A patient violence detection method and system based on a smart bracelet

Abstract

The application relates to the technical field of medical safety, in particular to a patient violence behavior detection method and system based on an intelligent bracelet, which comprises a data acquisition module, a behavior analysis module, an abnormality judgment module and a prewarning execution module. The physiological and motion signals of a patient are collected in real time through multidimensional sensors, a behavior characteristic matrix is generated, a matching degree index is calculated, a violence tendency is analyzed and judged in combination with a threshold interval and time continuity, and graded prewarning and multi-channel notification are realized. The application can significantly improve the recognition accuracy of complex behavior patterns, reduce misjudgment and false positives, ensure that medical staff can respond in time, improve the safety management efficiency, simplify the medical management process, and provide a reliable solution for the whole-process automatic detection and prewarning of patient violence behavior.

Description

Technical Field

[0001] This invention belongs to the field of medical health and smart device technology, specifically a method and system for detecting violent behavior in patients based on a smart bracelet. Background Technology

[0002] Traditional patient behavior monitoring methods have weak real-time detection capabilities for violent behavior, making it difficult to quickly grasp abnormal actions and changes in patient status. This leads to delayed responses from medical staff in emergency situations and makes it difficult to accurately locate potential risk areas. With the increasing demand for patient safety management in medical settings, management models relying on manual observation are no longer sufficient to meet the requirements of efficient, accurate, and safe behavior monitoring to achieve automated identification, analysis, and early warning of patient behavior throughout the entire process, thereby improving the overall efficiency of medical management. Therefore, it is necessary to analyze a patient violence behavior detection method and system based on smart bracelets.

[0003] Existing technology, such as the invention patent application with application number CN202310797320.3, discloses a medical monitoring system based on a smart bracelet, which relates to the field of medical monitoring technology. This system can configure a smart bracelet terminal onto a corresponding patient and acquire their medical history information and habitual characteristics. This enables personalized monitoring and management for each patient. The service center uses the stored patient medical history information and habitual characteristics to perform targeted analysis and calculations. Through analysis, the system can obtain a first physiological parameter threshold Q and a first habitual coefficient xgx, which can more accurately adapt to the patient's physiological condition and habitual characteristics. When the physiological monitoring information exceeds the first physiological parameter threshold Q and triggers an alarm, the miniature camera inside the smart bracelet terminal is activated to identify the patient's facial features and associate them with the first habitual coefficient xgx. By calculating the anomaly degree Yc, the system can more accurately determine the patient's abnormal condition and send enhanced alarm information according to a preset threshold.

[0004] Existing technologies can meet the basic requirements for a patient violence detection method and system based on smart bracelets, but there are also some potential defects and challenges, specifically in the following aspects: First, existing technologies do not pay enough attention to the real-time analysis of data collected by smart bracelets, which affects the rapid judgment of violent behavior, increases the time delay in preventing potential dangers, reduces the effectiveness of timely intervention by medical staff, and is more likely to cause human oversights, thus weakening the accuracy and reliability of patient safety management.

[0005] Second, existing technologies pay less attention to the comprehensive evaluation of multi-dimensional data from smart bracelets, which affects the screening of complex behavioral patterns, makes it difficult to fully identify the patient's action characteristics, reduces the speed of identifying abnormal behaviors, and may lead to biased behavioral judgments due to insufficient data analysis, affecting the efficiency of early warning of violent tendencies, resulting in wasted time due to misjudgment or omission, and weakening the level of precision in patient behavior management. Summary of the Invention

[0006] The purpose of this invention is to provide a method and system for detecting violent behavior in patients based on a smart bracelet, which solves the problem of insufficient real-time monitoring capability of violent behavior in patients in the prior art.

[0007] To solve the above-mentioned technical problems, the present invention adopts the following technical solution: The present invention provides a patient violent behavior detection system based on a smart bracelet, including: a data acquisition module, a behavior analysis module, an anomaly judgment module, and an early warning execution module.

[0008] Data acquisition module: The smart bracelet uses multi-dimensional sensors built into it to collect the patient's physiological and motion signals in real time, obtains a raw data set, and transmits the data set to the behavior analysis module.

[0009] Behavior analysis module: Based on the received raw data set, extract key feature parameters, generate a behavior feature matrix, and combine it with a pre-stored behavior pattern database to calculate the matching index of the patient's current behavior.

