CN108771543B - A method and system for elderly fall detection in real environment based on big data - Google Patents

Abstract

The invention discloses a big data-based fall detection method and system for the elderly in a real environment, and belongs to the technical field of pattern recognition and human behavior recognition. It cannot meet the individual differences and cannot obtain the fall test data of the elderly. The technical solutions adopted are: ① A big data-based fall detection method for the elderly in a real environment. This method is based on a big data platform, using wearable devices and the Internet of Things. The technology builds a fall detection system for the elderly in a real environment, collects the daily behavior data of the elderly in the real environment through wearable devices, and transmits the data to the big data platform through the Internet of Things technology. ② A fall detection system for the elderly in a real environment based on big data, the system includes wearable devices, peripheral alarm devices and a big data platform, and the wearable devices include sensors, microcontrollers and wireless transmission modules.

Description

A method and system for elderly fall detection in real environment based on big data

technical field

The invention relates to the technical field of pattern recognition and human behavior recognition, in particular to a big data-based fall detection method and system for the elderly in a real environment.

Background technique

Wearable device-based fall detection systems are generally worn on the chest, waist, legs, and wrists. By adding relevant attitude sensors to the device, a wearable detection device is formed, and the data output by the sensor during the movement of the experimental object is collected to analyze and detect the fall behavior. Another advantage of wearable devices is that they can collect physiological indicators, such as blood pressure and ECG, that cannot be collected by other video- or environment-based fall detection systems. Shang-Lin Hsieh et al. of Tatung University proposed a wrist fall detection system, which obtained good average sensitivity and specificity through experiments. Pannurat, N. et al. proposed an automatic fall watch control system. Shany, T. et al. used wearable sensors for human behavior recognition and fall detection. Habib, M.A., etc. use smartphones for fall detection. Casilari, E. et al. proposed a fall detection scheme based on Android system. Yuan, J, et al. proposed the use of an efficient interrupt-driven algorithm for fall detection and human daily behavior recognition. Vavoulas, G. et al. used smart phones to complete the fall detection of the human body. Medrano, C. et al. proposed a new method to use smart phones to obtain acceleration for fall detection. López, J.D. et al. constructed a behavior detection system using the fusion of multiple accelerations. Koshmak, G.A. et al. combined Android system and physiological detection for fall detection. Many domestic researchers have also proposed many methods for fall detection, but most of them are about algorithm improvement or the combination of different sensors to achieve fall detection, and they all use young volunteers to simulate falls and daily behaviors. According to experimental data, once these detection algorithms are applied to real elderly fall data samples, the accuracy of fall detection is greatly reduced.

The existing fall detection systems have the following problems:

(1) The existing fall detection systems all use young volunteers to simulate falls and daily behaviors to obtain experimental data. Although a high accuracy rate has been obtained, once the algorithm is used to detect falls of the elderly in a real environment, its accuracy rate is greatly reduced;

(2) The fall detection system cannot meet individual differences. Different users have different heights and weights in specific application scenarios, and the system parameters will also change during the fall detection process. The direct consequence of the unification of system parameters is to aggravate fall detection. error;

(3) During the experiment, it is impossible for the elderly to simulate falling to obtain experimental data, that is, it is difficult to obtain the real falling experimental data of the elderly.

SUMMARY OF THE INVENTION

The technical task of the present invention is to provide a big data-based fall detection method and system for the elderly in a real environment, so as to solve the problem that the existing fall detection systems and methods have poor accuracy, cannot meet individual differences, and cannot obtain fall experimental data for the elderly. The problem.

The technical task of the present invention is achieved in the following manner, a big data-based fall detection method for the elderly in a real environment, the method is based on a big data platform, using wearable devices and Internet of Things technology to build a real environment for falling detection methods for the elderly The system collects the daily behavior data of the elderly in the real environment through wearable devices, and transmits the data to the big data platform through the Internet of Things technology; the specific steps are as follows:

S1. The wearable device collects triaxial acceleration data generated by human activity and transmits it to the central processing system of the big data platform through a wireless network, and the next step is to perform step S2;

S2, the central processing system performs feature extraction and normalization on the collected X, Y, Z three-axis acceleration data, and the next step is to perform step S3;

S3. Determine whether it is a preset stage:

(1), if yes, then the big data platform is a preset stage, and the next step jumps to step S8;

