Fall early warning method and system based on real-time physiological data feature extraction

Abstract

The invention discloses a fall early warning method and system based on real-time physiological data feature extraction, and relates to the technical field of health monitoring, and the method comprises the steps: collecting physiological data in real time through a user wearing an intelligent bracelet, and carrying out the preprocessing; performing feature extraction on the preprocessed physiological data through a sliding window method to obtain real-time features; screening the real-time features by using a dynamic risk assessment method based on a sliding window to obtain real-time features conforming to rules; constructing a tumble early warning model by using a long-short term memory (LSTM) network model, and inputting the real-time features conforming to the rule into the tumble early warning model to obtain a tumble risk prediction value; according to the tumble early warning method based on real-time physiological data feature extraction, the three-axis acceleration, the attitude angle and the heart rate data collected by the smart bracelet in real time are fully utilized, the tumble risk of a user is accurately predicted in combination with preprocessing, feature extraction, rule screening and a tumble early warning model, and noise is eliminated through the preprocessing step.

Description

Technical Field

[0001] The present invention relates to the technical field of health monitoring, and in particular to a fall warning method and system based on real-time physiological data feature extraction. Background Art

[0002] With the increasing aging of society, the health management of the elderly has gradually become a focus of research in the medical and health fields, especially falls, which are one of the most common and fatal health risks among the elderly and have attracted widespread attention. Existing fall warning technologies mostly rely on traditional monitoring methods, such as video surveillance, ground sensors and mattress pressure sensors. In recent years, with the development of wearable technology and the Internet of Things, fall warning methods based on smart devices have gradually become a new research hotspot. Smart bracelets, as a convenient, wearable device with multiple sensing functions, can provide rich information by collecting users' physiological data in real time, helping to more accurately identify fall risks and thus reduce the harm of falls to the health of the elderly.

[0003] However, although the existing fall warning methods based on wearable devices can monitor the user's activity status in real time, they still have certain deficiencies in data processing and analysis. First, most of the existing technologies use simple motion detection algorithms, and usually can only rely on acceleration sensors to judge fall events, while the monitoring of users' physiological data is relatively neglected, and there is a lack of comprehensive analysis of multi-dimensional physiological signals. Secondly, existing methods usually use fixed thresholds to judge fall events, lack a dynamic risk assessment mechanism, and cannot fully adapt to the differences in physiological characteristics and real-time status changes of different individuals. In addition, traditional models have limited processing capabilities for time series data, making it difficult to capture behavioral trends over long time spans, and are even unable to have a deep understanding of complex physiological data, resulting in insufficient prediction accuracy and reliability. Summary of the invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a fall warning method based on real-time physiological data feature extraction to solve the problems of insufficient physiological data analysis and insufficient adaptation to individual differences in the prior art.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: In a first aspect, the present invention provides a fall warning method based on real-time physiological data feature extraction, which comprises: The user wears a smart bracelet to collect physiological data in real time and perform preprocessing; The preprocessed physiological data is subjected to feature extraction through a sliding window method to obtain real-time features; Use a dynamic risk assessment method based on sliding windows to screen real-time features and obtain real-time features that meet the rules; A fall warning model is constructed using the long short-term memory network (LSTM) model. Real-time features that meet the rules are input into the fall warning model to obtain the fall risk prediction value. Based on the fall risk prediction value, the fall risk level is determined and the corresponding emergency response mechanism is triggered.

[0007] As a preferred solution of the fall warning method based on real-time physiological data feature extraction described in the present invention, wherein: the real-time collection of physiological data by the user wearing a smart bracelet, the specific steps are: The user wears a smart bracelet equipped with a three-axis accelerometer, a gyroscope and a photoplethysmogram sensor; The physiological data refers to three-axis acceleration data, attitude angle data and heart rate data; The three-axis accelerometer is used to collect three-axis acceleration data in real time; Through the gyroscope, attitude angle data is collected in real time; Heart rate data is collected in real time through the photoelectric volumetric pulse wave sensor.

