Early warning and rescue nursing system for falling of old people

Abstract

The invention relates to the technical field of intelligent health monitoring, and discloses an early warning and rescue nursing system for falling of old people. The sensor module is used for collecting activity data and environment data of the old people in real time, and the data fusion and behavior analysis module is used for fusing the collected activity data and environment data and analyzing the health state and the falling risk of the old people; the intelligent early warning and personalized response module is used for triggering an early warning mechanism according to the data analysis result and generating a personalized response strategy; and the emergency rescue path planning and automatic response module is used for calculating an optimal rescue path and guiding rescue workers after the old people fall down. A unified sensing system is formed by adopting a multi-dimensional sensing fusion sensor module, and continuous and three-dimensional monitoring of daily behaviors and the environment of the old people is achieved. The integrity and accuracy of monitoring are remarkably improved, and the problems of single information, high false alarm rate and insufficient space coverage are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent health monitoring, and in particular to a fall warning and rescue care system for the elderly. Background Art

[0002] Falls among the elderly have become a leading cause of accidental injury and death. To address this issue, a variety of wearable fall monitoring products, such as smart bracelets and sensor belts, have emerged on the market. These products primarily utilize accelerometers, detecting changes in gravity to determine if a fall has occurred. However, in practice, these products have significant limitations. First, they are highly wearable, and elderly people may forget to wear them or wear them improperly, which can easily lead to blind spots in monitoring. Second, the data dimensionality of a single sensor is insufficient to fully reflect the elderly's environment and behavioral characteristics, making misjudgments and omissions highly likely. Especially in complex indoor spaces, relying solely on wearable devices to collect data is difficult to ensure complete coverage and continuous monitoring.

[0003] Most current behavior recognition systems use static thresholds or rule-matching strategies. For example, a fall is identified when acceleration exceeds a certain value. While simple to implement, this approach struggles to adapt to individual behavioral differences and gait variations, is susceptible to noise, and has a high rate of false positives. Furthermore, these methods lack the ability to model the temporal continuity of behavioral changes, making them unable to accurately identify pre-fall warning signs. This limits both the system's intelligence and robustness.

[0004] Furthermore, some existing early warning systems employ fixed strategies, such as sending notifications to guardians or the platform upon detecting an anomaly. This "single-track" response logic ignores individual differences and historical health records of users and fails to establish a dynamic adjustment mechanism. For example, for patients with chronic diseases or elderly people recovering from illness, their movement patterns may deviate from "normal" standards. Fixed thresholds are clearly inappropriate and can easily lead to false alarm fatigue, causing users to lose trust in the system. Summary of the Invention

[0005] In response to the shortcomings of the existing technology, the present invention provides a fall warning and rescue care system for the elderly, which solves the problems of the existing elderly fall monitoring methods in the technology, such as single perception information, low recognition accuracy, and rigid warning response.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: A fall warning and rescue care system for the elderly, comprising:

[0007] A sensor module for collecting real-time activity data and environmental data of the elderly. The sensor module includes a pressure-sensing floor, a wearable device, an environmental sensing device, and a positioning system.

[0008] The data fusion and behavior analysis module is connected to the sensor module to fuse the collected activity data and environmental data and analyze the health status and fall risk of the elderly;

[0009] The intelligent early warning and personalized response module is connected to the data fusion and behavior analysis module to trigger the early warning mechanism and generate personalized response strategies based on the data analysis results;

[0010] The emergency rescue path planning and automated response module is connected to the intelligent early warning and personalized response module to calculate the optimal rescue path and guide rescuers after the elderly fall.

[0011] Preferably, the pressure-sensing floor in the sensor module is used to monitor the changes in the pace of the elderly while walking, and the wearable device includes a smart bracelet or a smart insole with a built-in accelerometer and a gyroscope, which are used to collect the elderly's movement status and gait data in real time, and jointly analyze the acceleration signal output by the accelerometer and the angular velocity signal output by the gyroscope to determine a fall event.

[0012] Preferably, the data fusion and behavior analysis module includes:

[0013] Kalman filter unit, used to denoise multi-source data collected by sensors;

[0014] Gait recognition unit, used to identify the current gait characteristics of the elderly;

[0015] The fall risk assessment unit is used to assess the fall risk based on the identification results and health status.