[0010] Anomaly Detection Module: Based on the obtained matching degree index, it compares it with the preset violent behavior threshold range. If the matching degree index exceeds the violent behavior threshold range, it determines that the current behavior has a potential violent tendency and transmits the determination result to the early warning execution module.

[0011] Early warning execution module: Based on the received violent tendency determination result, it generates a corresponding early warning signal and sends the early warning signal to the terminal device of medical staff through the wireless communication module, while triggering the vibration reminder function on the smart bracelet.

[0012] Furthermore, the multi-dimensional sensors in the data acquisition module include an accelerometer, a gyroscope, a heart rate sensor, and a skin conductance sensor. The accelerometer collects acceleration change data along three axes; the gyroscope collects angular velocity change data; the heart rate sensor collects heart rate fluctuation data; and the skin conductance sensor collects skin conductivity change data. Each sensor is connected to the smart bracelet's main control chip via an I2C bus. The main control chip processes the received analog signals using a differential amplifier circuit and then converts them into digital signals using an analog-to-digital converter, forming the raw data set.

[0013] Furthermore, the method for generating the behavior feature matrix in the behavior analysis module is as follows: Based on the received raw data set, firstly, the accelerometer data is filtered to remove high-frequency noise and extract low-frequency motion features; secondly, the gyroscope data is integrated to obtain the angle change curve; thirdly, the heart rate sensor data is smoothed using a sliding window to extract the heart rate fluctuation amplitude; finally, the skin conductance response data is normalized to extract the conductivity change trend. These four types of feature data are arranged in a time series to form a four-dimensional feature vector matrix, i.e., the behavior feature matrix.

[0014] Furthermore, the specific calculation method of the matching degree index is as follows: Based on the generated behavioral feature matrix, several pre-stored violent behavior templates are extracted from the behavioral pattern database. The behavioral feature matrix is ​​compared element-by-element with each violent behavior template, and the Euclidean distance between the two is calculated. The average of the Euclidean distances of all templates is calculated, and then transformed into a matching degree index M through a nonlinear mapping function. The formula is: M = 1 / (1 + D), where D represents the average Euclidean distance, and the value range of M is [0,1].

[0015] Furthermore, the method for setting the threshold interval for violent behavior is as follows: based on historical data analysis, the distribution of the matching degree index of known violent behaviors is statistically analyzed, and the 95% confidence interval of the matching degree index is selected as the threshold interval for violent behavior. Specifically, the lower limit of the threshold interval for violent behavior is denoted as L, and the upper limit is denoted as U. If the matching degree index M satisfies M > U or M < L, it is determined that the current behavior has a potential tendency towards violence.

[0016] Furthermore, the warning signal generation method in the warning execution module is as follows: based on the received violent tendency determination result, a warning message containing the patient ID, timestamp, and behavioral description is first generated; secondly, the warning message is sent to the medical staff's terminal device via Bluetooth module; simultaneously, a micro vibration motor on the smart bracelet is driven by a PWM signal to generate a vibration reminder at a fixed frequency and amplitude. The vibration reminder frequency is set to 3Hz, the amplitude is set to 0.5g, and the duration is 5 seconds.

[0017] Furthermore, the method for constructing the behavioral pattern database is as follows: by collecting a large amount of behavioral data from patients in different scenarios, the data is labeled and classified, feature vectors of typical violent behaviors are extracted, and these feature vectors are stored as templates. The database supports dynamic updating; when new cases of violent behavior are added, the model can be retrained using machine learning algorithms to optimize the template library.

[0018] Furthermore, the violent tendency determination method in the abnormality determination module also includes the temporal continuity analysis of the patient's behavior. Specifically, based on the time series data of the matching degree index, the rate of change of the matching degree index at adjacent time points is calculated. If the rate of change exceeds a preset threshold, the current behavior is considered to have a sudden violent tendency, thereby further enhancing the warning priority.

[0019] Furthermore, the warning execution module also has a multi-level warning mechanism, which is divided into three warning levels according to the matching degree index: when the matching degree index is at the boundary of the violent behavior threshold range, a level one warning is triggered, and only the smart bracelet vibration reminder is activated; when the matching degree index deviates significantly from the threshold range, a level two warning is triggered, and a warning message is sent to the medical staff terminal device; when the matching degree index reaches an extreme value, a level three warning is triggered, and in addition to the above operations, the preset emergency contact number will be dialed automatically.