(2), if not, then the big data platform is not in the preset stage, and the next step is to perform step S4;

S4, use the classifier to detect human behavior classes on the behavior features extracted in step S2, and execute step S5 in the next step;

S5, according to the behavior class detected in step S4, the central processing system compares the behavior class in the behavior class database with the behavior class detected in step S4, and judges whether this behavior class is in the behavior class database:

(1), if it is, then execute step S6;

(2), if not, then jump to step S7;

S6, the central processing system compares the behavior class detected in step S4 with the fall behavior class in the behavior class database, and determines whether this behavior class is a fall behavior:

(1), if yes, then execute step S7;

(2), if not, then execute step S9;

S7, the central processing system triggers the alarm module to send an alarm signal to the peripheral alarm device, and the peripheral alarm device sends an alarm signal to remind the children of the elderly or the caregivers that the elderly may be in a dangerous situation, and the next step is to perform step S8;

S8, inquire and confirm the behavior class name, and store the behavior class in the behavior class database, and execute step S9 in the next step;

S9. Add the extracted sample features to the corresponding behavior sample database.

Preferably, the wearable device is integrated by a sensor, a wireless transmission module and a single-chip microcomputer.

Preferably, the sensor adopts ADXL345 sensor, and the wireless transmission module adopts wireless transmission module CC1000.

Preferably, in the step S2, the feature extraction of the collected X, Y, and Z acceleration data is specifically: according to the collected data, the mean value on the Y and Z axes are extracted from the time domain, and the average value of the Y and Z axes is extracted from the time domain. The quantile value and the quantile value below 75, in the frequency domain, extract the maximum frequency of the spectrum, the frequency component values below 5Hz, and the peak value of the spectrum below 5Hz on the Y-axis.

More preferably, the sample data is extracted on the Y and Z axes, that is, the acceleration mag on the Y and Z axes is calculated:

Among them, A _y refers to the acceleration value obtained by the wearable sensor on the Y axis; A _z refers to the acceleration value obtained by the wearable sensor on the Z axis.

More preferably, the mean value of the Y and Z axes extracted in the time domain is:

Among them, A _yi represents the ith acceleration sample value obtained by the wearable sensor on the Y axis; A _zi represents the ith acceleration sample value obtained by the wearable sensor on the Z axis; n refers to the wearable sensor in the Obtain n acceleration sample values on Y and Z respectively;

The 25th quantile value p25 and the 75th quantile value p75 of the mag data are calculated by the function prctile(), and the sums of squares sumsq25 and sumsq75 of the mag data below p25 and below p75 are calculated respectively.

More preferably, based on the feature extraction of the Y-axis in the frequency domain, the dispersion of the acceleration in the direction of the Y-axis is obtained:

Among them, A _y represents the acceleration value obtained by the wearable sensor on the Y axis; A _yi represents the ith acceleration sample value obtained by the wearable sensor on the Y axis; n refers to the wearable sensor on Y and Z respectively. Get n acceleration sample values;

After performing the fast Fourier transform on the dispersion, the maximum frequency of the spectrum maxFreq, the sum of the frequency components of 5 Hz or less, and the peak value numPeaks of the spectrum of 5 Hz or less are obtained, respectively.

More preferably, the time domain and frequency domain feature quantities extracted in the step S2 are combined into a feature vector, and the feature vector is normalized.

A fall detection system for the elderly in a real environment based on big data. The system includes wearable devices, peripheral alarm devices and a big data platform. The wearable device includes sensors, a single-chip microcomputer and a wireless transmission module. The big data platform includes a database and central processing. In the system, the single-chip microcomputer connects and controls the sensor and the wireless transmission module, and the wireless transmission module is wirelessly connected to the central processing system, and the central processing system sends the alarm information to the peripheral alarm equipment;

The sensor is used to collect data of triaxial acceleration generated by human activities;

The microcontroller is used to control the working state of the sensor and the wireless transmission module

The wireless transmission module is used to transmit the data collected by the sensor to the central processing system;

The database is used to classify and store the daily behavior data of the elderly in the real environment; the database includes a behavior database and a behavior sample database, the behavior database is used to store the behavior data, and the behavior sample database is used to store the behavior samples;

The central processing system is used to process the received data and issue processing orders;

The peripheral alarm device is used to receive the alarm information sent by the central processing system and send out an alarm signal to remind children or caregivers that the elderly may have dangerous situations. The peripheral alarm device adopts an alarm device or a mobile terminal.