[0016] Preferably, the Kalman filter unit performs denoising on the acceleration signal and the angular velocity signal using a Kalman filter algorithm, and the update formula of the Kalman filter is:

[0017]

[0018] in, represents the state estimate at the current time k, represents the state prediction at the current moment k, K k represents the Kalman gain, y k Indicates the measured value at the current moment, H k represents the measurement matrix, Represents the measurement residual.

[0019] Preferably, the gait recognition unit processes the gait data through a convolutional neural network and a long short-term memory network to analyze the movement state of the elderly and determine whether there is a risk of falling. The convolution operation of the convolutional neural network layer is:

[0020] h l =ReLU(W l *x l-1 +b l );

[0021] Among them, h l Represents the convolutional layer output of the first layer, W l represents the first layer of convolution kernel, x l-1 represents the input feature map of the l-1 layer, b l represents the bias term of the first layer, * represents the convolution operator, and ReLU(·) represents the activation function.

[0022] Preferably, the intelligent early warning and personalized response module includes:

[0023] Personalized health record generation unit, used to generate and dynamically adjust individualized risk models;

[0024] A deep learning assessment unit to predict fall probability and optimize early warning response strategies;

[0025] Personalized response strategy generation unit, used to formulate individualized intervention or rescue processes.

[0026] Preferably, the personalized health profile generation unit generates a personalized health profile based on the elderly's health data and historical activity data, and adjusts the warning threshold when the elderly have a higher risk of falling, so that the warning sensitivity matches the elderly's health status.

[0027] Preferably, the deep learning evaluation unit uses a reinforcement learning algorithm to optimize the fall warning mechanism, and dynamically adjusts the response threshold of the fall warning based on the health status and behavior patterns of the elderly. The update formula of the reinforcement learning is:

[0028]

[0029] Among them, Q(s t ,a t ) represents the current state s t Next take action a t Q value, α represents the learning rate, r t+1 represents the immediate reward after executing the action, γ represents the discount factor, s t Indicates the current state, s t+1 Indicates the state at the next moment, a t represents the action at the current moment, a′ represents the best action among all possible actions at the next moment, In state s t+1 The maximum Q value among all actions.

[0030] Preferably, the emergency rescue path planning and automated response module includes:

[0031] A rescue path planning unit, used to calculate the optimal path from the fall site to the rescuer;

[0032] Rescue personnel navigation unit, used to provide real-time route guidance and location feedback.

[0033] Preferably, the rescue path planning unit uses an A algorithm to calculate the shortest path from the fall scene to the location of the rescuer. The calculation formula of the A algorithm is:

[0034] f(n)=g(n)+h(n);

[0035] Among them, f(n) represents the evaluation value of node n, which represents the estimated total cost from the starting point to the target, g(n) represents the actual cost of node n from the starting point to the current node, and h(n) represents the heuristic estimated cost from node n to the target node.

[0036] The present invention provides a fall warning and rescue care system for the elderly. It has the following beneficial effects:

[0037] 1. This invention utilizes a multi-dimensional sensor module to integrate pressure-sensitive flooring, wearable devices, environmental sensing devices, and a positioning system into a unified sensing system, enabling continuous, three-dimensional monitoring of the elderly's daily behaviors and surroundings. This structure significantly improves the integrity and accuracy of monitoring without interfering with user behavior. Compared to existing solutions that rely on single-point devices or monitor only through wearable devices, this solution addresses the issues of limited information, high false alarm rates, and insufficient spatial coverage.

[0038] 2. This invention uses a joint model based on convolutional neural networks and long-short-term memory networks to model and recognize temporal features of elderly behavior. This technology automatically extracts key features from multi-source sensor data and dynamically captures abnormal changes, forming an intelligent judgment mechanism for pre-fall signs. Compared to existing methods that rely on fixed thresholds or static rules, this method not only effectively improves judgment accuracy but also addresses the problems of traditional methods such as slow response and weak generalization when dealing with sudden abnormal behaviors.