[0020] A second aspect of the present invention provides a method for executing the aforementioned patient violent behavior detection system based on a smart bracelet, characterized by comprising: Step 1, data acquisition: real-time acquisition of the patient's physiological and action signals using multi-dimensional sensors built into the smart bracelet to obtain a raw data set. Step 2, behavior analysis: based on the acquired raw data set, key feature parameters are extracted to generate a behavior feature matrix, and a matching degree index of the patient's current behavior is calculated in conjunction with a behavior pattern database. Step 3, anomaly determination: based on the obtained matching degree index, it is compared with a preset violent behavior threshold range to determine whether there is a violent tendency. Step 4, early warning execution: based on the violent tendency determination result, a corresponding early warning signal is generated and sent to the medical staff's terminal device, while simultaneously triggering a vibration reminder on the smart bracelet.

[0021] The beneficial effects of this invention are as follows: In the data acquisition module and behavior analysis module, by collecting patients' physiological and behavioral signals in real time from multiple dimensions, generating a behavioral feature matrix and calculating a matching degree index, the accuracy of recognizing complex behavioral patterns is significantly improved, and the risk of misjudgment due to a single data source is reduced. The anomaly detection module enhances the ability to capture sudden violent behavior by introducing a threshold range for violent behavior and time continuity analysis, reducing false alarms caused by data fluctuations. The early warning execution module ensures that medical staff can obtain early warning information in a timely manner through a hierarchical early warning mechanism and multi-channel notification methods, improving the response speed to emergencies. The overall design realizes automated detection and early warning of patient violent behavior throughout the entire process, simplifying medical management processes and improving the level of precision in safety management. Attached Figure Description

[0022] Figure 1This is a schematic diagram of the overall system architecture of the present invention, showing the connection relationship between the data acquisition module, behavior analysis module, anomaly detection module and early warning execution module.

[0023] Figure 2 The flowchart for the data acquisition module details the process by which multi-dimensional sensors acquire patients' physiological and behavioral signals and generate raw data sets.

[0024] Figure 3 The flowchart shows the method for generating the behavioral feature matrix, which describes the specific steps for extracting key feature parameters from the original dataset and forming a four-dimensional feature vector matrix.

[0025] Figure 4 The diagram shows the logic flow of the matching index calculation process, illustrating the comparison between the behavioral feature matrix and the violent behavior template, and the process of obtaining the matching index through a non-linear mapping function.

[0026] Figure 5 This diagram illustrates the hierarchical early warning mechanism of the early warning execution module, showing the operation process of triggering different levels of early warnings based on the matching degree index.

[0027] Figure 6 The flowchart illustrates the process of dynamically updating the behavior pattern database, demonstrating how to optimize the database and update templates by adding new violent behavior cases.

[0028] The attached diagram is labeled as follows: 1. Data acquisition module; 2. Behavior analysis module; 3. Anomaly detection module; 4. Early warning execution module; 5. Multi-dimensional sensor; 6. Behavior feature matrix; 7. Violent behavior template; 8. Matching index; 9. Hierarchical early warning mechanism; 10. Behavior pattern database. Detailed Implementation

[0029] This invention provides a patient violent behavior detection system and method based on a smart bracelet, the specific implementation of which is as follows. Figure 1 The diagram shows the overall system architecture. The system includes a data acquisition module 1, a behavior analysis module 2, an anomaly detection module 3, and an early warning execution module 4. These modules are connected via wireless communication protocols or internal buses to ensure efficient data transmission and automated detection and early warning throughout the entire process.

[0030] The data acquisition module 1 collects the patient's physiological and motion signals in real time using the multi-dimensional sensor 5 built into the smart bracelet. Combined with... Figure 2The working flowchart of the data acquisition module. The multi-dimensional sensor 5 includes an accelerometer, a gyroscope, a heart rate sensor, and a galvanic skin response sensor. The accelerometer is used to collect acceleration change data in three-axis directions, the gyroscope is used to collect angular velocity change data, the heart rate sensor is used to collect the heart rate fluctuation data of the patient, and the galvanic skin response sensor is used to collect the skin conductivity change data. Each sensor is connected to the main control chip of the smart bracelet through the I2C bus. After the main control chip processes the received analog signal through a differential amplification circuit, it uses an analog-to-digital converter to convert it into a digital signal, forming an original data set. This original data set is then transmitted to the behavior analysis module 2 for further processing.

