CN115462782A - Human body falling dynamic monitoring method and system based on multi-dimensional characteristic parameters - Google Patents

Human body falling dynamic monitoring method and system based on multi-dimensional characteristic parameters Download PDF

Info

Publication number
CN115462782A
CN115462782A CN202211085849.4A CN202211085849A CN115462782A CN 115462782 A CN115462782 A CN 115462782A CN 202211085849 A CN202211085849 A CN 202211085849A CN 115462782 A CN115462782 A CN 115462782A
Authority
CN
China
Prior art keywords
human body
characteristic
falling
parameter set
decision
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202211085849.4A
Other languages
Chinese (zh)
Other versions
CN115462782B (en
Inventor
李志�
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Sichuan Kanggu Kangye Technology Co ltd
Original Assignee
Sichuan Kanggu Kangye Technology Co ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Sichuan Kanggu Kangye Technology Co ltd filed Critical Sichuan Kanggu Kangye Technology Co ltd
Priority to CN202211085849.4A priority Critical patent/CN115462782B/en
Publication of CN115462782A publication Critical patent/CN115462782A/en
Application granted granted Critical
Publication of CN115462782B publication Critical patent/CN115462782B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • AHUMAN NECESSITIES
    • A61MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61BDIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00Measuring for diagnostic purposes; Identification of persons
    • A61B5/103Detecting, measuring or recording devices for testing the shape, pattern, colour, size or movement of the body or parts thereof, for diagnostic purposes
    • A61B5/11Measuring movement of the entire body or parts thereof, e.g. head or hand tremor, mobility of a limb
    • A61B5/1116Determining posture transitions
    • A61B5/1117Fall detection
    • AHUMAN NECESSITIES
    • A61MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61BDIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00Measuring for diagnostic purposes; Identification of persons
    • A61B5/02Detecting, measuring or recording pulse, heart rate, blood pressure or blood flow; Combined pulse/heart-rate/blood pressure determination; Evaluating a cardiovascular condition not otherwise provided for, e.g. using combinations of techniques provided for in this group with electrocardiography or electroauscultation; Heart catheters for measuring blood pressure
    • A61B5/0205Simultaneously evaluating both cardiovascular conditions and different types of body conditions, e.g. heart and respiratory condition
    • AHUMAN NECESSITIES
    • A61MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61BDIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00Measuring for diagnostic purposes; Identification of persons
    • A61B5/103Detecting, measuring or recording devices for testing the shape, pattern, colour, size or movement of the body or parts thereof, for diagnostic purposes
    • A61B5/11Measuring movement of the entire body or parts thereof, e.g. head or hand tremor, mobility of a limb
    • A61B5/1112Global tracking of patients, e.g. by using GPS
    • AHUMAN NECESSITIES
    • A61MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61BDIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00Measuring for diagnostic purposes; Identification of persons
    • A61B5/68Arrangements of detecting, measuring or recording means, e.g. sensors, in relation to patient
    • A61B5/6801Arrangements of detecting, measuring or recording means, e.g. sensors, in relation to patient specially adapted to be attached to or worn on the body surface
    • A61B5/6802Sensor mounted on worn items
    • AHUMAN NECESSITIES
    • A61MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61BDIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00Measuring for diagnostic purposes; Identification of persons
    • A61B5/68Arrangements of detecting, measuring or recording means, e.g. sensors, in relation to patient
    • A61B5/6801Arrangements of detecting, measuring or recording means, e.g. sensors, in relation to patient specially adapted to be attached to or worn on the body surface
    • A61B5/6802Sensor mounted on worn items
    • A61B5/6803Head-worn items, e.g. helmets, masks, headphones or goggles
    • AHUMAN NECESSITIES
    • A61MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61BDIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00Measuring for diagnostic purposes; Identification of persons
    • A61B5/68Arrangements of detecting, measuring or recording means, e.g. sensors, in relation to patient
    • A61B5/6801Arrangements of detecting, measuring or recording means, e.g. sensors, in relation to patient specially adapted to be attached to or worn on the body surface
    • A61B5/6802Sensor mounted on worn items
    • A61B5/681Wristwatch-type devices
    • AHUMAN NECESSITIES
    • A61MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61BDIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00Measuring for diagnostic purposes; Identification of persons
    • A61B5/72Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7235Details of waveform analysis
    • A61B5/7264Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
    • A61B5/7267Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems involving training the classification device
    • AHUMAN NECESSITIES
    • A61MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61BDIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00Measuring for diagnostic purposes; Identification of persons
    • A61B5/72Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7271Specific aspects of physiological measurement analysis
    • A61B5/7282Event detection, e.g. detecting unique waveforms indicative of a medical condition
    • AHUMAN NECESSITIES
    • A61MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61BDIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00Measuring for diagnostic purposes; Identification of persons
    • A61B5/74Details of notification to user or communication with user or patient ; user input means
    • A61B5/7405Details of notification to user or communication with user or patient ; user input means using sound
    • AHUMAN NECESSITIES
    • A61MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61BDIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00Measuring for diagnostic purposes; Identification of persons
    • A61B5/74Details of notification to user or communication with user or patient ; user input means
    • A61B5/742Details of notification to user or communication with user or patient ; user input means using visual displays
    • AHUMAN NECESSITIES
    • A61MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61BDIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00Measuring for diagnostic purposes; Identification of persons
    • A61B5/74Details of notification to user or communication with user or patient ; user input means
    • A61B5/746Alarms related to a physiological condition, e.g. details of setting alarm thresholds or avoiding false alarms
    • AHUMAN NECESSITIES
    • A61MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61BDIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00Measuring for diagnostic purposes; Identification of persons
    • A61B5/74Details of notification to user or communication with user or patient ; user input means
    • A61B5/7465Arrangements for interactive communication between patient and care services, e.g. by using a telephone network
    • A61B5/747Arrangements for interactive communication between patient and care services, e.g. by using a telephone network in case of emergency, i.e. alerting emergency services
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/08Learning methods
    • GPHYSICS
    • G08SIGNALLING
    • G08BSIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
    • G08B21/00Alarms responsive to a single specified undesired or abnormal condition and not otherwise provided for
    • G08B21/02Alarms for ensuring the safety of persons
    • G08B21/04Alarms for ensuring the safety of persons responsive to non-activity, e.g. of elderly persons
    • G08B21/0407Alarms for ensuring the safety of persons responsive to non-activity, e.g. of elderly persons based on behaviour analysis
    • G08B21/043Alarms for ensuring the safety of persons responsive to non-activity, e.g. of elderly persons based on behaviour analysis detecting an emergency event, e.g. a fall
    • GPHYSICS
    • G08SIGNALLING
    • G08BSIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
    • G08B21/00Alarms responsive to a single specified undesired or abnormal condition and not otherwise provided for
    • G08B21/02Alarms for ensuring the safety of persons
    • G08B21/04Alarms for ensuring the safety of persons responsive to non-activity, e.g. of elderly persons