[0008] As a preferred solution of the fall warning method based on real-time physiological data feature extraction of the present invention, the preprocessing comprises the following specific steps: The triaxial acceleration data, attitude angle data and heart rate data are normalized by using the minimum-maximum normalization method to obtain dimensionless triaxial acceleration data, attitude angle data and heart rate data; A low-pass filter is used to remove noise from the dimensionless three-axis acceleration data, attitude angle data and heart rate data to obtain pure three-axis acceleration data, attitude angle data and heart rate data; The linear interpolation algorithm is used to fill in the pure three-axis acceleration data, attitude angle data and heart rate data to obtain complete three-axis acceleration data, attitude angle data and heart rate data.

[0009] As a preferred solution of the fall warning method based on real-time physiological data feature extraction of the present invention, wherein: the feature extraction of the pre-processed physiological data by the sliding window method is performed to obtain the real-time feature, and the specific steps are: Set the sliding window length to , the sliding step size is set to , the data sampling frequency is set to , the total number of sampling points in each sliding window is ; The preprocessed three-axis acceleration data, attitude angle data and heart rate data are segmented using the sliding window method. The data structure expression of each window is: ; in, Indicates A sliding window, represents the index of the sliding window, The index of the time point. Indicates time point hour Axis acceleration data, Indicates time point hour Axis acceleration data, Indicates time point hour Axis acceleration data, Indicates time point The pitch angle at Indicates time point The rolling angle at Indicates time point The yaw angle at Indicates time point Heart rate data; For each window, according to the set sliding step size Move and generate a sliding window segmented data set ,in Indicates the total number of sliding windows; The real-time features are acceleration data amplitude, attitude angle data change rate and heart rate data change rate; For the Sliding Window , respectively extract the three-axis acceleration data amplitude, attitude angle data change rate and heart rate data change rate from the three-axis acceleration data, attitude angle data and heart rate data, and the expression is: ; in, Indicates Sliding Window Internal three-axis acceleration data amplitude; ; in, Indicates Sliding Window The rate of change of the internal attitude angle data, Indicates the time point The pitch angle at Indicates time point The rolling angle at Indicates time point The yaw angle at Indicates time point With time point The change in pitch angle between Indicates time point With time point The change in roll angle between Indicates time point With time point The change in yaw angle between ; in, Indicates Sliding Window The rate of change of the heart rate data, Indicates Sliding Window The average of all heart rate values ​​within the time frame.

[0010] As a preferred solution of the fall warning method based on real-time physiological data feature extraction described in the present invention, the real-time features are screened using a dynamic risk assessment method based on a sliding window to obtain real-time features that meet the rules. The specific steps are: In the Sliding Window In the calculation, a comprehensive characteristic value is constructed based on the three-axis acceleration amplitude, attitude angle change rate and heart rate change rate. The expression is: ; in, For the Sliding Window The comprehensive characteristic value of is the weight factor of the three-axis acceleration, is the weight factor of heart rate variability, is the weight factor of the attitude angle change rate, is the regulating factor of heart rate variability, represents the base of natural logarithms, is the average value of historical heart rate data; Based on the real-time features extracted from the user's historical physiological data, the mean and standard deviation of the real-time features are analyzed by quantile statistics. According to the analysis results, the risk window screening threshold is set. ; when > When , the sliding window is considered to be a high-risk window, which meets the rule conditions and retains the real-time characteristics of the window; when ≤ , the sliding window is considered to be a low-risk window and does not meet the rule conditions, and the real-time features of the window are removed.

