Abnormal behavior early warning method and system based on wearable device

Abstract

The application discloses an abnormal behavior early warning method and system based on a wearable device, and comprises the following steps: wearing a wearable monitoring device; collecting and recording daily activity data; establishing a daily activity level model according to the daily activity data; calculating an individual activity level baseline by using the daily activity level model; comparing the collected real-time activity data with the individual activity level baseline, and judging whether the activity data is abnormal; and if abnormal activity data appears, sending an early warning signal to a manager. The application collects motion acceleration data in real time, recognizes and analyzes human behaviors, constructs a machine learning model, identifies abnormal real-time behavior data which is significantly higher or lower than a normal activity level range, discovers risk behaviors in time, sends an early warning alarm, and prevents serious consequences from occurring.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of monitoring, in particular to an abnormal behavior early warning method and system based on wearable devices. BACKGROUND

[0002] Strengthening the construction of security alert facilities is an important prerequisite for maintaining operation and ensuring safety supervision.

[0003] The current degree of digitization and intelligentization is low, and intelligent and integrated devices and systems need to be introduced to achieve rapid identification and early warning of dangerous behavior. In the past, the identification and control of dangerous behavior were through monitoring cameras and relied on manual judgment based on experience, and there was no method that could automatically identify and warn dangerous behavior. SUMMARY

[0004] In order to solve the above technical problems, the present application provides an abnormal behavior early warning method and system based on wearable devices, which monitors and analyzes abnormal behavior patterns in real time, identifies abnormal behavior in time and sends warning alarms to the responsible person, so that appropriate measures can be taken in time to prevent serious consequences.

[0005] The technical solutions adopted are as follows:

[0006] An abnormal behavior early warning method based on wearable devices, wearing a wearable monitoring device; collecting and recording daily activity data; establishing a daily activity level model according to the daily activity data; calculating the daily activity level baseline using the daily activity level model; comparing the collected real-time activity data with the daily activity level baseline, and making a judgment on whether the activity data is abnormal; if abnormal activity data appears, an early warning signal is sent to the management personnel.

[0007] Further, the abnormal activity data includes activity level increase data and activity level decrease data that significantly exceed or are lower than the normal activity baseline level.

[0008] Preferably, acceleration data collected in real time in the wearable monitoring device is extracted, and acceleration mean, acceleration jerk rate and acceleration entropy are extracted; the daily activity level model is a multiple linear regression model constructed based on the acceleration mean, acceleration jerk rate and acceleration entropy extracted from the data collected in the previous three days, and its expression formula is as follows:

[0009] AL = β0 + β1 * meanACC + β2 jerkACC + β3 * entropyACC

[0010] Wherein: AL: daily activity level model;

[0011] meanACC: 1min acceleration average value;

[0012] β0 is a constant term in the multiple linear regression model;

[0013] β1, β2, β3 are regression coefficients of acceleration mean value, acceleration jerk and acceleration entropy characteristics;

[0014] jerkACC: acceleration jerk, that is, using linear regression, the average value of the first derivative of acceleration in 1 min is obtained;

[0015] entropyACC: entropy of acceleration value in 1 min;

[0016]

[0017] In the formula, N is the number of acceleration value samples in 1 min, and p i is the occurrence probability of the event of acceleration value = i in 1 min, that is, the number of points with value i ÷ N.

[0018] Further, when it is judged that the activity data of the day has no abnormal behavior, the daily activity data formed on the day is continuously superimposed on the normal activity data formed in the previous three days to form an updated daily activity level model for judging abnormal behavior.

[0019] Further, the activity level value of the previous three days is calculated using the constructed daily activity level model, and the activity level baseline of the individual is obtained by averaging the activity level values of the previous three days; the real-time collected data is compared with the obtained activity level baseline of the individual, and the absolute value of the difference between the real-time activity level and the daily activity level baseline is calculated; whether abnormal behavior occurs is judged according to the set threshold value.