[0031] Based on the received original data set, the behavior analysis module 2 extracts key feature parameters and generates a behavior feature matrix 6. Combining Figure 3 with the flowchart of the behavior feature matrix generation method, first, filter the accelerometer data to remove high-frequency noise to extract low-frequency motion features; second, perform an integration operation on the gyroscope data to obtain an angle change curve; third, perform a sliding window smoothing process on the heart rate sensor data to extract the heart rate fluctuation amplitude; finally, perform a normalization process on the galvanic skin response data to extract the conductivity change trend. The above four types of feature data are arranged in a time series to form a four-dimensional feature vector matrix, that is, the behavior feature matrix 6. The behavior feature matrix 6 is then input into the matching degree index calculation process. As Figure 4 shown, several pre-stored violent behavior templates 7 are extracted from the behavior pattern database 10, and the behavior feature matrix 6 is compared element by element with each violent behavior template 7 to calculate the Euclidean distance between the two. Calculate the average value of the Euclidean distances of all templates and convert it into a matching degree index 8 through a non-linear mapping function. The formula is M = 1 / (1 + D), where D represents the average value of the Euclidean distance, and the value range of M is [0,1]. The matching degree index 8 is then passed to the anomaly determination module 3.

[0032] The anomaly determination module 3 compares the obtained matching degree index 8 with the preset violent behavior threshold interval. The setting method of the violent behavior threshold interval is based on historical data analysis. Statistically analyze the distribution of the matching degree indexes of known violent behaviors, and select the 95% confidence interval of the matching degree index as the violent behavior threshold interval. Specifically, the lower limit of the violent behavior threshold interval is denoted as L, and the upper limit is denoted as U. If the matching degree index M satisfies M > U or M < L, it is determined that the current behavior has a potential violent tendency, and the determination result is transmitted to the warning execution module 4. In addition, the anomaly determination module 3 also includes an analysis of the time continuity of the patient's behavior. Based on the time series data of the matching degree index, calculate the change rate of the matching degree index at adjacent time points. If the change rate exceeds the preset threshold, it is considered that the current behavior has a sudden violent tendency, further enhancing the warning priority.

[0033] The early warning execution module 4 generates a corresponding early warning signal based on the received violent tendency assessment result and sends it to the medical staff's terminal device, while simultaneously triggering the vibration reminder function on the smart bracelet. Combined with... Figure 5 The diagram illustrates a tiered early warning mechanism. Early warning execution module 4 features a multi-level early warning system, categorized into three levels based on the matching degree index. When the matching degree index falls within the threshold range for violent behavior, a Level 1 warning is triggered, activating only the smart bracelet's vibration alert. When the matching degree index significantly deviates from the threshold range, a Level 2 warning is triggered, simultaneously sending an alert to the medical staff's terminal device. When the matching degree index reaches an extreme value, a Level 3 warning is triggered, automatically dialing a pre-set emergency contact number in addition to the above actions. The early warning signal is generated based on the received violent tendency assessment result. First, an alert message containing the patient ID, timestamp, and behavioral description is generated. Then, the alert message is sent to the medical staff's terminal device via Bluetooth. Simultaneously, a PWM signal drives a miniature vibration motor on the smart bracelet to generate a vibration alert at a fixed frequency and amplitude. The vibration alert frequency is set to 3Hz, the amplitude to 0.5g, and the duration to 5 seconds.

[0034] The construction method of behavioral pattern database 10 is as follows: Figure 6 As shown, by collecting a large amount of patient behavior data in different scenarios, the data is labeled and classified, and feature vectors of typical violent behaviors are extracted and stored as templates. The database supports dynamic updates; when new violent behavior cases are added, the model can be retrained using machine learning algorithms to optimize the template library. The behavior pattern database 10 interacts with the behavior analysis module 2 and the anomaly detection module 3 through an internal data interface to ensure the real-time performance and accuracy of the template library.