[0039] 3. By incorporating Q-learning reinforcement learning into the fall risk assessment and response process, this invention dynamically optimizes early warning strategies and intervention thresholds based on users' historical health data and behavioral feedback. The system continuously learns and adjusts during operation, forming a personalized risk decision-making logic. This significantly improves the adaptability and timeliness of early warning responses, particularly for high-risk elderly groups such as those with chronic diseases and mobility impairments, enabling more precise alignment with actual needs.

[0040] 4. This invention integrates an indoor positioning system with a dynamic A-path planning algorithm, combined with building floor plans and environmental perception information, to automatically calculate the optimal route from the rescuer's current location to the incident site. Real-time navigation instructions are then delivered via mobile terminals, enabling rapid and accurate on-site rescue. Compared to existing technologies that rely on voice notifications or manual dispatch by on-duty personnel, this significantly shortens response time and solves the problems of unclear routes and inefficient resource allocation. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is a system framework diagram of the present invention;

[0042] Figure 2 Schematic diagram of the data fusion and behavior analysis module of the present invention;

[0043] Figure 3 This is a schematic diagram of the intelligent early warning and personalized response module of the present invention;

[0044] Figure 4 Schematic diagram of the emergency rescue path planning and automated response module of the present invention. DETAILED DESCRIPTION

[0045] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the present specification. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0046] Please see the attached Figure 1 -Attached Figure 4 , an embodiment of the present invention provides a fall warning and rescue care system for the elderly, comprising;

[0047] A sensor module for collecting real-time activity data and environmental data of the elderly. The sensor module includes a pressure-sensing floor, a wearable device, an environmental sensing device, and a positioning system.

[0048] Specifically, the sensor module in this embodiment is responsible for collecting real-time activity and environmental data from the elderly and transmitting it to the subsequent processing module. The sensor module comprises multiple components, including a pressure-sensing floor, a wearable device, an environmental sensing device, and a positioning system. These components work together in the system to accurately monitor the elderly's activity status and environmental changes, providing data support for fall warning and emergency response.

[0049] In this embodiment, the sensor module consists of four main components: a pressure-sensing floor, a wearable device, an environmental sensing device, and a positioning system. Each component has independent functions, but together they provide real-time monitoring of the elderly's health status and environmental changes. Specifically, the sensor module collects multidimensional data to provide comprehensive awareness of the elderly's activities and environment. This data is then transmitted to the subsequent data fusion and behavior analysis module to facilitate fall risk assessment and early warning.

[0050] Pressure-sensing floors are installed in areas where seniors spend their daily time (such as living rooms, bedrooms, and bathrooms) to monitor their gait in real time. The pressure sensor array detects changes in the elderly's footsteps as they walk and detects gait characteristics such as standing time and walking speed. By monitoring these changes in footsteps in real time, the system can identify any abnormal behavior, such as a sudden pause or fall.

[0051] In one possible implementation, a pressure-sensing floor could be equipped with an array of multiple pressure sensors, combined with processing circuitry, to calculate the elderly person's gait parameters in real time. These gait parameters are then compared with historical data, allowing the system to determine whether the elderly person is at risk of falling. If the system detects a sudden pause and prolonged inactivity, it will issue a warning signal.

[0052] Wearable devices, such as smart bracelets or smart insoles, contain built-in accelerometers and gyroscopes. These devices collect dynamic information about the elderly's body, including movement status, gait changes, and tilt angles. The accelerometer measures the elderly's acceleration signal, while the gyroscope measures the elderly's angular velocity signal. These two signals work together to help the system determine the elderly's movement status.

[0053] Specifically, the acceleration and angular velocity signals provided by the accelerometer and gyroscope are jointly analyzed using a signal processing algorithm to accurately determine whether the elderly person has fallen. In some embodiments, the system analyzes movement status and gait changes in combination with historical behavioral data. If the system detects a sudden change inconsistent with a typical gait (such as a large change in acceleration and an abnormal change in gyroscope angular velocity), it will determine that the elderly person may have fallen.