Preferably, the central processing system includes a feature processing module, a preset stage discrimination module, a classifier, a behavioral data comparison module, a fall behavior discrimination module, an alarm module, and a database creation and maintenance module;

The feature processing module is used for feature extraction and normalization of the three-axis acceleration transmitted by the wireless transmission module to the central processing system;

The preset stage determination module is used to determine whether the elderly fall detection system is in the preset stage;

The classifier adopts SVM-based classifier, and the classifier is used to detect human behavior classes on the extracted behavior features;

The behavior class data comparison module is used to judge whether the behavior class detected by the classifier is in the existing behavior class database;

The fall behavior discrimination module is used to judge whether the behavior class detected by the classifier is a fall behavior;

The alarm module is used to send alarm information to the elderly's children or caregivers in abnormal situations;

The database creation and maintenance module is used to establish the database, add the behavior database and behavior sample database, and update the later management work.

Compared with the prior art, the method and system for detecting falls of the elderly in a real environment based on big data of the present invention have the following advantages:

(1), the present invention can effectively solve the problem that the accuracy of the algorithm is greatly reduced when many fall detection algorithms use young people to simulate falling to obtain data, and the accuracy of the algorithm is greatly reduced when applied to the elderly in the real environment. Store some daily behavior data of the elderly as training samples, and classify them into categories. In the subsequent use process, if a behavior category that is not in the database is detected, an alarm will be triggered, and it will be confirmed whether the action is a fall. If it is a fall, the database will One category is added to the system, and the data is stored in the database. With the growth of use time, the data samples in the database continue to increase, and the accuracy of fall detection will also continue to improve, realizing the intelligence of the real environment of the elderly under the background of big data. Fall detection greatly improves the stability, accuracy and detection efficiency of the elderly fall detection system in the real environment;

(2), the present invention takes the big data platform as the background, and applies the Internet of Things technology to identify and detect the behavior of the elderly in the real environment, especially for the fall of the elderly, to carry out real-time monitoring and control, and timely alarm processing when abnormal conditions occur. The real behavior data of the elderly can be used for detection and identification, which can vary from person to person, and can realize the individualization of samples, thereby improving the accuracy of behavior discrimination of the elderly;

(3) The present invention applies the big data platform, and can collect data samples that vary from person to person, so that everyone who uses the present invention collects his own data samples, and with the extension of the use time, the behavioral database and The amount of corresponding sample database data will be more and more, and the detection of the classifier will be more and more accurate, so it can achieve fall detection that varies from person to person, and also achieve sample collection in the real environment, which is more conducive to improving Fall detection accuracy.

Description of drawings

The present invention will be further described below in conjunction with the accompanying drawings.

Accompanying drawing 1 is a flowchart of a method for detecting falls of the elderly in a real environment based on big data;

FIG. 2 is a structural block diagram of a fall detection system for the elderly in a real environment based on big data.

Detailed ways

With reference to the accompanying drawings and specific embodiments of the description, a method and system for detecting a fall of the elderly in a real environment based on big data of the present invention will be described in detail below.

Example:

As shown in FIG. 1, the method for detecting the fall of the elderly in a real environment based on big data of the present invention is based on a big data platform, using wearable devices and Internet of Things technology to build a fall detection system for the elderly in a real environment. The wearable device collects the daily behavior data of the elderly in the real environment, and transmits the data to the big data platform through the Internet of Things technology; the specific steps are as follows:

S1. The ADXL345 sensor, the wireless transmission module CC1000 and the single-chip integrated wearable device collect the triaxial acceleration data generated by human activity and transmit it to the central processing system of the big data platform through the wireless network, and the next step is to perform step S2;

S2, the central processing system performs feature extraction and normalization on the collected X, Y, Z three-axis acceleration data, and the next step is to perform step S3;

S3. Determine whether it is a preset stage:

(1), if yes, then the big data platform is a preset stage, and the next step jumps to step S8;

(2), if not, then the big data platform is not in the preset stage, and the next step is to perform step S4;

S4, use the classifier to detect human behavior classes on the behavior features extracted in step S2, and execute step S5 in the next step;

S5, according to the behavior class detected in step S4, the central processing system compares the behavior class in the behavior class database with the behavior class detected in step S4, and judges whether this behavior class is in the behavior class database:

(1), if it is, then execute step S6;

(2), if not, then jump to step S7;

S6, the central processing system compares the behavior class detected in step S4 with the fall behavior class in the behavior class database, and determines whether this behavior class is a fall behavior:

(1), if yes, then execute step S7;

(2), if not, then execute step S9;

S7, the central processing system triggers the alarm module to send an alarm signal to the peripheral alarm device, and the peripheral alarm device sends an alarm signal to remind the children of the elderly or the caregivers that the elderly may be in a dangerous situation, and the next step is to perform step S8;

S8, inquire and confirm the behavior class name, and store the behavior class in the behavior class database, and execute step S9 in the next step;

S9. Add the extracted sample features to the corresponding behavior sample database.

Among them, the feature extraction of the collected X, Y, Z acceleration data in step S2 is specifically: according to the collected data, the mean value on the Y and Z axes and the quantile value below 25 are respectively extracted from the time domain. and the quantile value below 75, extract the maximum frequency of the spectrum, the frequency component values below 5Hz and the peak value of the spectrum below 5Hz on the Y-axis in the frequency domain, as follows:

(1) Extract sample data on the Y and Z axes, that is, calculate the acceleration mag on the Y and Z axes:

Among them, A _y refers to the acceleration value obtained by the wearable sensor on the Y axis; A _z refers to the acceleration value obtained by the wearable sensor on the Z axis.

(2), in the time domain, the mean value of the Y and Z axes is extracted as:

Among them, A _yi represents the ith acceleration sample value obtained by the wearable sensor on the Y axis; A _zi represents the ith acceleration sample value obtained by the wearable sensor on the Z axis; n refers to the wearable sensor in the Obtain n acceleration sample values on Y and Z respectively;

The 25th quantile value p25 and the 75th quantile value p75 of the mag data are calculated by the function prctile(), and the sums of squares sumsq25 and sumsq75 of the mag data below p25 and below p75 are calculated respectively.

(3), based on the feature extraction of the Y-axis in the frequency domain, to obtain the dispersion of the acceleration in the direction of the Y-axis:

Among them, A _y represents the acceleration value obtained by the wearable sensor on the Y axis; A _yi represents the ith acceleration sample value obtained by the wearable sensor on the Y axis; n refers to the wearable sensor on Y and Z respectively. Get n acceleration sample values;

After performing the fast Fourier transform on the dispersion, obtain the maximum frequency maxFreq of the spectrum, the sum5Hz of the frequency components below 5Hz, and the peak numPeaks of the spectrum below 5Hz;

(4) Combine the extracted time domain and frequency domain feature quantities into one feature vector, and normalize the feature vector.

Example 2:

As shown in FIG. 2, the big data-based fall detection system for the elderly in the real environment of the present invention includes a wearable device, a peripheral alarm device and a big data platform, and the wearable device includes a sensor, a single-chip microcomputer and a wireless transmission module, The big data platform includes a database and a central processing system. The single-chip microcomputer connects and controls the sensor and the wireless transmission module. The wireless transmission module is wirelessly connected to the central processing system. The central processing system sends alarm information to the peripheral alarm equipment; the sensor is used to collect the three The data of shaft acceleration; the single-chip microcomputer is used to control the working state of the sensor and the wireless transmission module. The wireless transmission module is used to transmit the data collected by the sensor to the central processing system; the database is used to classify and store the daily behavior data of the elderly in the real environment; the database includes Behavior database and behavior sample database, behavior database is used to store behavior data, behavior sample database is used to store behavior samples; central processing system is used to process received data and issue processing commands; peripheral alarm equipment is used to receive central Process the alarm information sent by the system and send out an alarm signal to remind children or caregivers that the elderly may be in a dangerous situation. The peripheral alarm device adopts an alarm or a mobile terminal.

The central processing system includes a feature processing module, a preset stage discrimination module, a classifier, a behavioral data comparison module, a fall behavior discrimination module, an alarm module, and a database creation and maintenance module; the characteristic processing module is used to transmit the wireless transmission module to the central processing unit. The three-axis acceleration of the system is feature extracted and normalized; the preset stage discrimination module is used to judge whether the elderly fall detection system is in the preset stage; the classifier adopts the classifier based on SVM, and the classifier is used for the extracted behavior. The feature is used to detect human behavior classes; the behavior class data comparison module is used to determine whether the behavior class detected by the classifier is in the existing behavior class database; the fall behavior discrimination module is used to determine whether the behavior class detected by the classifier is a fall Behavior; the alarm module is used to send alarm information to the children or caregivers of the elderly in abnormal situations; the database creation and maintenance module is used to establish a database, add behavior database and behavior sample database, and update management work later.