    • G08B21/0438Sensor means for detecting
    • G08B21/0446Sensor means for detecting worn on the body to detect changes of posture, e.g. a fall, inclination, acceleration, gait
    • GPHYSICS
    • G08SIGNALLING
    • G08BSIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
    • G08B21/00Alarms responsive to a single specified undesired or abnormal condition and not otherwise provided for
    • G08B21/02Alarms for ensuring the safety of persons
    • G08B21/04Alarms for ensuring the safety of persons responsive to non-activity, e.g. of elderly persons
    • G08B21/0438Sensor means for detecting
    • G08B21/0453Sensor means for detecting worn on the body to detect health condition by physiological monitoring, e.g. electrocardiogram, temperature, breathing
    • GPHYSICS
    • G08SIGNALLING
    • G08BSIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
    • G08B21/00Alarms responsive to a single specified undesired or abnormal condition and not otherwise provided for
    • G08B21/02Alarms for ensuring the safety of persons
    • G08B21/04Alarms for ensuring the safety of persons responsive to non-activity, e.g. of elderly persons
    • G08B21/0438Sensor means for detecting
    • G08B21/0492Sensor dual technology, i.e. two or more technologies collaborate to extract unsafe condition, e.g. video tracking and RFID tracking
    • GPHYSICS
    • G08SIGNALLING
    • G08BSIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
    • G08B25/00Alarm systems in which the location of the alarm condition is signalled to a central station, e.g. fire or police telegraphic systems
    • G08B25/01Alarm systems in which the location of the alarm condition is signalled to a central station, e.g. fire or police telegraphic systems characterised by the transmission medium
    • G08B25/08Alarm systems in which the location of the alarm condition is signalled to a central station, e.g. fire or police telegraphic systems characterised by the transmission medium using communication transmission lines
    • GPHYSICS
    • G08SIGNALLING
    • G08BSIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
    • G08B25/00Alarm systems in which the location of the alarm condition is signalled to a central station, e.g. fire or police telegraphic systems
    • G08B25/01Alarm systems in which the location of the alarm condition is signalled to a central station, e.g. fire or police telegraphic systems characterised by the transmission medium
    • G08B25/10Alarm systems in which the location of the alarm condition is signalled to a central station, e.g. fire or police telegraphic systems characterised by the transmission medium using wireless transmission systems
    • GPHYSICS
    • G08SIGNALLING
    • G08BSIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
    • G08B7/00Signalling systems according to more than one of groups G08B3/00 - G08B6/00; Personal calling systems according to more than one of groups G08B3/00 - G08B6/00
    • G08B7/06Signalling systems according to more than one of groups G08B3/00 - G08B6/00; Personal calling systems according to more than one of groups G08B3/00 - G08B6/00 using electric transmission, e.g. involving audible and visible signalling through the use of sound and light sources

Landscapes

  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Physics & Mathematics (AREA)
  • Engineering & Computer Science (AREA)
  • General Health & Medical Sciences (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Molecular Biology (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Animal Behavior & Ethology (AREA)
  • Medical Informatics (AREA)
  • Surgery (AREA)
  • Public Health (AREA)
  • Veterinary Medicine (AREA)
  • Pathology (AREA)
  • Physiology (AREA)
  • General Physics & Mathematics (AREA)
  • Emergency Management (AREA)
  • Business, Economics & Management (AREA)
  • Artificial Intelligence (AREA)
  • Gerontology & Geriatric Medicine (AREA)
  • Psychiatry (AREA)
  • Cardiology (AREA)
  • Evolutionary Computation (AREA)
  • Mathematical Physics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Pulmonology (AREA)
  • Signal Processing (AREA)
  • Dentistry (AREA)
  • Oral & Maxillofacial Surgery (AREA)
  • Theoretical Computer Science (AREA)
  • Fuzzy Systems (AREA)
  • Radar, Positioning & Navigation (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Nursing (AREA)
  • Software Systems (AREA)
  • Emergency Medicine (AREA)

Abstract

The invention discloses a human body falling dynamic monitoring method and a system based on multidimensional characteristic parameters, wherein the method comprises the following steps: constructing a human body falling monitoring mechanism by using an inertial sensor and a biosensor, and extracting characteristic parameters of data acquired by the inertial sensor and the biosensor consisting of an acceleration meter and a gyroscope to form a characteristic parameter set; constructing a human body falling decision characteristic threshold value parameter set convolutional neural network learning model, and deciding an initial human body falling decision characteristic threshold value parameter set; correcting the initial parameter set of the human body falling decision characteristic threshold value according to the self characteristics of different users to obtain a user falling decision real-time characteristic threshold value parameter set; and according to the parameter set of the real-time characteristic threshold value for the falling decision of the user, carrying out real-time judgment and decision on the activity and behavior state of the user, and outputting a falling state result. According to the invention, multi-dimensional fall decision characteristic parameters are introduced, so that the fall monitoring precision is effectively improved.

Description

Human body falling dynamic monitoring method and system based on multi-dimensional characteristic parameters
Technical Field
The invention relates to the technical field of human behavior identification and data processing, in particular to a human body falling dynamic monitoring method and system based on multi-dimensional characteristic parameters.
Background
By 2021, the population is over 2.67 hundred million, accounting for 18.9% of the national population, at the age of 60 years and above. The population over 65 years old reaches 20056 ten thousand people, and the proportion of the population reaches 14.2 percent; wherein about 2540 thousands of households of solitary old people are counted in 2020. In 2030 years, the number of the old people living alone in China is estimated to reach 3722 thousands of households. With the increasing of the population of the old people year by year, the safety problem of the old people living alone should be guaranteed, the falling is always an important hidden danger threatening the safety of the old people, and the falling of the old people can bring great medical burden to families and society. How to improve the falling detection and alarm of the old during the activity, improve the health level of the old to the maximum extent and reduce the medical load is a very important medical problem and a social problem.
The falling is the first reason of death of the old aged over 65 years old in China due to injury, the related death rate is up to 30%, and the falling incidence rate is increased along with the increase of the age, so that the injury of disability and death caused by falling can be greatly reduced by timely and accurately treating the fallen old, and the method has a very important effect on the elderly living alone.
Falling is the involuntary fall of a human body on the ground or some lower planes under the action of the loss of the balance of the center of gravity, and a person falling cannot respond in time. Because of the different postures of the person who falls, the way of falling is also very different. At present, in the research field of human behavior identification, a visual fall detection technology, an audio vibration fall detection technology and an inertial sensor fall detection technology are common for fall detection. The three technical methods have respective advantages and disadvantages, and the visual falling detection technology is easy to configure and high in judgment accuracy, but is limited by ambient light, space areas and the like, and is related to privacy problems and high in cost. The audio frequency vibration falling detection technology has poor practicability, more interference information sources, low accuracy and high cost, and cannot be applied outdoors. Although the inertial sensor fall detection technology has the advantages of low cost, wide coverage range, easiness in use and the like, the accuracy is low, and the energy consumption charging frequency is high.