[0020] Further, the set threshold value is 1.5std, if the absolute value of the difference > 1.5std: abnormal activity level is prompted, and if the absolute value of the difference ≦ 1.5std, the activity level is normal.

[0021] On the other hand, the present application also provides an abnormal behavior early warning system based on a wearable device, comprising a wearable monitoring device, an early warning device, a computer, and a data recording module, a data processing module, a model establishing module, a calculation module and an abnormal behavior discrimination module embedded in the computer;

[0022] The wearable monitoring device is worn on the body to collect motion signals;

[0023] The data recording module is used to receive the motion signals collected by the wearable monitoring device;

[0024] The data processing module is used for pre-processing the motion signal recorded by the data recording module, and extracting acceleration mean value, acceleration mutation rate and acceleration entropy feature data.

[0025] The model establishing module is used for modeling the extracted feature data, so as to obtain a daily activity level model.

[0026] The calculating module is used for calculating the individual activity level baseline by using the established daily activity level model.

[0027] The abnormal behavior judging module is used for comparing the collected real-time motion data with the individual activity level baseline, and judging whether there is abnormal behavior.

[0028] The early warning device receives the abnormal behavior signal judged by the abnormal behavior judging module, and sends an early warning signal to the manager.

[0029] Further, when the abnormal behavior judging module judges that there is no abnormal behavior, the collected normal activity data are transmitted to the model establishing module, so as to iteratively update the daily activity model.

[0030] The wearable monitoring device is a smart bracelet with a three-axis acceleration sensor.

[0031] The technical scheme of the present application has the following advantages:

[0032] A. The present application uses a wearable device to collect motion acceleration data in real time, recognizes and analyzes human behaviors, identifies abnormal real-time behavior data that significantly exceeds or is lower than the normal activity level range, extracts relevant features from three-axis acceleration sensor data by analyzing the motion mode of the human body, constructs a machine learning model, realizes the prediction and judgment of the current specific behavior, quickly and timely discovers the existing abnormal behavior, and sends an early warning alarm to the responsible person, so that the person can take corresponding measures in time to prevent serious consequences.

[0033] B. The daily activity level model used in the present application is established by using the normal activity data of the previous three days through a machine learning method, and the previous daily activity level model is further iterated based on the subsequently collected normal activity data, so that the established daily activity level model is always kept up-to-date, and as the data volume increases, the model is more accurate in fitting the daily activity level, so that the abnormal behavior can be more accurately judged; the present application can be used as an intelligent auxiliary tool for activity state, reduces the work load, and strengthens the management and prevention of problems. BRIEF DESCRIPTION OF DRAWINGS

[0034] In order to more clearly illustrate the specific embodiments of the present application, the following will briefly introduce the drawings needed to be used in the specific embodiments. Obviously, the drawings described in the following description are some embodiments of the present application, and other drawings can be obtained by those skilled in the art without any creative effort on the basis of these drawings.

[0035] Figure 1 is a flow chart of the abnormal behavior early warning method provided by the present application;

[0036] Figure 2 is a structural block diagram of the early warning system provided by the present application;

[0037] Figure 3 is a flow chart of the daily activity level model established by the present application. DETAILED DESCRIPTION

[0038] The technical solutions of the present application will be described in detail below with reference to the drawings. Obviously, the described embodiments are some embodiments of the present application, not all embodiments. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without any creative effort are within the scope of protection of the present application.

[0039] As shown in Figure 1 , the present application provides an abnormal behavior early warning method based on a wearable device, and the specific method is as follows:

[0040]

S001

[0041] A three-axis acceleration sensor is used, and the window length of the sliding window is preferably 60s, and the step length is preferably 5s.

[0042]

S002

[0043] The motion acceleration data collected by the wearable monitoring device is pre-processed. First, data calibration is performed, then Bi-Cubic-Interpolation is preferably used for data interpolation, and Butterworth filter is preferably used for band-pass filtering with a cutoff frequency of 0.2Hz and 15Hz.

