Indicates the first Consistency score of modal data; , , They represent the first The integrity score index coefficient, noise score index coefficient, and consistency score index coefficient of the modal data; This represents the cross-modal collaborative correction coefficient.

[0161] In this embodiment,

[0162]

[0163] in, Indicates the first Sensitivity coefficient for the integrity of modal data; Indicates the first The amount of missing data in each modality; Indicates the first The total amount of data for each modality; Indicates the first Outliers in modal data affect weights; Indicates the first The number of outlier types in each modality of data; Indicates the first The number of outliers of type k in a modal dataset; Indicates the first Standard threshold for the k-th class of data in a modal dataset;

[0164] In this embodiment,

[0165]

[0166] in, Indicates the first The actual signal-to-noise ratio of the modal data; Indicates the first The baseline signal-to-noise ratio for modal data; Indicates the first Signal-to-noise ratio sensitivity index for modal data; Indicates the number of noise frequency bands; Indicates the first Noise weighting of the f-th frequency band in modal data; No. The actual noise energy of the f-th frequency band in the modal data; Indicates the first The maximum permissible noise energy in the f-th frequency band of the modal data;

[0167] In this embodiment,

[0168]

[0169] in, Indicates the first Internal consistency weights for modal data; Indicates the number of associated sensors; Indicates the first The association weights between modal data and the j-th sensor; Indicates the first The deviation between the modal data and the actual value of the j-th sensor; Indicates the first The maximum permissible deviation between the modal data and the j-th sensor; Indicates the first The cross-correlation coefficient between one modality of data and other modality of data.

[0170] In this embodiment, It can be dynamically adjusted.

[0171] In this embodiment, The range of values ​​for is: 1≤ ≤5.

[0172] In this embodiment, The range of values ​​for is: .

[0173] In this embodiment, The range of values ​​for is: .

[0174] In this embodiment, The range of values ​​for is: .

[0175] In this embodiment, The range of values ​​for is: .

[0176] The working principle and beneficial effects of the above technical solution are as follows: By calculating the integrity score, noise score, and consistency score separately, and comprehensively considering multiple factors (such as the amount of missing data, outliers, signal-to-noise ratio, sensor bias, etc.) to determine the quality score of multimodal data, the quality of multimodal data can be comprehensively evaluated, avoiding the one-sidedness of single-factor evaluation. The integrity score reflects the completeness of the data, the noise score reflects the purity of the data, and the consistency score reflects the consistency of the data within and across modalities. The combined scores of these different aspects can accurately reflect the quality status of multimodal data in different dimensions, which is helpful for subsequent data processing, analysis, and decision-making.

[0177] Specifically, this embodiment also proposes a monitoring method for an integrated device for sarcopenia care in the elderly based on the Internet of Things (IoT), which includes the following steps:

[0178] According to the preset sampling frequency, multimodal data including muscle micro-vibration signals, joint angles, gait phases and plantar pressure data are collected synchronously and uploaded to the cloud analysis unit.

[0179] After receiving multimodal data, the cloud-based analysis unit performs a quality assessment. If the data quality does not meet the standards, it is re-collected. The qualified multimodal data is pre-processed and key features are extracted.

[0180] The extracted feature data is input into a lightweight LSTM network, which outputs the sarcopenia progression risk index (SRI) and fall risk index (FRI), and the risks are classified.

[0181] Based on the Sarcopenia Progression Risk Index (SRI) and Fall Risk Index (FRI), a personalized care plan is automatically generated, which includes resistance training, nutritional prescriptions, medication reminders, and environmental modification suggestions. This plan is then displayed to the user on the device's main screen 4, while electrical stimulation is applied via the patch 3.

[0182] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0183] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention.