Currently, in human body fall monitoring research, a fall detection method based on an inertial sensor is one of the most common monitoring means. A series of falling detection methods or algorithms are carried out according to the obvious changes of weightlessness, overweight, violent collision and the like of the motion posture of the human body in the falling process. For example, the acceleration sensor is used for acquiring human activity behavior data, the human activity behavior data are converted into angle values, inclination gradient variances are obtained to serve as characteristic values, normal activity characteristic values are compared, and two fixed threshold values at two moments after falling and falling are generated are set to judge the falling detection algorithm of the human body. For example, gait data of human activity behaviors are collected through an acceleration sensor, a gyroscope, a geomagnetic instrument and a pressure sensor, and characteristic parameters are extracted according to normal gait data and abnormal gait data to train a support vector quaternary classifier, so that the condition of possible falling is prompted in advance, and then the user is subjected to real-time falling detection under the condition of possible falling, and further the falling detection of the human body is realized. For example, the data of the activities and behaviors of the human body are collected through an acceleration sensor, the acceleration threshold value is set in the falling stage of the falling period and the collision stage in the falling process to realize judgment, and the falling detection is realized by analyzing the standing and lying of the human body in combination with the angle. For example, the data of the activity behaviors of the human body are acquired through an acceleration sensor, a gyroscope and a pressure sensor, the activity posture of the human body can be changed sharply in the falling process of the human body, and the falling state of the human body can be detected by acquiring and analyzing information such as the acceleration, the angular velocity, the plantar pressure and the like in the change process of the posture of the human body through a motion sensor.
Although the above fall detection method can detect fall behaviors of most human bodies, the actual situations of the fall are complex, the individual differences are different, the number of introduced decisive elements for fall detection and judgment is small, and a judgment decision algorithm is simple, so that the fall detection accuracy is difficult to guarantee.
Disclosure of Invention
The invention aims to provide a dynamic human body falling monitoring method based on multi-dimensional characteristic parameters, and aims to solve the technical problems of low accuracy of falling false alarm, missed alarm and the like caused by few decisive elements, simple algorithm and single function in current human body falling monitoring.
The first purpose of the invention is realized by adopting the following technical scheme: a dynamic human body falling monitoring method based on multi-dimensional characteristic parameters comprises the following steps:
a human body falling monitoring mechanism is constructed by an inertial sensor and a biosensor which are composed of an acceleration and a gyroscope, and comprises a human body dynamic three-dimensional space monitoring mechanism and a biomedical vital sign heart rate monitoring mechanism;
constructing a human body falling decision characteristic threshold value parameter set convolutional neural network learning model, extracting characteristic parameters of data acquired by an inertial sensor and a biosensor consisting of an acceleration and a gyroscope to form a characteristic parameter set, and introducing the characteristic parameter set into the human body falling decision characteristic threshold value parameter set convolutional neural network learning model for deep learning, training, analyzing and deciding to obtain a human body falling decision characteristic threshold value initial parameter set;
correcting the initial parameter set of the human body falling decision characteristic threshold value through a human body falling decision characteristic threshold value parameter set convolutional neural network learning model according to the self characteristics of different users to obtain a user falling decision real-time characteristic threshold value parameter set;
and judging and deciding whether the activity behavior state of the user falls or not in real time according to the real-time characteristic threshold value parameter set for the falling decision of the user, and outputting a falling state result.
Furthermore, the human body dynamic three-dimensional space monitoring mechanism respectively corresponds three space coordinates of x, y and z of an inertial sensor consisting of an acceleration instrument and a gyroscope to the human body dynamic three-dimensional space coordinates. An x axis of an inertial sensor consisting of an acceleration and a gyroscope represents the change of the moving acceleration of the human body in the left and right directions; the acceleration and an inertial sensor y axis formed by the gyroscope represent the movement acceleration change of the human body space in the front-back direction; an inertial sensor z axis formed by the acceleration and the gyroscope represents the gravity acceleration change in the vertical direction of the human body space; the angular change of the z-axis of an inertial sensor consisting of an acceleration instrument and a gyroscope relative to the vertical direction represents the change of the attitude angular velocity of the inclination, rotation or offset of the human body.
Further, the human biomedical vital signs heart rate monitoring mechanism corresponds the PPG heart rate biosensor to a human vital signs heart rate value parameter state.
Further, the method for constructing the convolutional neural network learning model of the human fall decision characteristic threshold parameter set comprises the following steps: and constructing a human body falling decision characteristic threshold value parameter set convolutional neural network learning model through a non-falling daily life activity database and a falling activity behavior database.
Further, the method for extracting the characteristic parameters of the data acquired by the inertial sensor and the biosensor, which are composed of the acceleration and the gyroscope, includes: extracting characteristic parameters from activity behavior acceleration data, angular velocity data and vital sign heart rate data acquired from human body non-falling daily life activity and human body falling activity to form a characteristic parameter set, wherein the characteristic parameter set comprises a movement acceleration component difference characteristic parameter delta g x And Δ g y And a characteristic parameter delta g of the difference value of the gravity acceleration components z Attitude angular velocity difference characteristic parameter delta theta and synthetic acceleration vector amplitude characteristic parameter SVM k Synthetic acceleration vector amplitude extreme difference characteristic parameter delta SVM R Differential acceleration vector amplitude absolute average value characteristic parameter DSVM and synthetic acceleration vector amplitude standard deviation characteristic parameter SVM sd And a real-time heart rate difference characteristic parameter delta HRV.
Further, the extraction is acceleratedAfter characteristic parameters of data acquired by an inertial sensor and a biosensor consisting of a gyroscope and a gyroscope are input into a human body falling decision characteristic threshold parameter set convolutional neural network learning model to carry out deep learning, training, analysis and decision on a human body falling decision characteristic threshold initial parameter set, wherein the human body falling decision characteristic threshold initial parameter set comprises delta g xFall-TH Threshold value characteristic parameter, Δ g yFall-TH Characteristic threshold parameter, Δ g zFall-TH Characteristic threshold parameter, Δ θ Fall-TH Characteristic threshold value parameter, SVM k Fall-TH Characteristic threshold value parameter, delta SVM RFall-TH Characteristic threshold parameter, DSVM Fall-TH Characteristic threshold value parameter, SVM sdFall-TH Characteristic threshold parameter, Δ HRV Fall-TH A characteristic threshold parameter.
Further, the method for obtaining the user fall decision real-time characteristic threshold value parameter set to replace the human fall decision characteristic threshold value initial parameter set by correcting the human fall decision characteristic threshold value initial parameter set through the human fall decision characteristic threshold value parameter set convolutional neural network learning model according to the self characteristics of different users comprises the following steps: different users import activity behavior acceleration data, angular velocity data and vital sign heart rate data acquired by non-falling daily life activities and falling activities into a human body falling decision characteristic threshold value parameter set convolutional neural network learning model for learning, training, analyzing and modifying according to own characteristics to obtain a user falling decision real-time characteristic threshold value parameter set and replace a human body falling decision characteristic threshold value initial parameter set; the user fall decision real-time characteristic threshold parameter set comprises' delta g x Fall-TH Characteristic threshold parameter,' Δ g y Fall-TH Characteristic threshold parameter,' Δ g zFall-TH Characteristic threshold parameter,' Delta theta Fall-TH Characteristic threshold parameter,' SVM k Fall-TH Characteristic threshold parameter,' delta SVM RFall-TH Characteristic threshold parameter,' DSVM Fall-TH Characteristic threshold parameter,' SVM sdFall-TH Characteristic threshold parameter,' Δ HRV Fall-TH A characteristic threshold parameter.