[0011] As a preferred solution of the fall warning method based on real-time physiological data feature extraction described in the present invention, wherein: the long short-term memory network LSTM model is used to build a fall warning model, and the real-time features that meet the rules are input into the fall warning model to obtain the fall risk prediction value. The specific steps are: The long short-term memory network LSTM model is used as the core framework of fall warning; The fall warning model includes an input layer, an LSTM layer, a feature weighted processing layer, an activation function layer, and an output layer; For the high-risk sliding windows screened out, extract the triaxial acceleration data amplitude, attitude angle data change rate, and heart rate data change rate of each window, and arrange them in chronological order to form a real-time feature matrix , the expression is: ; in, Represents the total number of high-risk sliding windows; The real-time feature matrix Passed as input to the input layer of the fall warning model; The real-time feature matrix of the LSTM layer input , input the real-time feature matrix step by time Each row to the LSTM unit; The LSTM unit generates hidden states through the input gate, forget gate, and output gate mechanism. , the expression is: ; in, Represents the real-time feature matrix No. A high-risk sliding window, Indicates the index of the high-risk sliding window, with a value range of , Indicates High-risk sliding windows The hidden state of Indicates High-risk sliding windows The hidden state of Indicates High-risk sliding windows The memory unit, Represents the calculation function of the LSTM model unit of the long short-term memory network, Indicates High-risk sliding windows The amplitude of the three-axis acceleration data, Indicates High-risk sliding windows The change rate of attitude angle data, Indicates High-risk sliding windows The rate of change of heart rate data; go through The LSTM layer outputs a hidden state sequence. ; The hidden state sequence of the LSTM layer Input feature weighted processing layer, the hidden state sequence Aggregate into a fixed-dimensional feature vector ; Fixed-dimensional feature vector Encapsulates the three-axis acceleration data amplitude, attitude angle data change rate, and heart rate data change rate of the high-risk sliding window; From a fixed-dimensional feature vector In the process, the amplitude of the three-axis acceleration data, the change rate of the attitude angle data and the change rate of the heart rate data of the high-risk sliding window are extracted, and the weighted combination value is obtained through the weighted combination formula. ; The weighted combination value Input activation function layer, weighted combination value Apply the Sigmoid activation function to map the fall risk prediction value , the expression is: ; in, represents the Sigmoid activation function, Indicates at the global time point Prediction value of fall risk.

[0012] As a preferred solution of the fall warning method based on real-time physiological data feature extraction described in the present invention, wherein: based on the fall risk prediction value, the fall risk level is judged and the corresponding emergency response mechanism is triggered, the specific steps are: Set the fall risk level threshold range to , determine the fall risk level and trigger the corresponding emergency response mechanism; in, Indicates the lower threshold of the fall risk level, represents the upper threshold of the fall risk level; when < When , it indicates a low risk level, and normal monitoring and data recording will continue; when ≤ ≤ When it is on, it indicates a medium risk level, and a reminder notification will be sent through the smart bracelet to remind the user to pay attention to the surrounding environment and take preventive measures such as rest and adjust posture; when > , indicating a high risk level, immediately issues a fall warning to the user and notifies emergency contacts.

[0013] In a second aspect, the present invention provides a fall warning system based on real-time physiological data feature extraction, including a data acquisition module, a feature extraction module, a dynamic screening module, a risk prediction module and a response mechanism module; The data acquisition module is used to collect the user's physiological data in real time and perform normalization, denoising and interpolation processing on the collected data; The feature extraction module is used to perform segmentation processing on the preprocessed physiological data using a sliding window method and extract real-time features; The dynamic screening module is used to screen the real-time features based on the dynamic risk assessment method of the sliding window, select the features that meet the conditions, and classify the risk windows; The risk prediction module is used to construct a fall warning model using a long short-term memory network (LSTM) model, process the screened high-risk sliding windows, and accurately predict the risk of falling; The response mechanism module is used to set a threshold range according to the predicted fall risk value, determine the risk level of the fall, and trigger a corresponding response mechanism.

[0014] In a third aspect, the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, any step of the fall warning method based on real-time physiological data feature extraction as described in the first aspect of the present invention is implemented.

[0015] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, any step of the fall warning method based on real-time physiological data feature extraction as described in the first aspect of the present invention is implemented.

[0016] The beneficial effects of the present invention are as follows: the present invention uses a fall warning method based on real-time physiological data feature extraction, fully utilizes the three-axis acceleration, posture angle and heart rate data collected in real time by the smart bracelet, and combines preprocessing, feature extraction, rule screening and fall warning model to achieve accurate prediction of the user's fall risk. Through the preprocessing step, noise is eliminated and the integrity and accuracy of the data are guaranteed; the feature extraction process accurately captures the user's movement changes through a sliding window method to help identify abnormal behaviors before falling; a dynamic risk assessment method is used to screen high-risk real-time features to improve the accuracy of risk assessment; the fall warning model further combines time series information to accurately predict the probability of falling and trigger a corresponding emergency response mechanism. The implementation of this series of steps not only improves the timeliness and accuracy of the warning, but also can judge the risk level in real time and take emergency response measures to minimize the harm caused by fall accidents, especially providing important safety guarantees for the elderly. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0018] Figure 1 This is a flow chart of the fall warning method based on real-time physiological data feature extraction in Example 1.