[0035] In practical application, the system's workflow is as follows: After the patient wears the smart bracelet, the data acquisition module 1 begins to collect the patient's physiological and behavioral signals in real time, forming a raw dataset which is then merged and transmitted to the behavior analysis module 2. The behavior analysis module 2 extracts key feature parameters to generate a behavior feature matrix 6, and calculates the matching degree index 8 of the patient's current behavior in conjunction with the behavior pattern database 10. The matching degree index 8 is transmitted to the anomaly judgment module 3. If the matching degree index 8 exceeds the violent behavior threshold range, it is determined that the current behavior has a potential violent tendency, and the judgment result is transmitted to the early warning execution module 4. The early warning execution module 4 triggers different levels of early warning operations based on the matching degree index 8, including generating early warning information, sending it to medical staff's terminal devices, and triggering the smart bracelet to vibrate. Throughout the process, the modules achieve fully automated detection and early warning of the patient's violent behavior through efficient data transmission and logical judgment.

[0036] In this embodiment, the multi-dimensional sensor 5 of the data acquisition module 1 is connected to the main control chip via a hardware interface. The main control chip converts analog signals into digital signals through a differential amplifier circuit and an analog-to-digital converter, ensuring the accuracy and real-time performance of data acquisition. The behavior analysis module 2 processes the raw data set using software algorithms to generate a behavior feature matrix 6, and compares it with the behavior pattern database 10 through a matching degree index calculation process, ensuring the accuracy of behavior recognition. The anomaly judgment module 3 enhances the ability to capture sudden violent behavior through violent behavior threshold range and time continuity analysis, reducing false alarms. The early warning execution module 4 ensures that medical staff can obtain early warning information in a timely manner through a hierarchical early warning mechanism and multi-channel notification methods, improving the response speed to emergencies. The overall design realizes automated detection and early warning of patient violent behavior throughout the entire process, simplifying medical management processes and improving the level of precision in safety management. In order to enable those skilled in the art to fully understand and implement this invention, the implementation principle of this invention is further explained below in conjunction with specific application scenarios.

[0037] In real-world medical scenarios, after a patient wears a smart bracelet, the data acquisition module 1 collects the patient's physiological and behavioral signals in real time through multi-dimensional sensors 5. Combined with... Figure 2 The data acquisition module workflow diagram shows that the accelerometer collects acceleration change data in three axes to capture the patient's vigorous limb movements; the gyroscope collects angular velocity change data to sense changes in the patient's posture; the heart rate sensor collects heart rate fluctuation data to monitor emotional excitement or tension; and the skin conductance sensor collects skin conductivity change data to reflect the patient's emotional fluctuations. Each sensor is connected to the main control chip via an I2C bus. The main control chip uses a differential amplifier circuit to process the received analog signals and then converts them into digital signals through an analog-to-digital converter, forming the raw data set. This process ensures the real-time performance and accuracy of data acquisition, providing a reliable foundation for subsequent analysis.

[0038] Subsequently, behavior analysis module 2 generates behavior feature matrix 6 based on the received raw data set. Combined with... Figure 3The flowchart of the behavioral feature matrix generation method is as follows: First, accelerometer data is low-pass filtered to remove high-frequency noise and extract low-frequency motion features, which reflect the slow movement trend of the patient's limbs. Second, gyroscope data is integrated to obtain angle change curves, used to identify the patient's rotation or tilting movements. Third, heart rate sensor data is smoothed using a sliding window to extract heart rate fluctuation amplitude, used to determine whether the patient's emotional state is abnormal. Finally, skin conductance data is normalized to extract conductivity change trends, further assisting in the judgment of emotional fluctuations. The above four types of feature data are arranged in time series to form a four-dimensional feature vector matrix, namely the behavioral feature matrix 6. This matrix is ​​transmitted to the matching degree index calculation process through an internal data interface, such as... Figure 4 As shown, several pre-stored violent behavior templates 7 are extracted from the behavior pattern database 10. The behavior feature matrix 6 is compared element-by-element with each violent behavior template 7, and the Euclidean distance between them is calculated. After averaging the Euclidean distances of all templates, they are transformed into a matching degree index 8 through a non-linear mapping function, the formula of which is M=1 / (1+D), where D represents the average Euclidean distance. The matching degree index 8 ranges from [0,1], and the closer the value is to 1, the higher the similarity between the current behavior and the violent behavior template.

[0039] Anomaly detection module 3 determines violent tendencies based on a matching index of 8. Combined with... Figure 4 The logic diagram shows that the threshold range for violent behavior is set based on historical data analysis, using a 95% confidence interval of the matching degree index distribution of known violent behaviors as a reference. If the matching degree index M exceeds the upper or lower limit of the violent behavior threshold range, it is determined that the current behavior has a potential violent tendency, and the determination result is transmitted to the early warning execution module 4. In addition, the anomaly judgment module 3 also analyzes the temporal continuity of the patient's behavior, calculating the rate of change between adjacent time points by combining the time series data of the matching degree index. If the rate of change exceeds a preset threshold, it is considered that the current behavior has a sudden violent tendency, thereby increasing the early warning priority. This mechanism effectively reduces false alarms caused by single data fluctuations.