Further, the method for judging and deciding whether the activity behavior state of the user falls down in real time according to the real-time characteristic threshold parameter set for the user fall decision comprises the following steps:
a. by the difference of the moving acceleration components in the feature parameter set, the feature parameter Δ g x And Δ g y And a characteristic parameter Δ g of the gravity acceleration component difference z Judging the falling state of the human body;
b. further judging the falling state of the human body according to the attitude angular velocity difference characteristic parameter delta theta in the characteristic parameter set;
c. further judging the falling state of the human body through a real-time heart rate difference value characteristic parameter delta HRV in the characteristic parameter set;
d. through the synthetic acceleration vector amplitude feature parameter SVM in the feature parameter set k And synthesizing the extreme difference characteristic parameter delta SVM of the acceleration vector amplitude R And the differential acceleration vector amplitude absolute average characteristic parameter DSVM further judges the falling state of the human body;
e. characteristic parameter SVM by synthetic acceleration vector amplitude standard deviation in characteristic parameter set sd And further judging the falling state of the human body to obtain whether the human body is in the falling state at present.
The second objective of the present invention is to provide a dynamic human body fall monitoring system based on multi-dimensional characteristic parameters, so as to solve the technical problems of high cost, low efficiency and low accuracy due to environmental impact, space and environment in the current human body fall detection method.
The second purpose of the invention is realized by the following technical means: a dynamic human body falling monitoring system based on multi-dimensional characteristic parameters adopts a wearable mode to construct a human body falling monitoring mechanism module, a human body falling decision characteristic threshold value model module, a human body falling decision characteristic threshold value correction module and a human body falling judgment and decision module, wherein the human body falling monitoring mechanism module is used for constructing a human body dynamic three-dimensional space monitoring mechanism and a biomedical vital sign heart rate monitoring mechanism; the human body falling decision characteristic threshold value model building module is used for building a human body falling decision characteristic threshold value parameter set convolution neural network learning model, extracting characteristic parameters of data acquired by an inertial sensor and a biosensor which are composed of an acceleration and a gyroscope to form a characteristic parameter set, and importing the characteristic parameter set into the human body falling decision characteristic threshold value parameter set convolution neural network learning model for deep learning, training, analyzing and deciding to obtain a human body falling decision characteristic threshold value initial parameter set; the human body falling decision characteristic threshold value correction module is used for correcting the initial parameter set of the human body falling decision characteristic threshold value through the human body falling decision characteristic threshold value parameter set convolutional neural network learning model according to the self characteristics of different users to obtain a user falling decision real-time characteristic threshold value parameter set; the human body falling judgment and decision module is used for judging and deciding whether the activity behavior state of the user falls in real time according to the real-time characteristic threshold value parameter set of the user falling decision, and outputting a falling state result.
The invention has the beneficial effects that: the invention not only utilizes the wearable human body activity behavior identification technology, but also utilizes the human body biomedical technology to monitor the human body falling behavior state, thereby solving the problem that the current human body falling detection method is limited by environment, space, high cost, low efficiency and accuracy and the like; the multi-dimensional falling decision characteristic parameters are introduced, so that the problems of low accuracy such as falling false alarm, missing alarm and the like caused by few decisive elements, simple algorithm and single in the current human body falling monitoring are solved; the invention can also utilize a positioning system and a wireless communication system to solve the problem that the information is not transmitted to the associated service or monitoring platform, mechanism and relatives and friends timely or comprehensively after the human body is dynamically monitored to fall.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the structures shown in the drawings without creative efforts.
FIG. 1 is a flow chart of the present invention;
fig. 2 is a framework diagram of a human fall monitoring logical structure;
fig. 3 is a graph of a convolutional neural network learning model of a human fall decision characteristic threshold parameter set;
FIG. 4 is a circuit frame diagram of a dynamic monitoring device for human body falling;
FIG. 5 is a flow chart of a human fall determination and decision making process;
FIG. 6 is a system framework diagram of the present invention;
fig. 7 is a flow chart of a dynamic monitoring device for human body falling.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. The components of embodiments of the present invention generally described and illustrated in the figures herein may be arranged and designed in a wide variety of different configurations.
It should be noted that: like reference numbers and letters refer to like items in the following figures, and thus, once an item is defined in one figure, it need not be further defined and explained in subsequent figures.
Some embodiments of the invention are described in detail below with reference to the accompanying drawings. The embodiments described below and the features of the embodiments can be combined with each other without conflict.
Example 1:
referring to fig. 1 to 3, a method for dynamically monitoring human body falling based on multi-dimensional characteristic parameters includes the following steps:
a human body falling monitoring mechanism is constructed by an inertial sensor and a biosensor consisting of an acceleration instrument and a gyroscope, and comprises a human body dynamic three-dimensional space monitoring mechanism and a biomedical vital sign heart rate monitoring mechanism;
constructing a human body falling decision characteristic threshold value parameter set convolutional neural network learning model, extracting characteristic parameters of data acquired by an inertial sensor and a biosensor consisting of an acceleration and a gyroscope to form a characteristic parameter set, and introducing the characteristic parameter set into the human body falling decision characteristic threshold value parameter set convolutional neural network learning model for deep learning, training, analyzing and deciding to obtain a human body falling decision characteristic threshold value initial parameter set;
correcting the initial parameter set of the human body falling decision characteristic threshold value through a human body falling decision characteristic threshold value parameter set convolutional neural network learning model according to the self characteristics of different users to obtain a user falling decision real-time characteristic threshold value parameter set;
and judging and deciding whether the activity behavior state of the user falls or not in real time according to the real-time characteristic threshold value parameter set for the falling decision of the user, and outputting a falling state result.
In this embodiment, the dynamic three-dimensional space monitoring mechanism of the human body respectively corresponds three spatial coordinates of x, y, and z of an inertial sensor composed of an acceleration and a gyroscope to the dynamic three-dimensional space coordinates of the human body. An x axis of an inertial sensor consisting of an acceleration instrument and a gyroscope represents the change of the moving acceleration of the human body in the left and right directions; an inertial sensor y axis formed by the acceleration and the gyroscope represents the movement acceleration change of the human body space in the front-back direction; an inertial sensor z axis formed by an acceleration instrument and a gyroscope represents the change of the gravity acceleration in the vertical direction of the human body space; the angular change of the z-axis of an inertial sensor consisting of an acceleration instrument and a gyroscope relative to the vertical direction represents the change of the attitude angular velocity of the inclination, rotation or offset of the human body.
In this embodiment, the biomedical vital signs heart rate monitoring mechanism corresponds the PPG heart rate biosensor to the human vital signs heart rate value parameter state.