Specific working process:

(1) Use the single-chip microcomputer and sensors to collect the data of the three-axis acceleration generated by human activities, and use the wireless transmission module to transmit the data to the central processing system of the big data platform;

(2) The central processing system performs feature extraction and normalization processing on the received data through the feature processing module, and judges whether the system is in the preset stage through the preset stage discrimination module:

(I) If the central processing system is set to the preset stage at the beginning of use, the database creation and maintenance module needs to add and collect data samples for some daily behaviors and actions of individuals, so as to provide samples for the later classifier detection;

(II) When in the normal use stage, use the classifier to detect the collected data features, and judge whether the behavior class is in the database through the behavior class data comparison module:

(i) If this behavior class is in the behavior class database, continue to judge whether it is a fall behavior through the fall behavior discrimination module:

①. If it falls, it will trigger the alarm module to alarm;

2. If it is not a fall behavior, add the extracted data features to the data sample library of the corresponding behavior class;

(ii) If the behavior class is not in the behavior class database, that is, if the classifier detects that the behavior class is an unknown behavior, it will first trigger the alarm module to alarm, and at the same time, the peripheral alarm device will send an alarm signal (because the daily behavior in the behavior class database has been If an unknown behavior class is detected, it is suspected of falling, and the alarm is triggered first for safety); then the name of the behavior class is confirmed, and the new behavior class is stored in the behavior class database through the database creation and maintenance module , and add the data features to the corresponding behavior database.

Through the above specific embodiments, those skilled in the art can easily implement the present invention. However, it should be understood that the present invention is not limited to the specific embodiments described above. On the basis of the disclosed embodiments, those skilled in the technical field can arbitrarily combine different technical features to realize different technical solutions.

Except for the technical features described in the specification, they are all known technologies by those skilled in the art.

Claims (9)

1. a fall detection method for the elderly under a real environment based on big data, is characterized in that the method is based on a big data platform, utilizes wearable equipment and Internet of Things technology to build a fall detection system for the elderly under a real environment, collects by wearable equipment The daily behavior data of the elderly in the real environment is transmitted to the big data platform through the Internet of Things technology; the specific steps are as follows: S1. The wearable device collects triaxial acceleration data generated by human activity and transmits it to the central processing system of the big data platform through a wireless network, and the next step is to perform step S2; S2, the central processing system performs feature extraction and normalization on the collected X, Y, Z three-axis acceleration data, and the next step is to perform step S3; S3. Determine whether it is a preset stage: (1), if yes, then the big data platform is a preset stage, and the next step jumps to step S8; (2), if not, then the big data platform is not in the preset stage, and the next step is to perform step S4; S4, use the classifier to detect human behavior classes on the behavior features extracted in step S2, and execute step S5 in the next step; S5, according to the behavior class detected in step S4, the central processing system compares the behavior class in the behavior class database with the behavior class detected in step S4, and judges whether this behavior class is in the behavior class database: (1), if it is, then execute step S6; (2), if not, then jump to step S7; S6, the central processing system compares the behavior class detected in step S4 with the fall behavior class in the behavior class database, and determines whether this behavior class is a fall behavior: (1), if yes, then execute step S7; (2), if not, then execute step S9; S7, the central processing system triggers the alarm module to send an alarm signal to the peripheral alarm device, and the peripheral alarm device sends an alarm signal to remind the children of the elderly or the caregivers that the elderly may be in a dangerous situation, and the next step is to perform step S8; S8, inquire and confirm the behavior class name, and store the behavior class in the behavior class database, and execute step S9 in the next step; S9. Add the extracted sample features to the corresponding behavior sample database. 2 . The method for detecting falls of the elderly in a real environment based on big data according to claim 1 , wherein the wearable device is integrated by a sensor, a wireless transmission module and a single-chip microcomputer. 3 . 3 . The method for detecting falls of the elderly in a real environment based on big data according to claim 2 , wherein the sensor adopts an ADXL345 sensor, and the wireless transmission module adopts a wireless transmission module CC1000. 4 . 4. the fall detection method of the elderly based on the real environment of big data according to claim 1, it is characterized in that carrying out feature extraction to the X, Y, Z three-axis acceleration data collected in described step S2 specifically: according to collecting The obtained data extracts the mean value on the Y and Z axes, the quantile value below 25, and the quantile value below 75 from the time domain, and extracts the maximum frequency of the spectrum based on the Y axis in the frequency domain, and frequencies below 5Hz Component values and peaks of the spectrum below 5Hz. 5. the fall detection method of the elderly based on the real environment of big data according to claim 4, is characterized in that extracting sample data on Y, Z two axes, namely calculates the acceleration mag on Y, Z two axes:

Among them, A _y refers to the acceleration value obtained by the wearable sensor on the Y axis; A _z refers to the acceleration value obtained by the wearable sensor on the Z axis. 6. according to claim 4 or 5, it is characterized in that the mean value on Y, Z two axes is extracted in the time domain as:

Among them, A _yi represents the ith acceleration sample value obtained by the wearable sensor on the Y axis; A _zi represents the ith acceleration sample value obtained by the wearable sensor on the Z axis; n refers to the wearable sensor in the Obtain n acceleration sample values on Y and Z respectively; The 25th quantile value p25 and the 75th quantile value p75 of the mag data are calculated by the function prctile(), and the sums of squares sumsq25 and sumsq75 of the mag data below p25 and below p75 are calculated respectively. 7. the fall detection method of the elderly under the real environment based on big data according to claim 6, is characterized in that in frequency domain based on the feature extraction of Y-axis, namely seeks the dispersion of the acceleration on Y-axis direction:

Among them, A _y represents the acceleration value obtained by the wearable sensor on the Y axis; A _yi represents the ith acceleration sample value obtained by the wearable sensor on the Y axis; n refers to the wearable sensor on Y and Z respectively. Get n acceleration sample values; After performing the fast Fourier transform on the dispersion, the maximum frequency of the spectrum maxFreq, the sum of the frequency components of 5 Hz or less, and the peak value numPeaks of the spectrum of 5 Hz or less are obtained, respectively. 8. the fall detection method of the elderly under the real environment based on big data according to claim 7 is characterized in that the time domain and the frequency domain feature quantity extracted in the described step S2 are merged into a feature vector, and the feature vector is normalized. Unified processing. 9. A fall detection system for the elderly in a real environment based on big data, characterized in that the system comprises a wearable device, an external alarm device and a big data platform, the wearable device comprises a sensor, a single-chip microcomputer and a wireless transmission module, and the big data platform Including the database and the central processing system, the single-chip microcomputer connects and controls the sensor and the wireless transmission module, the wireless transmission module wirelessly connects the central processing system, and the central processing system sends the alarm information to the peripheral alarm equipment; The sensor is used to collect data of triaxial acceleration generated by human activities; The microcontroller is used to control the working state of the sensor and the wireless transmission module The wireless transmission module is used to transmit the data collected by the sensor to the central processing system; The database is used to classify and store the daily behavior data of the elderly in the real environment; the database includes a behavior database and a behavior sample database, the behavior database is used to store the behavior data, and the behavior sample database is used to store the behavior samples; The central processing system is used to process the received data and issue processing orders; the central processing system includes a feature processing module, a preset stage discrimination module, a classifier, a behavior data comparison module, a fall behavior discrimination module, an alarm module, and database creation and maintenance. module; The feature processing module is used for feature extraction and normalization of the three-axis acceleration transmitted by the wireless transmission module to the central processing system; The preset stage determination module is used to determine whether the elderly fall detection system is in the preset stage; The classifier adopts SVM-based classifier, and the classifier is used to detect human behavior classes on the extracted behavior features; The behavior class data comparison module is used to judge whether the behavior class detected by the classifier is in the existing behavior class database; The fall behavior discrimination module is used to judge whether the behavior class detected by the classifier is a fall behavior; The alarm module is used to send alarm information to the elderly's children or caregivers in abnormal situations; Database creation and maintenance module is used to establish database, add behavior database and behavior sample database, and update management work later; The peripheral alarm device is used to receive the alarm information sent by the central processing system and send out an alarm signal to remind children or caregivers that the elderly may be in danger. The peripheral alarm device adopts an alarm device or a mobile terminal.

Images