[0054] Environmental sensing devices collect multidimensional data about the elderly's surroundings. This data includes temperature and humidity, light intensity, air quality, and the status of doors and windows. This data helps the system further analyze whether the elderly's activities are consistent with normal living conditions. For example, temperature and humidity sensors can measure the comfort level of the living environment, while light intensity sensors can help determine whether the elderly are moving in dimly lit environments, a change that could increase the risk of falls.

[0055] In one possible implementation, environmental sensing devices can also include smoke detectors and door magnetic switches. By monitoring these sensor data in real time, the system can automatically analyze and respond accordingly when an abnormality occurs. If the system detects a sudden high temperature or smoke signal, it may be considered that the elderly are experiencing an emergency, thereby triggering the corresponding emergency warning mechanism.

[0056] The positioning system plays a crucial role in the sensor module of this invention, particularly in emergency response situations. This system typically utilizes a Wi-Fi or ultra-wideband (UWB)-based positioning module to accurately track the location of the elderly. In the event of a fall, the system can immediately determine the specific location of the fall based on the location information provided by the positioning system, providing data support for subsequent emergency rescue efforts.

[0057] In one possible implementation, the positioning system can be combined with environmental sensing devices to assist in determining whether a fall has occurred by monitoring the indoor environment and the elderly's movements in real time. For example, if the system detects an elderly person entering a high-risk area (such as a slippery surface or an obstacle area), it will automatically adjust the warning strategy and report their location to relevant personnel.

[0058] The data fusion and behavior analysis module is connected to the sensor module to fuse the collected activity data and environmental data and analyze the health status and fall risk of the elderly;

[0059] Specifically, the data fusion and behavior analysis module is electrically connected to the sensor module and is used to jointly process the activity and environmental data collected by the sensor module. This module forms the core unit for behavior recognition and risk analysis in the fall warning system, enabling continuous operations such as data noise reduction, feature recognition, and risk assessment.

[0060] The data fusion and behavior analysis module includes a Kalman filter unit, a gait recognition unit, and a fall risk assessment unit. The three functional units are connected in a logical sequence and work together to complete a closed-loop process from raw data input to risk output.

[0061] The Kalman filter unit includes a prediction processing unit and an update processing unit, which are used to construct the state prediction and observation correction processes respectively. This unit uses the Kalman filter algorithm to perform time-series filtering on the acceleration and angular velocity signals, eliminating high-frequency noise and irregular disturbances, thereby improving data stability.

[0062] In the implementation, the Kalman filter update step follows the following formula:

[0063]

[0064] in, represents the state estimate at the current time k, represents the state prediction at the current moment k, K k represents the Kalman gain, y k Indicates the measured value at the current moment, H k represents the measurement matrix, Represents the measurement residual.

[0065] The filtered multi-dimensional motion data is transmitted to the gait recognition unit in real time for analyzing the current motion pattern of the elderly.

[0066] The gait recognition unit includes a feature extraction unit and a behavior modeling unit. The feature extraction unit constructs a multi-layer convolution structure through a convolutional neural network (CNN) to extract local change patterns in multi-source sensor data.

[0067] In the exemplary structure, the convolution operation is as follows:

[0068] h l =ReLU(W l *x l-1 +b l );

[0069] Among them, h1 represents the convolutional layer output of the first layer, W l represents the first layer of convolution kernel, x l-1 represents the input feature map of the l-1 layer, b l represents the bias term of the first layer, * represents the convolution operator, and ReLU(·) represents the activation function.

[0070] In some embodiments, the CNN output is used as a time series embedding vector and input into an LSTM network for time series modeling.

[0071] The LSTM internal gating mechanism captures the correlation between previous and subsequent gaits, retains and forgets behavioral features, and forms a dynamic response to behavioral variations.

[0072] In a specific implementation, the recognition module can distinguish between normal gait and patterns such as rapid acceleration and rotation angle changes before a fall, providing a basis for subsequent risk assessment.

[0073] The fall risk assessment unit receives the gait recognition results and makes a multi-dimensional judgment based on the health status data and environmental factors.

[0074] In some embodiments, the unit includes a feature assessment subunit and a risk scoring subunit, and quantitatively weights multiple risk factors by constructing a weight function.

[0075] In general, risk factors include but are not limited to gait instability, body lean angle, and gait continuity interruption rate.