Referring to fig. 3, the core algorithm of the invention is to construct a human body falling decision characteristic threshold parameter set convolutional neural network learning model, firstly, feature parameters of the known human body non-falling daily life activity behavior and the physical quantity acceleration and angular velocity of the human body falling activity behavior and the human body vital sign heart rate value data are extracted, and the feature parameters of the known human body non-falling daily life activity behavior and the feature parameters of the falling activity behavior are comprehensively analyzed and compared, and a human body falling decision characteristic threshold value initial parameter set is obtained through automatic learning, training, analysis and decision of the human body falling decision characteristic threshold value parameter set convolutional neural network learning model. Because individuals have different activities of daily life due to age, height, weight, mental state and agility of activity, the initial parameter set of the human body falling decision characteristic threshold value needs to be corrected through the human body falling decision characteristic threshold value parameter set convolutional neural network learning model according to the characteristics of the individuals when the individuals actually monitor the falling of the human body, and the human body falling decision real-time characteristic threshold value parameter set of the individuals is established. Therefore, more accurate and efficient judgment and decision of fall monitoring can be realized.
In this embodiment, the method for constructing the convolutional neural network learning model of the human fall decision characteristic threshold parameter set includes: and constructing a human body falling decision characteristic threshold value parameter set convolutional neural network learning model through a non-falling daily life activity database and a falling activity behavior database. Further, the non-falling daily life activity and falling activity database is established by acquiring a large amount of data of the non-falling daily life activity and falling activity of the human body through an inertial sensor composed of an acceleration instrument and a gyroscope, a PPG heart rate biosensor and the like, and the non-falling daily life activity of the human body at least comprises walking, running, sitting, standing, squatting, bending, lying, climbing stairs, descending stairs and the like.
In this embodiment, the method for extracting characteristic parameters of data acquired by an inertial sensor and a biosensor, which are composed of an acceleration sensor and a gyroscope, includes: extracting characteristic parameters from activity behavior acceleration data, angular velocity data and vital sign heart rate value data (see figure 2) acquired from human body non-falling daily activity activities and human body falling activity activities to form a characteristic parameter set, wherein the characteristic parameter set comprises a movement acceleration component difference characteristic parameter delta g x And Δ g y Characteristic parameter delta g of gravity acceleration component difference z Attitude angular velocity difference characteristicParameter delta theta, synthetic acceleration vector amplitude characteristic parameter SVM k And synthesizing the extreme difference characteristic parameter delta SVM of the acceleration vector amplitude R Differential acceleration vector amplitude absolute mean characteristic parameter DSVM and synthetic acceleration vector amplitude standard deviation characteristic parameter SVM sd And a real-time heart rate difference characteristic parameter delta HRV. Further, extracting characteristic parameters: firstly, acquiring acceleration data and angular velocity data by an inertial sensor, wherein the acceleration data and the angular velocity data comprise human body movement acceleration component data (g) of human body activity x And g y ) Data g of acceleration of gravity of human body z The data information such as the human body rotation and deviation attitude angular velocity data (theta represents the angle change of the human body relative to the vertical direction) and interference noise and the like adopts a three-order moving average filtering method to preprocess the original acceleration data and the angular velocity data, so that a good smoothing effect is obtained, the noise is reduced, and meanwhile, the human body activity behavior data contained in the original data is reserved. Then extracting the difference value delta g of the moving component of the moving and the gravity acceleration from the three-dimensional space acceleration data and the angular velocity data R x ,Δg R y ,Δg R z (ii) a An attitude angular velocity pole difference value delta theta; composite acceleration vector amplitude SVM k Synthesizing the acceleration vector amplitude pole difference value Delta SVM R Differential synthesis acceleration vector amplitude absolute mean value DSVM and synthesis acceleration vector amplitude standard deviation SVM sd And taking the characteristic parameters such as the real-time heart difference value delta HRV and the like as decision-making characteristic parameters for monitoring the human body falling.
Further, the specific implementation method for extracting the feature parameters is as follows: (1) Characteristic parameter delta g of moving acceleration component difference x =Δg x(k) -Δg xAVE(t-1) ,Δg y =Δg y(k) -Δg yAVE(t-1) (ii) a Representing the difference between the acceleration of the x and y axes at the current moment k in the sliding window period t and the average value in the previous window period (t-1); reflecting the moving space change conditions of the activities of the daily life of the human body to the left or the right and to the front or the back, if the value is larger, the moving space change of the activities of the human body is larger; (2) Gravity acceleration component difference characteristic parameter delta g z =Δg z(k)gzAVE(t-1) The difference between the acceleration of the z axis at the current k moment in the sliding window t period and the average value in the previous window (t-1) time period is represented; the degree of change of activities such as walking, squatting, sitting, climbing up and down steps and falling of the human body in the vertical direction is reflected, and if the value is larger, the larger the activity weightlessness or overweight change of the human body in the vertical space direction is indicated. (3) Synthetic acceleration vector magnitude feature parameters
Figure BDA0003835010550000091
Wherein the SVM k Representing the vector amplitude of the synthesized acceleration signal at the current k moment in the sliding window t period; wherein g is xk 、g yk 、g zk Respectively representing the acceleration in the x, y and z axis directions at the k moment; if the value is larger, the human body is more active. (4) Synthetic acceleration vector amplitude extreme difference characteristic parameter delta SVM R =SVM MAX -SVM MIN (ii) a Wherein the SVM MAX And SVM MIN Representing the maximum value and the minimum value of the amplitude of the synthetic acceleration vector of the human body activity in the sliding window t period; representing the difference between the maximum value and the minimum value of the amplitude of the synthesized acceleration vector in the sliding window t period; reflects the intensity of the human body activity, and if the value is larger, the human body activity is more intense. (5) Absolute mean characteristic parameter of differential acceleration vector amplitude
Figure BDA0003835010550000101
The degree of human body activity is represented, the intensity of activity state change in the human body sliding window t time period is reflected, and if the value is larger, the intensity of human body activity is indicated to be more intense. (6) Synthetic acceleration vector magnitude standard deviation characteristic parameter
Figure BDA0003835010550000102
Figure BDA0003835010550000103
Representing the standard deviation of the amplitude of the synthetic acceleration vector in the t period of the sliding window when the human body moves, wherein the mean value of the amplitude of the synthetic acceleration vector SVM ave represents the synthetic acceleration in the t period of the sliding window when the human body movesThe average of the magnitude of the degree vector. The characteristic parameter of the synthesized acceleration vector amplitude standard deviation reflects the intensity of state change when the human body moves, and if the value is smaller, the motion change is more intense. (7) Attitude angular velocity difference characteristic parameter delta theta = theta MAX –θ MIN Wherein
Figure BDA0003835010550000104
Figure BDA0003835010550000105
The angle change size relative to the vertical direction in the sliding window t period in the human body activity process is represented, and the angle change of the human body relative to the vertical direction, namely the degree of the body inclination, rotation or deviation, is reflected, and if the value is larger, the larger the body inclination, rotation or deviation is indicated. (8) Human vital sign heart rate data characteristic parameter delta HRV = HRV (K) –HRV (t-1) The method is characterized by representing the difference value between the real-time heart rate at the current k moment and the real-time heart rate at the previous moment (t-1) in a sliding window t period in the activity process of a human body, reflecting the panic and excitement emotion of the body, leading to the enhancement of nervous excitement, wherein the larger the value is, the more violent the activity change is, and the larger the change of the heart rate value is.