Claims

1. An integrated care device for sarcopenia in the elderly based on the Internet of Things, comprising a main body of the device (1), characterized in that, The device body (1) is provided with a handrail (11) on the top, a foot plate (2) is provided on the bottom of the back of the device body (1), patches (3) are provided on both sides of the device body (1), and a main screen (4) is provided on the top of the front of the device body (1). The IoT-based integrated nursing device for sarcopenia in the elderly also includes a multimodal sensing unit, a cloud analysis unit, and an intervention execution unit. The multimodal sensing unit includes a plantar distributed pressure sensor A and a joint elastic bandage B. The plantar distributed pressure sensor A is set on the plantar plate (2), and the joint elastic bandage B is set on the back of the main body (1) of the device. The multimodal sensing unit is configured to acquire muscle micro-vibration signals, joint angles and gait phases through the joint elastic band B, acquire plantar pressure data through the plantar distributed pressure sensor A, obtain multimodal data based on the acquired information, and perform data quality assessment, processing and fusion. Among them, the various types of sensors integrated on the plantar distributed pressure sensor A and the joint elastic bandage B are calibrated, and the data quality is comprehensively scored based on the data quality assessment results. The score reflects the quality status of multimodal data in different dimensions. A comprehensive score is given for data quality, including: Based on the data integrity assessment results, noise detection and assessment results, and data consistency assessment results, the quality score of the collected multimodal data is calculated. in, This indicates the quality score of the collected multimodal data; This represents the total number of modes in multimodal data; Indicates the first Weights of modal data; Indicates the first Time decay factor for modal data; Indicates the first Completeness score of modal data; Indicates the first Noise scoring for modal data; Indicates the first Consistency score of modal data; , , They represent the first The integrity score index coefficient, noise score index coefficient, and consistency score index coefficient of the modal data; Indicates the cross-modal collaborative correction coefficient; The cloud-based analysis unit is configured to receive and analyze multimodal data acquired by the multimodal perception unit, input the multimodal data into a lightweight LSTM network, and output the sarcopenia progression risk index (SRI) and fall risk index (FRI). The intervention execution unit is configured to automatically generate a personalized nursing plan that includes resistance training, nutritional prescriptions, medication adherence reminders and environmental modification suggestions based on the results output by the cloud analysis unit, and display it through the main screen (4), while performing electrical stimulation nursing intervention through the patch (3).

2. The integrated IoT-based nursing device for sarcopenia in the elderly according to claim 1, characterized in that, The multimodal sensing unit includes: The data acquisition module is configured to simultaneously acquire muscle micro-vibration signals, joint angles, and gait phases using a joint elastic bandage B at a pre-set sampling frequency. At the same time, it uses a plantar distributed pressure sensor A to acquire plantar pressure data in real time. Among them, the joint elastic band B integrates a triaxial accelerometer, a triaxial gyroscope, a miniature electromyography sensor array and an inertial measurement unit (IMU). The miniature electromyography sensor array acquires muscle micro-vibration signals, the triaxial accelerometer and triaxial gyroscope perform gait phase recognition, and the inertial measurement unit (IMU) monitors joint angle changes. The quality assessment module is configured to perform quality assessment on the collected multimodal data, including data integrity check, noise detection and assessment, and data consistency assessment, and to give a comprehensive score for data quality based on the assessment results; The data processing module is configured to preprocess multimodal data for quality assessment.

3. The integrated IoT-based nursing device for sarcopenia in the elderly according to claim 2, characterized in that, The quality assessment module specifically includes: The system checks whether the collected data is missing, contains outliers, or is incomplete. Analyze the noise level in the data, assess the signal-to-noise ratio of the data, and determine whether the data is severely interfered with; Check whether the data collected by different sensors are consistent in time and logic; Based on the above evaluation results, a comprehensive score is given for the data quality, and the data quality of the current multimodal data is judged based on the comprehensive score. If the overall score is within the preset score range, then the data quality of the current multimodal data is considered high. If the overall score is not within the preset score range, the data quality of the current multimodal data is judged to be high or low, and multimodal data is collected again.

4. The integrated IoT-based nursing device for sarcopenia in the elderly according to claim 2, characterized in that, The data acquisition module also includes a calibration module, which is used to periodically calibrate the various sensors integrated on the plantar distributed pressure sensor A and the joint elastic bandage B to ensure the accuracy and reliability of the collected data. The calibrated sensors are then repeatedly tested, and the calibrated data is compared with the data before calibration to evaluate the calibration effect. Based on the comparison results, it is determined whether the calibration has achieved the expected goal. If not, the calibration is repeated.