[0019] Figure 2 Schematic diagram of a fall warning system based on real-time physiological data feature extraction in Example 1. DETAILED DESCRIPTION

[0020] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0021] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0022] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0023] Example 1, reference Figure 1 and Figure 2 , which is the first embodiment of the present invention, provides a fall warning method based on real-time physiological data feature extraction, comprising the following steps: S1. Real-time collection of physiological data by a user wearing a smart bracelet and preprocessing; the user wearing a smart bracelet is equipped with a three-axis accelerometer, a gyroscope and a photoelectric volumetric pulse wave sensor; The physiological data refers to three-axis acceleration data, attitude angle data and heart rate data; The three-axis accelerometer is used to collect three-axis acceleration data in real time; Through the gyroscope, attitude angle data is collected in real time; Collect heart rate data in real time through photoplethysmography sensor; By wearing a smart bracelet, the three-axis acceleration data, posture angle data and heart rate data are collected in real time, ensuring the continuity and accuracy of the data source, providing reliable data support for subsequent fall risk analysis; The triaxial acceleration data, attitude angle data and heart rate data are normalized by using the minimum-maximum normalization method to obtain dimensionless triaxial acceleration data, attitude angle data and heart rate data; Normalization solves the problem of inconsistent dimensions of different data and improves the adaptability to multimodal data; A low-pass filter is used to remove noise from the dimensionless three-axis acceleration data, attitude angle data and heart rate data to obtain pure three-axis acceleration data, attitude angle data and heart rate data; The application of low-pass filter eliminates the interference of environmental noise and equipment jitter on the original data, and enhances the robustness to harsh acquisition conditions; The linear interpolation algorithm is used to fill in the pure three-axis acceleration data, attitude angle data and heart rate data to obtain complete three-axis acceleration data, attitude angle data and heart rate data; The linear interpolation algorithm ensures the continuity of time series data and avoids misjudgment due to temporary failure of acquisition equipment; Through data preprocessing, the user's real status can be captured more accurately and the interference of abnormal data on the analysis results can be reduced.

[0024] S2, extracting features from the preprocessed physiological data using a sliding window method to obtain real-time features; Set the sliding window length to , the sliding step size is set to , the data sampling frequency is set to , the total number of sampling points in each sliding window is ; The present invention sets the sliding window length to seconds, the sliding step size is set to seconds, the data sampling frequency is Hz, the total number of sampling points in each sliding window × =150; The pre-processed three-axis acceleration data, attitude angle data and heart rate data are processed in sections using the sliding window method. The data structure expression of each window is: ; in, Indicates A sliding window, represents the index of the sliding window, The index of the time point. Indicates time point hour Axis acceleration data, Indicates time point hour Axis acceleration data, Indicates time point hour Axis acceleration data, Indicates time point The pitch angle at Indicates time point The rolling angle at Indicates time point The yaw angle at Indicates time point Heart rate data; For each window, according to the set sliding step size Move and generate a sliding window segmented data set ,in Indicates the total number of sliding windows; Real-time features refer to the acceleration data amplitude, attitude angle data change rate, and heart rate data change rate; For the Sliding Window , respectively extract the three-axis acceleration data amplitude, attitude angle data change rate and heart rate data change rate from the three-axis acceleration data, attitude angle data and heart rate data, and the expression is: ; in, Indicates Sliding Window The internal three-axis acceleration data amplitude is used to measure the average value of the motion intensity within the sliding window; ; in, Indicates Sliding Window The rate of change of the internal attitude angle data, Indicates the time point The pitch angle at Indicates time point The rolling angle at Indicates time point The yaw angle at Indicates time point With time point The change in pitch angle between Indicates time point With time point The change in roll angle between Indicates time point With time point The change in yaw angle between ; in, Indicates Sliding Window The rate of change of the heart rate data is used to measure the amplitude of the heart rate fluctuation in the sliding window. Indicates Sliding Window The average of all heart rate values ​​in the The sliding window method extracts dynamic features with time series according to the time window, accurately captures the instantaneous state of human movement and physiological changes, enhances response speed and accuracy, and provides more reliable data support for subsequent risk screening and prediction.