The time length of the sliding window of the method can be self-adaptive, and the overlapping rate of the window is 50 percent; i.e. the last 50% of the data of the current window as the first 50% of the data of the following window. The time length of the sliding window can be adjusted according to the characteristic parameter of the amplitude standard deviation of the synthesized acceleration vector, namely the SVM (support vector machine) of the amplitude standard deviation of the synthesized acceleration vector of the sliding window in the current t time period sd(t) And (t + 1) time period sliding window synthetic acceleration vector amplitude standard deviation SVM s d(t+1) The ratio therebetween; delta SVM s d =SVM s d(t) -SVM s d(t+1) (ii) a Such as a delta SVM sd Not less than 0; keeping the time length of the previous sliding window; and conversely, reducing the time length by half as the time length of the sliding window of the next time period.
In the present embodiment, after extracting the characteristic parameters of the data collected by the acceleration, the inertial sensor composed of a gyroscope, and the biosensor, the characteristic parameters need to be extractedImporting the sign parameter set into a human body falling decision characteristic threshold parameter set convolutional neural network learning model for deep learning, training, analyzing and deciding to obtain a human body falling decision characteristic threshold initial parameter set, wherein the human body falling decision characteristic threshold initial parameter set comprises delta g x Fall-TH Threshold value characteristic parameter, Δ g y Fall-TH Characteristic threshold value parameter, Δ g z Fall-TH Characteristic threshold parameter, Δ θ Fall-TH Characteristic threshold value parameter, SVM k Fall-TH Characteristic threshold value parameter, delta SVM R Fall-TH Characteristic threshold parameter, DSVM Fall-TH Characteristic threshold value parameter, SVM s d Fall-TH Characteristic threshold parameter, Δ HRV Fall-TH A characteristic threshold parameter.
In this embodiment, the method for obtaining the user fall decision real-time characteristic threshold parameter set by modifying the human fall decision characteristic threshold initial parameter set through the human fall decision characteristic threshold parameter set convolutional neural network learning model according to the characteristics of different users comprises: different users lead the activity behavior acceleration data, the angular velocity data and the vital sign heart rate value data collected by non-falling daily life activity and falling activity behaviors into a human falling decision characteristic threshold value parameter set convolutional neural network learning model for learning, training, analyzing and correcting according to the characteristics of the users, so that a user falling decision real-time characteristic threshold value parameter set is obtained and replaces a human falling decision characteristic threshold value initial parameter set; the user falling decision real-time characteristic threshold value parameter set comprises' delta g x Fall-TH Characteristic threshold parameter,' Δ g y Fall-TH Characteristic threshold parameter,' Δ g z Fall-TH Characteristic threshold parameter,' Delta theta Fall-TH Characteristic threshold parameter,' SVM k Fall-TH Characteristic threshold parameter,' Δ SVM RFall-TH Characteristic threshold parameter,' DSVM Fall-TH Characteristic threshold parameter,' SVM s d Fall-TH Characteristic threshold parameter,' Δ HRV Fall-TH A characteristic threshold parameter. Furthermore, the method for correcting the characteristic threshold value parameter set of the human body falling decision characteristic according to the characteristics of the user per se is used for walking, running, sitting, standing, squatting, bending, lying on back, lying down and lying downThe operation of 'climbing stairs' and 'descending stairs' is carried out, the relevant human activity behavior data is collected, learning, training, analysis and correction are carried out according to the human body falling decision characteristic threshold value parameter set convolutional neural network learning model, and the user self falling decision real-time characteristic threshold value parameter set is determined.
The invention constructs a human body falling monitoring judgment and decision model based on characteristic parameters and characteristic threshold value parameters, wherein the human body falling judgment and decision model is a hierarchical superposition progressive algorithm, each level of algorithm independently carries out falling judgment and decision once, and the latter level of algorithm only carries out calculation, judgment and decision on the judged and decided state of the former level; the human body falling monitoring and judging and decision model construction characteristic parameter set and the human body falling decision characteristic threshold value parameter set (refer to fig. 5). Specifically, the real-time judgment and decision of the activity behavior state of whether the user falls or not according to the real-time characteristic threshold value parameter set for the user fall decision comprises the following steps:
a. firstly, the characteristic parameter delta g is obtained through the difference value of the moving acceleration components in the characteristic parameter set x And Δ g y And a characteristic parameter Δ g of the gravity acceleration component difference z Judging the falling state of the human body; because the body of the human body loses balance and inclines before falling over, the physical quantity of the human body is obviously changed in moving acceleration and gravitational acceleration, and therefore when delta g is used, the body is not inclined, and the body is not inclined when falling over x Or Δ g y Is less than or equal to' Δ g x Fall-TH Threshold sum of y Fall-TH Threshold value, Δ g z Greater than' Δ g z Fall-TH When the threshold value is reached, the decision is that the human body is in a non-falling state, otherwise, the human body is considered to fall, and if the human body is accurately judged to be in the falling state, further judgment and decision are needed.
b. Further judging the falling state of the human body through the attitude angular velocity difference characteristic parameter delta theta in the characteristic parameter set; because the rotation and movement angles of the body will be changed violently after the body tilts in the falling process of the human body, the angle of the body posture relative to the vertical direction will be changed by about 90 degrees, which is reflected in that the posture angular velocity of the physical quantity gyroscope will be changed remarkably, and therefore, when delta theta is less than or equal to' delta theta Fall-TH And when the threshold value is reached, the decision is that the human body is in a non-falling state, otherwise, the human body is considered to fall, and if the judgment on the falling state is accurate, further judgment and decision are needed.
c. Secondly, further judging the falling state of the human body through a real-time heart rate difference value characteristic parameter delta HRV in the characteristic parameter set; because the autonomic nervous system of the human body has obvious regulation and control effects on the heart rate, when the human body is in motion, in a panic or excited, the excitability of sympathetic nerves is enhanced, and the heart rate is accelerated to different degrees. The method is characterized in that the heart rate value of vital signs of a human body is drastically changed, namely, when the human body falls down, the state is sudden and occurs under the non-self-help condition, the human body generates panic exciting mood, the excitability of sympathetic nerves is enhanced, the heart rate value is rapidly increased, after the human body falls down stably, the excitability of vagus nerves is enhanced, and the heart rate value is rapidly reduced. Therefore, when Δ HRV is less than or equal to' Δ HRV Fall-TH And (4) threshold value, wherein the decision is that the human body is in a non-falling state, otherwise, the human body is considered to fall, and if the falling state is accurately judged, further judgment and decision are needed.