[0076] In one possible implementation, the risk score uses a linear weighted model:

[0077] R=α1S var +α2θ dev +α3T loss ;

[0078] Among them, R is the comprehensive risk score; S var is the stride variation coefficient; θ dev is the attitude deviation angle; T loss is the loss ratio of continuous actions; α1, α2, and α3 are experience weight coefficients.

[0079] In this embodiment, the scoring result is compared with a threshold value and a risk level label is output. The system decides whether to trigger an early warning response based on the label.

[0080] Furthermore, the system can also store historical recognition results for use by the long-term health trend analysis module to form an individualized behavioral portrait.

[0081] The data fusion and behavior analysis module communicates bidirectionally with the sensor module in a logical structure, and has a data interface with the response module or remote monitoring platform to achieve a closed loop of behavior perception and risk control.

[0082] The intelligent early warning and personalized response module is connected to the data fusion and behavior analysis module to trigger the early warning mechanism and generate personalized response strategies based on the data analysis results;

[0083] Specifically, the Intelligent Early Warning and Personalized Response Module in this embodiment is connected to the output of the Data Fusion and Behavior Analysis Module, receiving the fall risk score and behavior recognition results provided by it. It then generates differentiated response strategies based on individual health characteristics and triggers corresponding early warning or rescue processes. Through dynamic learning and strategy optimization mechanisms, this module achieves closed-loop management from risk perception to emergency response, ensuring timely and adaptable intervention.

[0084] In this embodiment, the intelligent early warning and personalized response module consists of a personalized health record generation unit, a deep learning evaluation unit, and a personalized response strategy generation unit. The three are interconnected through a data bus and a logic interface to form a progressive processing link.

[0085] Generally speaking, the personalized health record generation unit is used to integrate historical health data and real-time behavioral characteristics to build an individualized risk model.

[0086] Specifically, the unit reads the user's basic health parameters from a local or cloud database, including but not limited to age, chronic disease history, balance ability test results, and past fall records, and performs correlation analysis with real-time collected gait stability indicators and environmental risk factors.

[0087] In one possible implementation, the process of generating a health record can be expressed as follows:

[0088] P profile =β1H base +β2B current +β3E risk ;

[0089] Among them, P profile is the output of the individual risk model, H base is the static health parameter vector, B current is the current behavior feature vector, E risk is the environmental risk factor, β1, β2, and β3 are weight coefficients obtained through historical data training.

[0090] The deep learning assessment unit receives the risk model output and real-time behavior data, predicts the probability of falling and optimizes the response strategy.

[0091] As an option, the unit uses a deep reinforcement learning framework to dynamically adjust warning thresholds and response actions through a Q-learning algorithm.

[0092] In a specific implementation, the Q value update formula is:

[0093]

[0094] Among them, Q(s t ,a t ) represents the current state s t Next take action a t Q value, α represents the learning rate, r t+1 represents the immediate reward after executing the action, γ represents the discount factor, s t Indicates the current state, s t+1 Indicates the state at the next moment, a t represents the action at the current moment, a′ represents the best action among all possible actions at the next moment, In state s t+1 The maximum Q value among all actions

[0095] In some embodiments, the reward function r t+1 The design includes false positive penalty and missed negative penalty to ensure that the system strikes a balance between frequent false positives and response delays.

[0096] The personalized response strategy generation unit formulates specific intervention measures based on the assessment results.

[0097] Specifically, when the system determines that the fall risk exceeds a threshold, the unit generates a response strategy based on the following factors:

[0098] Risk level (low / medium / high);

[0099] The elderly person’s current location (e.g., bathroom, bedroom);

[0100] Whether you live alone;

[0101] Environmental risk factor (such as the slippery degree of the ground).

[0102] In one possible implementation, the response strategy includes a hierarchical alarm mechanism:

[0103] When the risk is low, users are reminded to pay attention to safety through voice on smart devices;

[0104] When the risk is medium, an early warning notification will be sent to the preset contacts;

[0105] When there is a high risk or a fall has occurred, an emergency call is automatically triggered and the door is unlocked.