[0044]

S003

[0045] ​The following features are extracted from the data per minute: meanACC, jerkACC and entropyACC. The daily activity level model is built by extracting the above three features from the normal activity data of the previous three days. The expression of the individual activity level model is:

[0046] AL = β0 + β1*meanACC + β2jerkACC + β3*entropyACC

[0047] wherein: AL: daily activity level model;

[0048] meanACC: mean value of acceleration in 1 minute;

[0049] β0: constant term in the multiple linear regression model;

[0050] β1, β2, β3: regression coefficients of each feature;

[0051] jerkACC: jerk, i.e. the average value of the first derivative of acceleration in 1 minute obtained by linear regression;

[0052] entropyACC: entropy of acceleration values in 1 minute;

[0053]

[0054] In the formula, N is the number of acceleration values in 1 minute, and p i is the occurrence probability of the event i in 1 minute, i.e. the number of points with value i ÷ N.

[0055]

S004

[0056] The average value of the daily activity level is calculated by the individual daily activity level model built above, and the individual activity level baseline is obtained by collecting the activity data of the previous three days, which is not limited to the previous three days. The data of the previous four days or more days can also be collected.

[0057] When it is determined that the activity data of the day has no abnormal behavior, the daily activity data formed on the day is preferably continued to be stacked on the normal activity data formed in the previous three days to form an updated activity level model for judging abnormal behavior, and the activity level baseline value is calculated according to the updated activity level model.

[0058]

S005

[0059] The abnormal activity data here includes both activity level increase and activity level decrease. The activity level increase includes frequent walking, frequent squatting, etc. The activity level decrease is as follows: for example, the lying time becomes longer, the time is kept in one posture (such as sitting, lying, etc.).

[0060] The individual risk behavior is determined as follows:

[0061] The individual real-time activity level calculated from the collected real-time data is compared with the activity baseline level, the absolute value of the difference (|real-time activity level-baseline activity level|) is calculated, and the threshold value of the absolute value of the difference is set to 1.5std. According to the threshold value and the absolute value of the difference, it is determined whether abnormal behavior occurs:

[0062] When the absolute value of the difference is >1.5std: abnormal activity level is prompted;

[0063] When the absolute value of the difference is ≦1.5std: the activity level is normal.

[0064] For the above various abnormal behaviors, timely warning will be given to inform the personnel to avoid serious situations. The setting of the threshold value is not limited to the above given value, and it can also be adjusted according to the actual situation.

[0065] In addition, for example, Figure 2As shown, the application also provides a wearable device-based abnormal behavior early warning system, comprising a wearable monitoring device, an early warning device, a computer, and a data recording module, a data processing module, a model establishing module, a calculation module and an abnormal behavior discrimination module embedded in the computer; the wearable monitoring device is worn on the body to collect motion signals in real time; the data recording module is used to receive the motion signals collected by the wearable monitoring device; the data processing module is used to preprocess the motion signals recorded by the data recording module and extract feature data; the model establishing module is used to model the extracted feature data to obtain a daily activity level model; the calculation module calculates the individual activity level baseline by using the established daily activity level model, and the abnormal behavior discrimination module is used to compare the collected real-time motion data with the individual activity level baseline to determine whether there is abnormal behavior; the early warning device receives the abnormal behavior signal determined by the abnormal behavior discrimination module and sends an early warning signal to the manager. When the abnormal behavior discrimination module determines that there is no abnormal behavior, the collected normal activity data is transmitted to the model establishing module for iterative updating of the daily activity model to further optimize the model, and the individual activity level baseline is calculated by using the updated model each time to make abnormal behavior comparison and discrimination.

[0066] With the development of wearable monitoring devices, the hardware devices used for behavior recognition data collection are becoming smaller and smaller in size and stronger and stronger in endurance, which can meet the needs of long-time continuous data collection in daily life. The wearable monitoring device used in the application adopts a self-developed device of Zhongke Xinian.

[0067] The unmentioned parts of the application are applicable to the prior art.