d. Secondly, the characteristic parameter SVM is determined by the composite acceleration vector amplitude value in the characteristic parameter set k Synthetic acceleration vector amplitude extreme difference characteristic parameter delta SVM R And the differential acceleration vector amplitude absolute average characteristic parameter DSVM further judges the falling state of the human body; the falling is realized after the gravity center is unbalanced and inclined, and the falling time is short, so that the human body is strongly impacted with the ground after falling, and is reflected in that a plurality of peak values, valley values and maximum peak values appear in the amplitude of the synthesized acceleration vector under the interaction of force; therefore, when SVM k Less than or equal to' SVM Fall-TH Threshold value, and Δ SVM R Less than or equal to' Δ SVM R Fall-TH A threshold value, and DSVM is less than or equal to' DSVM Fall-TH And (4) threshold value, wherein the decision is that the human body is in a non-falling state, otherwise, the human body is considered to fall, and if the falling state is accurately judged, further judgment and decision are needed.
e. When the human body is in the active state deltag x 、Δg y 、Δg z 、Δθ、ΔHRV、SVM MAX 、ΔSVM R DSVM characteristic parameters meet the real-time characteristic threshold value parameter delta g of user falling decision x Fall-TH 、′Δg y Fall-TH 、′Δg z Fall-TH 、′Δθ Fall-TH 、′ΔHRV Fall-TH 、′SVM K Fall-TH 、′ΔSVM RFall-TH 、′DSVM Fall-TH After the judgment and decision conditions are met, the SVM is finally passed s d The characteristic parameters further judge the falling state, and after the human body falls, the human body can be in a stationary and stable state for a period of time, so that the acceleration curve can not fluctuate violently and tends to be stable, and the characteristic parameters are reflected in that the standard deviation of the acceleration vector amplitude is minimum; therefore, when SVM s d Greater than or equal to' SVM s d Fall-TH And in the threshold, the decision is that the human body is in an active state, and the current human body is judged to be in a non-falling state, otherwise, the decision is that the human body is in a falling state.
Referring to fig. 5, based on the same inventive concept, the invention provides a dynamic human body falling monitoring system based on multidimensional characteristic parameters, which is used for realizing the above dynamic human body falling monitoring method based on multidimensional characteristic parameters, and comprises a human body falling monitoring mechanism building module, a human body falling decision characteristic threshold value model building module, a human body falling decision characteristic threshold value correcting module and a human body falling judgment and decision module, wherein the human body falling monitoring mechanism building module is used for building a human body dynamic three-dimensional space monitoring mechanism and a biomedical vital sign heart rate monitoring mechanism; the human body falling decision characteristic threshold value model building module is used for building a human body falling decision characteristic threshold value parameter set convolutional neural network learning model, extracting characteristic parameters of data acquired by an inertial sensor and a biosensor which are composed of an acceleration and a gyroscope to form a characteristic parameter set, and introducing the characteristic parameter set into the human body falling decision characteristic threshold value parameter set convolutional neural network learning model for deep learning, training, analyzing and deciding to obtain a human body falling decision characteristic threshold value initial parameter set; the human body falling decision characteristic threshold value correction module is used for correcting the initial parameter set of the human body falling decision characteristic threshold value through the human body falling decision characteristic threshold value parameter set convolutional neural network learning model according to the self characteristics of different users to obtain a user falling decision real-time characteristic threshold value parameter set; the decision module is used for judging and deciding whether the activity behavior state of the user falls down in real time according to the real-time characteristic threshold value parameter set for the user fall decision, and outputting the fall state result.
Further, referring to fig. 4, the invention also provides a dynamic monitoring device for human body falling, which comprises a system microprocessor control unit, a human body activity and behavior data monitoring unit, a human body vital sign and heart rate data monitoring unit, a navigation and positioning system, a wireless communication system, a falling alarm system and a system power supply. The system microprocessor control unit controls and manages all functional modules of the dynamic human body falling monitoring system based on the multi-dimensional characteristic parameters. Specifically, the working principle of the dynamic monitoring device for human body falling (see fig. 7) is as follows: after the inertial sensor and the PPG heart rate biosensor are adopted to obtain human activity behavior data and human vital sign heart rate data, whether a human body falls down is judged and decided through a dynamic human body falling monitoring method of multidimensional characteristic parameters, once the human body falls down is judged and decided, the positioning system is automatically triggered, after geographical position information is obtained, the wireless communication system (one or more of 3G, 4G, 5G, bluetooth, wi-Fi and the like) and voice and light alarm are automatically triggered, the human body falling information including the geographical position information during falling down and basic vital sign dynamic heart rate value data information after falling down are timely sent to a related service or monitoring platform, a mechanism and relatives, information support of a rescue base and scheme is improved for the related service or monitoring platform, the mechanism and the relatives, and loss caused by falling down is reduced as much as possible. Preferably, the dynamic monitoring device for human body falling is suitable for wearable equipment, and can be suitable for wearable equipment worn on wrists, wearable equipment worn on breasts, and wearable equipment worn on heads.
The invention not only utilizes the recognition technology of human body behaviors, but also utilizes the biomedical technology of the human body to monitor the falling state of the human body, thereby solving the problems that the current human body falling detection method is limited by environment, space, high cost, low efficiency and accuracy and the like; the invention introduces multi-dimensional fall decision characteristic parameters, and solves the problems of low accuracy such as fall false alarm, false alarm and the like caused by few decisive elements, simple algorithm and single in current human fall monitoring. The device of the invention utilizes the positioning system and the wireless communication system to solve the problem that the information transmission to the associated service or monitoring platform, mechanism and relatives and friends is not timely or comprehensive after the falling of the human body is dynamically monitored. The invention constructs a dynamic three-dimensional (x, y and z axis) space coordinate system of a human body, an inertial sensor consisting of an acceleration and a gyroscope is adopted to obtain the activity behavior data of the human body, and the multi-dimensional falling decision characteristic parameters of the human body represented by the three-dimensional space coordinate system are extracted from the activity behavior data of the human body, wherein the multi-dimensional falling decision characteristic parameters comprise a moving acceleration component, a gravity acceleration component, a synthesized acceleration vector amplitude, a differential synthesized acceleration vector amplitude absolute average value, a synthesized acceleration vector amplitude range value, a synthesized acceleration vector amplitude standard deviation, an attitude angle, an angular velocity and other characteristic parameters; the method adopts a PPG heart rate biosensor to acquire real-time, dynamic and continuous vital sign heart rate value HRV data of a human body, and extracts characteristic parameters such as a real-time heart rate difference value and the like from the vital sign heart rate value HRV data. According to the characteristic parameters, a human body falling decision characteristic threshold value parameter set is established by establishing a human body falling decision characteristic threshold value parameter set convolutional neural network learning model, and finally, the falling and non-falling states of the behaviors are judged and decided through daily life activity data and the human body falling decision characteristic threshold value parameter set. The invention finally corrects the human body falling decision characteristic threshold value parameter set according to the self characteristics (height, weight, life habit, spirit and the like) of the user, provides powerful basis for accurately and quickly deciding non-falling monitoring such as falling, fast sitting, fast squatting and the like, and efficiently and accurately provides detection and monitoring of falling and non-falling states and results. The detection and monitoring device implemented by the method has the advantages of low cost, easy mobility (no limitation of a space area), feasibility and comfort. The invention can be implemented in a wearable device mode and has wide application prospect.