[0106] Furthermore, the unit can combine indoor positioning data to push navigation information containing the optimal path to rescuers. The path planning algorithm shares computing resources with the emergency rescue path planning and automated response module.

[0107] The above technical solution ensures the precise matching of early warning strategies with individual health status through dynamic modeling, reinforcement learning and hierarchical response mechanism, while avoiding the rigid response problem in existing technologies.

[0108] The emergency rescue path planning and automated response module is connected to the intelligent early warning and personalized response module to calculate the optimal rescue path and guide rescuers after the elderly fall;

[0109] Specifically, the Emergency Rescue Path Planning and Automated Response Module is connected to the output of the Intelligent Warning and Personalized Response Module. It receives the fall event confirmation signal triggered by the module and the elderly person's real-time location information. It then uses a path planning algorithm to generate the optimal rescue path and drives the navigation device to guide rescuers to the scene quickly. This module ensures efficient and reliable rescue response through dynamic environmental perception and multi-objective optimization.

[0110] In this embodiment, the emergency rescue path planning and automated response module includes a rescue path planning unit and a rescuer navigation unit, which work in coordination with the logic control unit via a data interface.

[0111] Generally, the rescue path planning unit is used to calculate the optimal path from the rescuer's current location to the fall location based on indoor map data and real-time obstacle information.

[0112] Specifically, the unit has built-in building floor plan data, including key information such as room layout, access control location, staircase passages, etc., and interacts with environmental perception devices in real time to obtain the status of temporary obstacles (such as mobile furniture and cleaning equipment).

[0113] In one possible implementation, path planning uses the A algorithm, whose evaluation function is defined as:

[0114] f(n)=g(n)+h(n);

[0115] Among them, f(n) represents the evaluation value of node n, which represents the estimated total cost from the starting point to the target, g(n) represents the actual cost of node n from the starting point to the current node, and h(n) represents the heuristic estimated cost from node n to the target node.

[0116] As an option, when there are dynamic obstacles in the environment, the system periodically updates the map data and recalculates the path to ensure path validity.

[0117] The rescuer navigation unit receives the planned path data and converts it into visual or voice guidance information.

[0118] In some embodiments, the unit pushes navigation instructions to rescuers through a mobile terminal APP, including turn prompts, distance reminders, and obstacle warnings.

[0119] Specifically, the process of generating navigation instructions includes the following steps:

[0120] Convert the path node sequence into a natural language description (e.g., "turn left and go through the second door");

[0121] Combined with real-time positioning data from UWB or Bluetooth beacons, guidance content can be dynamically modified;

[0122] Highlight the optimal route and key landmarks on the electronic map.

[0123] Furthermore, the system can also be linked with the intelligent access control system to automatically unlock the passage doors in relevant areas during the rescue process, reducing traffic obstructions.

[0124] In an extended implementation, if multiple elderly people fall at the same time, the module uses a multi-objective optimization algorithm to allocate rescue resources.

[0125] The priority weight is calculated using the following formula:

[0126] W i =λ1Si +λ2D i +λ3H i ;

[0127] Among them, W i represents the rescue priority of the i-th fall event; S i Score severity; D i is the path length to the nearest rescuer; H i is the health risk coefficient; λ1, λ2, and λ3 are configurable weight parameters.

[0128] The emergency rescue path planning and automated response module integrates indoor map data and real-time environmental perception information, uses algorithm A to dynamically calculate the optimal path from the rescuer's current location to the fall location, and combines mobile terminal navigation instructions with intelligent access control to achieve rapid passage and precise guidance; for multiple concurrent fall incidents, the module allocates rescue resources through a priority weight model, comprehensively evaluates the severity of the incident, distance and health risk factor, and generates a multi-target collaborative response plan to ensure an efficient and orderly rescue process.

[0129] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A fall warning and rescue care system for the elderly, characterized by: include; A sensor module for collecting real-time activity data and environmental data of the elderly. The sensor module includes a pressure-sensing floor, a wearable device, an environmental sensing device, and a positioning system. The data fusion and behavior analysis module is connected to the sensor module to fuse the collected activity data and environmental data and analyze the health status and fall risk of the elderly; The intelligent early warning and personalized response module is connected to the data fusion and behavior analysis module to trigger the early warning mechanism and generate personalized response strategies based on the data analysis results; The emergency rescue path planning and automated response module is connected to the intelligent early warning and personalized response module to calculate the optimal rescue path and guide rescuers after the elderly fall.