[0025] S3. Use a dynamic risk assessment method based on a sliding window to screen the real-time features and obtain real-time features that meet the rules; In the Sliding Window Based on the three-axis acceleration amplitude, attitude angle change rate and heart rate change rate, the real-time feature fusion formula is used to calculate the comprehensive feature value, which is expressed as follows: ; in, For the Sliding Window The comprehensive characteristic value of is used to measure the possibility of falling risk within the window. is the weight factor of the three-axis acceleration, is the weight factor of heart rate variability, is the weight factor of the attitude angle change rate, is the regulating factor of heart rate variability, represents the base of natural logarithms, is the average value of historical heart rate data; Based on the real-time features extracted from the user's historical physiological data, the mean and standard deviation of the real-time features are analyzed by quantile statistics. According to the analysis results, the risk window screening threshold is set. , The present invention sets a threshold =1.0; when > When , the sliding window is considered to be a high-risk window, which meets the rule conditions and retains the real-time characteristics of the window; when ≤ When , the sliding window is considered to be a low-risk window and does not meet the rule conditions, and the real-time features of the window are removed; Combined with the user's historical physiological data, by calculating the comprehensive feature value and setting the screening threshold, the low-risk sliding window is eliminated and only the high-risk features are retained, which reduces the interference of irrelevant data on the early warning model and significantly improves the computing efficiency. At the same time, it meets the user's personalized needs and reduces the false alarm rate.

[0026] S4. Use the long short-term memory network LSTM model to build a fall warning model, input the real-time features that meet the rules into the fall warning model, and obtain the fall risk prediction value; The long short-term memory network LSTM model is used as the core framework of fall warning; The fall warning model includes an input layer, an LSTM layer, a feature weighted processing layer, an activation function layer, and an output layer; For the high-risk sliding windows screened out, extract the triaxial acceleration data amplitude, attitude angle data change rate, and heart rate data change rate of each window, and arrange them in chronological order to form a real-time feature matrix , the expression is: ; in, Represents the total number of high-risk sliding windows; The real-time feature matrix Passed as input to the input layer of the fall warning model; The real-time feature matrix of the LSTM layer input , input the real-time feature matrix step by time Each row to the LSTM unit; The LSTM unit generates a hidden state through the input gate, forget gate, and output gate mechanism, and the expression is: ; in, Represents the real-time feature matrix No. A high-risk sliding window, Indicates the index of the high-risk sliding window, with a value range of , Indicates High-risk sliding windows The hidden state of Indicates High-risk sliding windows The hidden state of Indicates High-risk sliding windows The memory unit, Represents the calculation function of the LSTM model unit of the long short-term memory network, Indicates High-risk sliding windows The amplitude of the three-axis acceleration data, Indicates High-risk sliding windows The change rate of attitude angle data, Indicates High-risk sliding windows The rate of change of heart rate data; go through The LSTM layer outputs a hidden state sequence. , the expression is: ; The hidden state sequence of the LSTM layer Input feature weighted processing layer, the hidden state sequence Aggregate into a fixed-dimensional feature vector , the expression is: ; in, represents the weight matrix, represents the bias vector; Fixed-dimensional feature vector Encapsulates the three-axis acceleration data amplitude, attitude angle data change rate, and heart rate data change rate of the high-risk sliding window; From a fixed-dimensional feature vector In the process, the amplitude of the three-axis acceleration data, the change rate of the attitude angle data and the change rate of the heart rate data of the high-risk sliding window are extracted, and the weighted combination value is obtained through the weighted combination formula. , the expression is: ; The weighted combination value Input activation function layer, weighted combination value Apply the Sigmoid activation function to map it into a fall risk prediction value , the expression is: ; in, represents the Sigmoid activation function, Indicates at the global time point The predicted value of fall risk; The fall warning model is constructed through the long short-term memory network (LSTM) model, which can capture the long-term and short-term dynamic changes of physiological data in the time series and accurately analyze the fall risk trend, thereby effectively improving the accuracy and timeliness of risk prediction and providing a reliable basis for taking preventive measures in a timely manner.