5. The integrated IoT-based nursing device for sarcopenia in the elderly according to claim 1, characterized in that, The cloud-based analytics unit includes: The model loading module is configured to load a pre-built and trained lightweight LSTM network, input the extracted feature data into the lightweight LSTM network, perform forward propagation through the lightweight LSTM network, extract time series features, and finally output intermediate feature vectors. The risk assessment module is configured based on a lightweight LSTM network to determine the sarcopenia progression risk index (SRI) and fall risk index (FRI), specifically: By using specific neural network layers or regression models, the intermediate feature vectors output by the lightweight LSTM network are mapped to the sarcopenia progression risk index and the fall risk index. Based on the established Sarcopenia Progression Risk Index (SRI) and Fall Risk Index (FRI), the risk is classified into low risk, medium risk, and high risk. The results output module is configured to generate a risk assessment report and output the results to the main screen (4) for display.

6. The integrated IoT-based nursing device for sarcopenia in the elderly according to claim 5, characterized in that, The risk determination module verifies and calibrates the determined sarcopenia progression risk index (SRI) and fall risk index (FRI). By comparing the determined risk indices with historical data or known standards, the risk indices are calibrated and adjusted based on the comparison results. Abnormal values ​​or unreasonable data in the determination results are detected, and finally, the verified and calibrated sarcopenia progression risk index (SRI) and fall risk index (FRI) are output.

7. The integrated IoT-based nursing device for sarcopenia in the elderly according to claim 1, characterized in that, The intervention execution unit includes: The treatment plan generation module is configured to generate personalized care plans based on the Sarcopenia Progression Risk Index (SRI) and Fall Risk Index (FRI) output by the cloud-based analytics unit. Specifically: Based on the user's health condition, a suitable resistance training plan is generated, including training frequency, intensity, and duration; Based on the user's nutritional needs, a personalized nutritional prescription is generated, including recommendations for daily intake of protein, vitamins, and minerals. Based on the user's health condition and medication needs, a medication reminder plan is generated, including suggestions for home environment modifications and assistive devices; The solution display module is configured to display the generated personalized care solution to the user through the main screen (4) and provide interactive functions, allowing the user to view detailed information, adjust the plan and set reminders; The nursing intervention module is configured to pre-store electrical stimulation parameters that are compatible with the sarcopenia progression risk index (SRI) and fall risk index (FRI), including intensity, frequency and duration. Based on the sarcopenia progression risk index (SRI) and fall risk index (FRI) output by the cloud analysis unit, the control patch (3) is adjusted to the corresponding electrical stimulation parameters to perform nursing intervention on the user, and the use of electrical stimulation is recorded at the same time.

8. The integrated IoT-based nursing device for sarcopenia in the elderly according to claim 4, characterized in that, The calibration module includes: The model training submodule is used for: The sensitive environmental factors of the plantar distributed pressure sensor A and the various sensors integrated on the joint elastic bandage B are acquired, and the parameters of the sensitive environmental factors of each sensor are collected to obtain the sensitive environmental parameters; wherein, the various sensors integrated on the joint elastic bandage B include a triaxial accelerometer, a triaxial gyroscope, a miniature electromyography sensor array and an inertial measurement unit (IMU); Obtain product parameter information and working principle information of each sensor; extract core hardware parameters based on the product parameter information and working principle information of each sensor, and determine the core hardware parameters corresponding to each sensor. A coupling analysis was performed on the sensitive environmental parameters and core hardware parameters corresponding to each sensor to determine the coupling influence law for each sensor. A hybrid feature set is constructed based on the aforementioned coupling effect law; the hybrid feature set includes an environmental feature set, a hardware feature set, and a coupling feature set. Sensitive environmental parameters, core hardware parameters, and hybrid feature sets are divided into training datasets and validation datasets according to a preset ratio. The neural network model is iteratively trained based on the training dataset to obtain a data-driven sub-model. Based on the working principle of the sensor, the error formula is derived, the quantitative relationship between the embedded hardware parameters and the environmental parameters is determined, and a physical sub-model is constructed. By fusing the data-driven sub-model with the physical sub-model, a multi-dimensional environment-hardware hybrid compensation model is obtained. The compensation value calculation submodule is used for: Select any sensor as the target sensor; acquire the measurement value dataset of the target sensor within a preset period; input the measurement value dataset into a multi-dimensional environment-hardware hybrid compensation model to determine the compensation value corresponding to each measurement value in the measurement value dataset; Calculate the mean of the compensation values ​​corresponding to each measurement value in the measurement value dataset, and use it as the compensation mean of the measurement value dataset; The calibration submodule is used for: The compensation coefficient of the target sensor is determined based on the compensation mean. The target sensor is calibrated based on the compensation coefficient; It traverses all sensors and completes the calibration of various types of sensors.

Images