2. The elderly fall warning and rescue care system according to claim 1 is characterized in that: The pressure-sensing floor in the sensor module is used to monitor the changes in the elderly's walking pace. The wearable device includes a smart bracelet or smart insole with a built-in accelerometer and gyroscope, which are used to collect the elderly's movement status and gait data in real time, and jointly analyze the acceleration signal output by the accelerometer and the angular velocity signal output by the gyroscope to determine a fall event.

3. The elderly fall warning and rescue care system according to claim 1 is characterized in that: The data fusion and behavior analysis module includes: Kalman filter unit, used to denoise multi-source data collected by sensors; Gait recognition unit, used to identify the current gait characteristics of the elderly; The fall risk assessment unit is used to assess the fall risk based on the identification results and health status.

4. The elderly fall warning and rescue care system according to claim 3 is characterized in that: The Kalman filter unit performs denoising on the acceleration signal and the angular velocity signal using a Kalman filter algorithm. The update formula of the Kalman filter is: in, represents the state estimate at the current time k, represents the state prediction at the current moment k, K k represents the Kalman gain, y k Indicates the measured value at the current moment, H k represents the measurement matrix, Represents the measurement residual.

5. The elderly fall warning and rescue care system according to claim 3 is characterized in that: The gait recognition unit processes the gait data through a convolutional neural network and a long short-term memory network to analyze the movement state of the elderly and determine whether there is a risk of falling. The convolution operation of the convolutional neural network layer is: h l =ReLU(W l *x l-1 +b l ); Among them, h l Represents the convolutional layer output of the first layer, W l represents the first layer of convolution kernel, x l-1 represents the input feature map of the l-1 layer, b l represents the bias term of the first layer, * represents the convolution operator, and ReLU(·) represents the activation function.

6. The elderly fall warning and rescue care system according to claim 1 is characterized in that: The intelligent early warning and personalized response module includes: Personalized health record generation unit, used to generate and dynamically adjust individualized risk models; A deep learning assessment unit to predict fall probability and optimize early warning response strategies; Personalized response strategy generation unit, used to formulate individualized intervention or rescue processes.

7. The elderly fall warning and rescue care system according to claim 6 is characterized in that: The personalized health profile generation unit generates a personalized health profile based on the health data and historical activity data of the elderly, and adjusts the warning threshold when the elderly have a high risk of falling, so that the warning sensitivity matches the health status of the elderly.

8. The elderly fall warning and rescue care system according to claim 6 is characterized in that: The deep learning evaluation unit uses a reinforcement learning algorithm to optimize the fall warning mechanism and dynamically adjusts the fall warning response threshold based on the health status and behavior patterns of the elderly. The update formula of the reinforcement learning is: Among them, Q(s t ,a t ) represents the current state s t Next take action a t Q value, α represents the learning rate, r t+1 represents the immediate reward after executing the action, γ represents the discount factor, s t Indicates the current state, s t+1 Indicates the state at the next moment, a t represents the action at the current moment, a′ represents the best action among all possible actions at the next moment, In state s t+1 The maximum Q value among all actions.

9. The elderly fall warning and rescue care system according to claim 1 is characterized in that: The emergency rescue path planning and automated response module includes: A rescue path planning unit, used to calculate the optimal path from the fall site to the rescuer; Rescue personnel navigation unit, used to provide real-time route guidance and location feedback.

10. The elderly fall warning and rescue care system according to claim 9, characterized in that: The rescue path planning unit uses the A algorithm to calculate the shortest path from the fall scene to the rescuer's location. The calculation formula of the A algorithm is: f(n)=g(n)+h(n); Among them, f(n) represents the evaluation value of node n, which represents the estimated total cost from the starting point to the target, g(n) represents the actual cost of node n from the starting point to the current node, and h(n) represents the heuristic estimated cost from node n to the target node.

Images