[0027] S5. Based on the fall risk prediction value, determine the fall risk level and trigger the corresponding emergency response mechanism; Set the fall risk level threshold range to , judge the fall risk level and trigger the corresponding emergency response mechanism. The present invention sets the fall risk level threshold range based on the fall risk prediction value. ; in, Indicates the lower threshold of the fall risk level, represents the upper threshold of the fall risk level; when < When , it indicates a low risk level, and normal monitoring and data recording will continue; when ≤ ≤ When it is on, it indicates a medium risk level, and a reminder notification will be sent through the smart bracelet to remind the user to pay attention to the surrounding environment and take preventive measures such as rest and adjust posture; when > When it is on, it indicates a high risk level, and the user is immediately warned of a fall and the emergency contact is notified; The risk level classification can take different degrees of response to different risk situations, improving practicality; Through the automated alarm mechanism, relevant parties can be notified quickly after high-risk situations or falls, shortening the rescue response time, which is of great significance to high-risk patient groups.

[0028] This embodiment also provides a fall warning system based on real-time physiological data feature extraction, including: a data acquisition module, a feature extraction module, a dynamic screening module, a risk prediction module and a response mechanism module; The data acquisition module is used to collect the user's physiological data in real time and perform normalization, denoising and interpolation processing on the collected data; A feature extraction module is used to segment the preprocessed physiological data using a sliding window method and extract real-time features; Dynamic screening module, which is used by the rule model to filter out qualified features and classify risk windows; The risk prediction module uses the long short-term memory network (LSTM) model to build a fall warning model, processes the high-risk sliding windows screened out, and accurately predicts the risk of falling; The response mechanism module is used to set the threshold range according to the predicted fall risk value, determine the risk level of fall, and trigger the corresponding response mechanism.

[0029] This embodiment also provides a computer device, which is suitable for the case of a fall warning method based on real-time physiological data feature extraction, including: a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the fall warning method based on real-time physiological data feature extraction as proposed in the above embodiment.

[0030] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covered on the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0031] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the fall warning method based on real-time physiological data feature extraction as proposed in the above embodiment is implemented; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memory, referred to as SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, referred to as EEPROM), erasable programmable read-only memory (Erasable Programmable Read Only Memory, referred to as EPROM), programmable read-only memory (Programmable Red-Only Memory, referred to as PROM), read-only memory (Read-Only Memory, referred to as ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0032] In summary, the present invention achieves accurate prediction of the user's fall risk through: a fall warning method based on real-time physiological data feature extraction, making full use of the three-axis acceleration, posture angle and heart rate data collected in real time by the smart bracelet, combining preprocessing, feature extraction, rule screening and fall warning model. Through the preprocessing step, noise is eliminated and the integrity and accuracy of the data are guaranteed; the feature extraction process accurately captures the user's movement changes through the sliding window method to help identify abnormal behaviors before falling; a dynamic risk assessment method is used to screen high-risk real-time features to improve the accuracy of risk assessment; the fall warning model further combines time series information to accurately predict the probability of falling and trigger the corresponding emergency response mechanism. The implementation of this series of steps not only improves the timeliness and accuracy of the warning, but also can judge the risk level in real time and take emergency response measures to minimize the harm caused by fall accidents, especially providing important safety guarantees for the elderly.

[0033] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A fall warning method based on real-time physiological data feature extraction, characterized in that: include, The user wears a smart bracelet to collect physiological data in real time and perform preprocessing; The preprocessed physiological data is subjected to feature extraction through a sliding window method to obtain real-time features; Use a dynamic risk assessment method based on sliding windows to screen real-time features and obtain real-time features that meet the rules; A fall warning model is constructed using the long short-term memory network (LSTM) model. Real-time features that meet the rules are input into the fall warning model to obtain the fall risk prediction value. Based on the fall risk prediction value, the fall risk level is determined and the corresponding emergency response mechanism is triggered.