[0068] Obviously, the above embodiments are only examples for clearly illustrating, but not limit the embodiments. For those skilled in the art, other different forms of changes or variations can be made on the basis of the above description. Here, all the embodiments need not and cannot be exhausted. The obvious changes or variations derived therefrom are still within the protection scope of the application.

Claims

1. A wearable device-based abnormal behavior early warning method, characterized by, Wearing a wearable monitoring device; collecting and recording daily activity data; establishing a daily activity level model according to the daily activity data; calculating an individual activity level baseline using the daily activity level model; Comparing the collected real-time activity data with the individual activity level baseline, and determining whether the activity data is abnormal; If abnormal activity data occurs, an early warning signal is sent to the management personnel; acceleration data collected in real time by the wearable monitoring device is extracted, and acceleration mean, acceleration jerk and acceleration entropy are extracted; the daily activity level model is a multiple linear regression model constructed based on the acceleration mean, acceleration jerk and acceleration entropy extracted from the data collected in the previous three days, and the expression formula is as follows: AL=β0+β1*meanACC+β2jerkACC+β3*entropyACC Where: AL: daily activity level model; β0 is the constant term in the multiple linear regression model; β1, β2, β3 are the regression coefficients of acceleration mean, acceleration jerk and acceleration entropy characteristics; meanACC: 1min acceleration mean; jerkACC: acceleration jerk, i.e. the average value of 1min acceleration first derivative obtained by using linear regression; entropyACC: 1min acceleration value entropy; In the formula, N is the number of acceleration value samples in 1 min, and p i is the occurrence probability of the event of acceleration value = i in this 1 min, i.e. the number of points with value i ÷ N. 2.The wearable device based abnormal behavior early warning method of claim 1, wherein, When it is determined that the activity data of the day has no abnormal behavior, the daily activity data formed on the day is continued to be stacked on the normal activity data formed in the previous three days to form an updated daily activity level model for judging abnormal behavior. 3.The wearable device based abnormal behavior early warning method of claim 2, wherein, The daily activity level model is used to calculate the activity level value of the previous three days, and the individual activity level baseline is obtained by averaging the activity level values of the previous three days; the real-time collected data is compared with the obtained individual activity level baseline, and the absolute value of the difference between the real-time activity level and the individual activity level baseline is calculated; whether abnormal behavior occurs is determined according to the set threshold value. 4.The wearable device based abnormal behavior early warning method of claim 3, wherein, The set threshold value is 1.5std, if the absolute value of the difference is >1.5std: abnormal activity level is prompted, if the absolute value of the difference is ≤1.5std, then the activity level is normal.

5. A wearable device based abnormal behavior early warning system, characterized in that, The system is used for the abnormal behavior early warning method based on the wearable device of any one of claims 1-4, and the system comprises a wearable monitoring device, an early warning device, a computer, and a data recording module, a data processing module, a model establishing module, a calculation module and an abnormal behavior discrimination module embedded in the computer; The wearable monitoring device is worn on the body to collect motion signals; The data recording module is used to receive the motion signals collected by the wearable monitoring device; The data processing module is used to pre-process the motion signals recorded by the data recording module, and extract acceleration mean, acceleration jerk and acceleration entropy characteristic data; The model establishing module is used to model the extracted characteristic data to obtain a daily activity level model; The calculation module uses the established daily activity level model to calculate an individual activity level baseline; The abnormal behavior judging module is configured to compare the collected real-time motion data with an individual activity level baseline to determine whether there is abnormal behavior. The early warning device receives the abnormal behavior signal determined by the abnormal behavior judging module and sends an early warning signal to a manager.

6. The wearable device based pre-alarm system for abnormal behavior according to claim 5, wherein, When the abnormal behavior judging module determines that there is no abnormal behavior, the collected normal activity data is transmitted to the model establishing module to iteratively update the daily activity model.

7. The wearable device based pre-alarm system for abnormal behavior according to claim 6, wherein, The wearable monitoring device is a smart bracelet with a three-axis acceleration sensor.

Images