[0040] The early warning execution module 4 triggers different levels of early warning operations based on the matching degree index 8. Combined with... Figure 5The diagram illustrates a tiered early warning mechanism. When the matching index falls within the threshold range for violent behavior, a Level 1 warning is triggered, activating only the vibration alert function of the smart bracelet. At this time, a miniature vibration motor, driven by a PWM signal, generates a 5-second vibration alert at a frequency of 3Hz and an amplitude of 0.5g, prompting the patient to adjust their behavior. When the matching index significantly deviates from the threshold range, a Level 2 warning is triggered, generating an alert message containing the patient's ID, timestamp, and behavioral description, which is then sent to the healthcare worker's terminal device via Bluetooth. When the matching index reaches an extreme value, a Level 3 warning is triggered, automatically dialing a pre-set emergency contact number in addition to the above actions, ensuring that healthcare workers can intervene immediately. This tiered early warning mechanism significantly improves the response efficiency to emergencies.

[0041] The dynamic update capability of the behavioral pattern database 10 further enhances the system's adaptability. Combined with... Figure 6 The system dynamically updates its behavioral pattern database. By collecting a large amount of patient behavioral data in different scenarios, the system labels and classifies the data, extracts feature vectors of typical violent behaviors, and stores them as templates. When new violent behavior cases are added, the system retrains the model using machine learning algorithms to optimize the template library, thereby improving the accuracy and coverage of violent behavior identification. The behavioral pattern database 10 interacts efficiently with the behavioral analysis module 2 and the anomaly detection module 3 through internal data interfaces, ensuring the real-time performance and accuracy of the template library.

[0042] The overall design achieves automated detection and early warning of patient violent behavior throughout the entire process through multi-dimensional sensors 5 in the data acquisition module 1, feature extraction and matching in the behavior analysis module 2, threshold comparison and temporal continuity analysis in the anomaly judgment module 3, and a hierarchical early warning mechanism in the early warning execution module 4. This process not only simplifies medical management procedures but also significantly improves the level of precision in safety management, ensuring that medical staff can respond to potential risks in a timely manner and protect the safety of patients and the medical environment.

Claims

1. A patient violent behavior detection system based on a smart bracelet, characterized in that, include: Data acquisition module (1): Real-time acquisition of the patient's physiological and action signals through the multi-dimensional sensor (5) built into the smart bracelet, obtaining the raw data set, and transmitting the data set to the behavior analysis module (2); Behavior analysis module (2): Based on the received raw data set, extract key feature parameters, generate a behavior feature matrix (6), and combine it with the pre-stored behavior pattern database (10) to calculate the matching degree index (8) of the patient's current behavior; Abnormal judgment module (3): Based on the obtained matching degree index (8), compare it with the preset violent behavior threshold range. If the matching degree index exceeds the violent behavior threshold range, it is determined that the current behavior has a potential violent tendency, and the judgment result is transmitted to the early warning execution module (4); Early warning execution module (4): Based on the received violent tendency judgment result, generate the corresponding early warning signal, and send the early warning signal to the medical staff terminal device through the wireless communication module, while triggering the vibration reminder function on the smart bracelet.

2. The patient violent behavior detection system based on a smart bracelet according to claim 1, characterized in that, The multi-dimensional sensor (5) includes an accelerometer, a gyroscope, a heart rate sensor, and a skin conductance sensor. The accelerometer is used to collect acceleration change data in three axes, the gyroscope is used to collect angular velocity change data, the heart rate sensor is used to collect heart rate fluctuation data, and the skin conductance sensor is used to collect skin conductivity change data. Each sensor is connected to the main control chip of the smart bracelet via an I2C bus. The main control chip processes the received analog signal through a differential amplifier circuit and then converts it into a digital signal using an analog-to-digital converter to form the original data set.