2. The fall warning method based on real-time physiological data feature extraction according to claim 1, characterized in that: The specific steps of collecting physiological data in real time by having the user wear a smart bracelet are as follows: The user wears a smart bracelet equipped with a three-axis accelerometer, a gyroscope and a photoplethysmogram sensor; The physiological data refers to three-axis acceleration data, attitude angle data and heart rate data; The three-axis accelerometer is used to collect three-axis acceleration data in real time; Through the gyroscope, attitude angle data is collected in real time; Heart rate data is collected in real time through the photoelectric volumetric pulse wave sensor.

3. The fall warning method based on real-time physiological data feature extraction as claimed in claim 2, characterized in that: The pre-processing comprises the following specific steps: The triaxial acceleration data, attitude angle data and heart rate data are normalized by using the minimum-maximum normalization method to obtain dimensionless triaxial acceleration data, attitude angle data and heart rate data; A low-pass filter is used to remove noise from the dimensionless three-axis acceleration data, attitude angle data and heart rate data to obtain pure three-axis acceleration data, attitude angle data and heart rate data; The linear interpolation algorithm is used to fill in the pure three-axis acceleration data, attitude angle data and heart rate data to obtain complete three-axis acceleration data, attitude angle data and heart rate data.

4. The fall warning method based on real-time physiological data feature extraction as claimed in claim 3, characterized in that: The sliding window method is used to extract features from the preprocessed physiological data to obtain real-time features. The specific steps are as follows: Set the sliding window length to , the sliding step size is set to , the data sampling frequency is set to , the total number of sampling points in each sliding window is ; The preprocessed three-axis acceleration data, attitude angle data and heart rate data are segmented using the sliding window method. The data structure expression of each window is: ； in, Indicates A sliding window, represents the index of the sliding window, The index of the time point. Indicates time point hour Axis acceleration data, Indicates time point hour Axis acceleration data, Indicates time point hour Axis acceleration data, Indicates time point The pitch angle at Indicates time point The rolling angle at Indicates time point The yaw angle at Indicates time point Heart rate data; For each window, according to the set sliding step size Move and generate a sliding window segmented data set ,in Indicates the total number of sliding windows; The real-time features are acceleration data amplitude, attitude angle data change rate and heart rate data change rate; For the Sliding Window , respectively extract the three-axis acceleration data amplitude, attitude angle data change rate and heart rate data change rate from the three-axis acceleration data, attitude angle data and heart rate data, and the expression is: ； in, Indicates Sliding Window Internal three-axis acceleration data amplitude; ； in, Indicates Sliding Window The characteristic value of the rate of change of the internal attitude angle data, Indicates the time point The pitch angle at Indicates time point The rolling angle at Indicates time point The yaw angle at Indicates time point With time point The change in pitch angle between Indicates time point With time point The change in roll angle between Indicates time point With time point The change in yaw angle between ； in, Indicates Sliding Window The rate of change of the heart rate data, Indicates Sliding Window The average of all heart rate values ​​within the time frame.

5. The fall warning method based on real-time physiological data feature extraction as claimed in claim 4, characterized in that: The dynamic risk assessment method based on sliding windows is used to screen the real-time features to obtain the real-time features that meet the rules. The specific steps are as follows: In the Sliding Window Based on the three-axis acceleration amplitude, attitude angle change rate and heart rate change rate, the real-time feature fusion formula is used to calculate the comprehensive feature value, which is expressed as follows: ； in, For the Sliding Window The comprehensive characteristic value of is the weight factor of the three-axis acceleration, is the weight factor of heart rate variability, is the weight factor of the attitude angle change rate, is the regulating factor of heart rate variability, represents the base of natural logarithms, is the average value of the change rate of historical heart rate data; Based on the real-time features extracted from the user's historical physiological data, the mean and standard deviation of the real-time features are analyzed by quantile statistics. According to the analysis results, the risk window screening threshold is set. ; when > When , the sliding window is considered to be a high-risk window, which meets the rule conditions and retains the real-time characteristics of the window; when ≤ , the sliding window is considered to be a low-risk window and does not meet the rule conditions, and the real-time features of the window are removed.

6. The fall warning method based on real-time physiological data feature extraction according to claim 5, characterized in that: The long short-term memory network LSTM model is used to build a fall warning model, and the real-time features that meet the rules are input into the fall warning model to obtain the fall risk prediction value. The specific steps are as follows: The long short-term memory network LSTM model is used as the core framework of fall warning; The fall warning model includes an input layer, an LSTM layer, a feature weighted processing layer, an activation function layer, and an output layer; For the high-risk sliding windows screened out, extract the triaxial acceleration data amplitude, attitude angle data change rate, and heart rate data change rate of each window, and arrange them in chronological order to form a real-time feature matrix , the expression is: ； in, Represents the total number of high-risk sliding windows; The real-time feature matrix Passed as input to the input layer of the fall warning model; The real-time feature matrix of the LSTM layer input , input the real-time feature matrix step by time Each row to the LSTM unit; The LSTM unit generates hidden states through the input gate, forget gate, and output gate mechanism. , the expression is: ； in, Represents the real-time feature matrix No. A high-risk sliding window, Indicates the index of the high-risk sliding window, with a value range of , Indicates High-risk sliding windows The hidden state of Indicates High-risk sliding windows The hidden state of Indicates High-risk sliding windows The memory unit, Represents the calculation function of the LSTM model unit of the long short-term memory network, Indicates High-risk sliding windows The amplitude of the three-axis acceleration data, Indicates High-risk sliding windows The change rate of attitude angle data, Indicates High-risk sliding windows The rate of change of heart rate data; go through The LSTM layer outputs a hidden state sequence. ; The hidden state sequence of the LSTM layer Input feature weighted processing layer, the hidden state sequence Aggregate into a fixed-dimensional feature vector ; Fixed-dimensional feature vector Encapsulates the three-axis acceleration data amplitude, attitude angle data change rate, and heart rate data change rate of the high-risk sliding window; From a fixed-dimensional feature vector The three-axis acceleration data amplitude, attitude angle data change rate and heart rate data change rate of the high-risk sliding window are extracted, and the weighted combination value is obtained through the weighted combination formula ; The weighted combination value Input activation function layer, weighted combination value Apply the Sigmoid activation function and map it to the fall risk prediction value , the expression is: ； in, represents the Sigmoid activation function, Indicates at the global time point Prediction value of fall risk.

7. The fall warning method based on real-time physiological data feature extraction according to claim 6, characterized in that: The specific steps of judging the fall risk level based on the fall risk prediction value and triggering the corresponding emergency response mechanism are as follows: Set the fall risk level threshold range to , determine the fall risk level and trigger the corresponding emergency response mechanism; in, Indicates the lower threshold of the fall risk level, represents the upper threshold of the fall risk level; when < When , it indicates a low risk level, and normal monitoring and data recording will continue; when ≤ ≤ When it is on, it indicates a medium risk level, and a reminder notification will be sent through the smart bracelet to remind the user to pay attention to the surrounding environment and take preventive measures such as rest and adjust posture; when > , indicating a high risk level, immediately issues a fall warning to the user and notifies emergency contacts.

8. A fall warning system based on real-time physiological data feature extraction, based on the fall warning method based on real-time physiological data feature extraction according to any one of claims 1 to 7, characterized in that: Including data collection module, feature extraction module, dynamic screening module, risk prediction module and response mechanism module; The data acquisition module is used to collect the user's physiological data in real time and perform normalization, denoising and interpolation processing on the collected data; The feature extraction module is used to perform segmentation processing on the preprocessed physiological data using a sliding window method and extract real-time features; The dynamic screening module is used to screen the real-time features based on the dynamic risk assessment method of the sliding window, select the features that meet the conditions, and classify the risk windows; The risk prediction module is used to construct a fall warning model using a long short-term memory network (LSTM) model, process the screened high-risk sliding windows, and accurately predict the risk of falling; The response mechanism module is used to set a threshold range according to the predicted fall risk value, determine the risk level of the fall, and trigger a corresponding response mechanism.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the fall warning method based on real-time physiological data feature extraction according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the fall warning method based on real-time physiological data feature extraction according to any one of claims 1 to 7 are implemented.

Images