3. The patient violent behavior detection system based on a smart bracelet according to claim 1, characterized in that, The method for generating the behavior feature matrix (6) is as follows: based on the received original data set, the accelerometer data is filtered to remove high-frequency noise, low-frequency motion features are extracted, the gyroscope data is integrated to obtain the angle change curve, the heart rate sensor data is smoothed by sliding window to extract the heart rate fluctuation amplitude, the skin conductance response data is normalized to extract the conductivity change trend, and the above four types of feature data are arranged in time series to form a four-dimensional feature vector matrix, namely the behavior feature matrix (6).

4. The patient violent behavior detection system based on a smart bracelet according to claim 1, characterized in that, The matching degree index (8) is calculated as follows: Based on the generated behavioral feature matrix (6), several pre-stored violent behavior templates (7) are extracted from the behavioral pattern database (10). The behavioral feature matrix (6) is compared element by element with each violent behavior template (7), and the Euclidean distance between the two is calculated. The average of the Euclidean distances of all templates is calculated, and the matching degree index M is converted into a matching degree index M through a nonlinear mapping function. The formula is M=1 / (1+D), where D represents the average value of the Euclidean distance, and the value range of M is [0,1].

5. A patient violent behavior detection system based on a smart bracelet according to claim 1, characterized in that, The method for setting the violent behavior threshold range is as follows: Based on historical data analysis, statistically analyze the distribution of the matching degree index of known violent behaviors, and select the 95% confidence interval of the matching degree index as the violent behavior threshold range. Among them, the lower limit of the violent behavior threshold range is denoted as L, and the upper limit is denoted as U. If the matching degree index M satisfies M > U or M < L, it is determined that the current behavior has a potential violent tendency.

6. The patient violent behavior detection system based on a smart bracelet according to claim 1, characterized in that, The method for generating the warning signal is as follows: Based on the received violent tendency determination result, first generate a warning message containing the patient ID, timestamp, and behavior description. Secondly, send the warning message to the medical staff's terminal device through the Bluetooth module, and at the same time drive the micro vibration motor on the smart bracelet through the PWM signal to generate vibration reminders at a fixed frequency and amplitude. The frequency of the vibration reminder is set to 3 Hz, the amplitude is set to 0.5 g, and the duration is 5 seconds.

7. A patient violent behavior detection system based on a smart bracelet according to claim 1, characterized in that, The method for constructing the behavior pattern database (10) is as follows: By collecting a large amount of patients' behavior data in different scenarios, annotating and classifying the data, extracting the feature vectors of typical violent behaviors, and storing these feature vectors as templates. The database supports the dynamic update function. When new violent behavior cases are added, the model can be retrained through machine learning algorithms to optimize the template library.

8. A patient violent behavior detection system based on a smart bracelet according to claim 1, characterized in that, The abnormal determination module (3) further includes the analysis of the time continuity of the patient's behavior, specifically: Based on the time series data of the matching degree index, calculate the change rate of the matching degree index between adjacent time points. If the change rate exceeds the preset threshold, it is considered that the current behavior has a sudden violent tendency, and the warning priority is further enhanced.

9. A patient violence behavior detection system based on a smart bracelet according to claim 1, characterized in that, The warning execution module (4) has a multi-level warning mechanism, which is divided into three warning levels according to the level of the matching degree index: When the matching degree index is at the boundary of the violent behavior threshold range, the first-level warning is triggered, and only the vibration reminder of the smart bracelet is activated; when the matching degree index deviates significantly from the threshold range, the second-level warning is triggered, and at the same time, a warning message is sent to the medical staff's terminal device; when the matching degree index reaches an extreme value, the third-level warning is triggered, and in addition to the above operations, a preset emergency contact number will be automatically dialed.

10. A method for implementing the patient violence detection system based on a smart bracelet according to any one of claims 1 to 9, characterized in that, Including: Step 1, data collection: Real-time collect the physiological signals and motion signals of the patient through the multi-dimensional sensor (5) built in the smart bracelet to obtain the original data set; Step 2, behavior analysis: Based on the collected original data set, extract the key feature parameters, generate the behavior feature matrix (6), and calculate the matching degree index (8) of the patient's current behavior in combination with the behavior pattern database (10); Step 3, abnormal determination: Based on the obtained matching degree index (8), compare it with the preset violent behavior threshold range to determine whether there is a violent tendency; Step 4, warning execution: Based on the violent tendency determination result, generate the corresponding warning signal and send it to the medical staff's terminal device, and at the same time trigger the vibration reminder of the smart bracelet.

Images