The invention adopts an inertial sensor and a biosensor to obtain human activity behavior and vital sign heart rate data, and extracts real-time characteristic parameters including movement and gravity acceleration components, a synthetic acceleration vector amplitude, a differential synthetic acceleration vector amplitude absolute average value, a synthetic acceleration vector amplitude extreme value and standard deviation and an attitude angular velocity extreme value from the human activity behavior and heart rate data in a sliding window mode; real-time heart rate difference and other characteristic parameters; and (3) importing the real-time characteristic parameters into a human body falling decision characteristic threshold parameter set convolutional neural network learning model, correcting the human body falling decision characteristic threshold parameter set, then carrying out judgment and decision making, outputting a falling result, and transmitting the information to an association party in time to reduce loss as much as possible.
It should be noted that, for the sake of simplicity, the foregoing embodiments are described as a series of combinations of acts, but those skilled in the art should understand that the present application is not limited by the described order of acts, as some steps may be performed in other orders or simultaneously according to the present application. Further, those skilled in the art will recognize that the embodiments described in this specification are preferred embodiments and that no acts are necessarily required of the application.
In the above embodiments, the basic principle and the main features of the present invention and the advantages of the present invention are described. It will be understood by those skilled in the art that the present invention is not limited to the embodiments described above, which are merely illustrative of the principles of the invention, and that modifications and variations can be made by one skilled in the art without departing from the spirit and scope of the invention as defined by the appended claims.

Claims (9)

1. A dynamic human body falling monitoring method based on multi-dimensional characteristic parameters is characterized by comprising the following steps:
a human body falling monitoring mechanism is constructed by an inertial sensor and a biosensor consisting of an acceleration instrument and a gyroscope, and comprises a human body dynamic three-dimensional space monitoring mechanism and a biomedical vital sign heart rate monitoring mechanism;
constructing a human body falling decision characteristic threshold value parameter set convolutional neural network learning model, extracting characteristic parameters of data acquired by an inertial sensor and a biosensor consisting of an acceleration and a gyroscope to form a characteristic parameter set, and introducing the characteristic parameter set into the human body falling decision characteristic threshold value parameter set convolutional neural network learning model for deep learning, training, analyzing and deciding to obtain a human body falling decision characteristic threshold value initial parameter set;
correcting the initial parameter set of the human body falling decision characteristic threshold value through a human body falling decision characteristic threshold value parameter set convolutional neural network learning model according to the self characteristics of different users to obtain a user falling decision real-time characteristic threshold value parameter set;
and judging and deciding whether the activity behavior state of the user falls or not in real time according to the real-time characteristic threshold value parameter set for the falling decision of the user, and outputting a falling state result.
2. The method for dynamically monitoring the falling of the human body based on the multidimensional characteristic parameters as claimed in claim 1, wherein the mechanism for dynamically monitoring the three-dimensional space of the human body respectively corresponds three space coordinates of x, y and z of an inertial sensor consisting of an acceleration sensor and a gyroscope to the dynamic three-dimensional space coordinates of the human body, and an x axis of the inertial sensor consisting of the acceleration sensor and the gyroscope represents the change of the moving acceleration of the human body in the left-right direction; an inertial sensor y axis formed by the acceleration and the gyroscope represents the movement acceleration change of the human body space in the front-back direction; an inertial sensor z axis formed by the acceleration and the gyroscope represents the change of the gravitational acceleration in the vertical direction of the human body space; the angular change of the z-axis of the inertial sensor consisting of the acceleration and the gyroscope relative to the vertical direction represents the change of the attitude angular velocity of the inclination, rotation or deviation of the human body.
3. The dynamic human fall monitoring method based on multi-dimensional feature parameters as claimed in claim 1, wherein the human biomedical vital signs heart rate monitoring mechanism corresponds PPG heart rate biosensors to human vital signs heart rate value parameter states.
4. A dynamic human fall monitoring method based on multidimensional characteristic parameters as claimed in claim 1, wherein the method for constructing the convolutional neural network learning model based on the characteristic threshold parameter set for human fall decision making is as follows: and constructing a human body falling decision characteristic threshold value parameter set convolutional neural network learning model through a non-falling daily life activity database and a falling activity behavior database.
5. The dynamic human fall monitoring method based on multidimensional characteristic parameters as claimed in claim 1, wherein the method for extracting the characteristic parameters of the data acquired by the inertial sensor and the biosensor consisting of an accelerometer and a gyroscope comprises: extracting characteristic parameters from activity behavior acceleration data, angular velocity data and vital sign heart rate data acquired from human body non-falling daily life activity and human body falling activity to form a characteristic parameter set, wherein the characteristic parameter set comprises a movement acceleration component difference characteristic parameter delta g x And Δ g y Characteristic parameter delta g of gravity acceleration component difference z Attitude angular velocity difference characteristic parameter delta theta and synthetic acceleration vector amplitude characteristic parameter SVM k Synthetic acceleration vector amplitude extreme difference characteristic parameter delta SVM R Differential acceleration vector amplitude absolute average value characteristic parameter DSVM and synthetic acceleration vector amplitude standard deviation characteristic parameter SVM sd And a real-time heart rate difference characteristic parameter delta HRV.
6. The dynamic human fall monitoring method based on the multidimensional characteristic parameters, as claimed in claim 5, wherein after extracting the characteristic parameters of data acquired by the inertial sensor and the biosensor comprising the accelerometer and the gyroscope, the characteristic parameter set needs to be imported into a human fall decision characteristic threshold parameter set convolutional neural network learning model for deep learning, training, analyzing and deciding to obtain a human fall decision characteristic threshold initial parameter set; the initial parameter set of the human body falling decision characteristic threshold value comprises delta g x Fall-TH Threshold value characteristic parameter, Δ g y Fall-TH Characteristic threshold parameter, Δ g z Fall-TH Characteristic threshold parameter, Δ θ Fall-TH Characteristic threshold value parameter, SVM k Fall-TH Characteristic threshold value parameter, delta SVM R Fall-TH Characteristic threshold parameter, DSVM Fall-TH Characteristic threshold value parameter, SVM sd Fall-TH Characteristic threshold parameter, Δ HRV Fall-TH A characteristic threshold parameter.
7. A dynamic human fall monitoring method based on multidimensional characteristic parameters as claimed in claim 1, wherein the initial parameter set of the human fall decision characteristic threshold is modified by the human fall decision characteristic threshold parameter convolutional neural network learning model according to the characteristics of different users themselves to obtain a real-time characteristic threshold parameter set of the user fall decision to replace the initial parameter set of the human fall decision characteristic threshold, and the method comprises: different users import activity behavior acceleration data, angular velocity data and vital sign heart rate data acquired by non-falling daily life activities and falling activities into a human body falling decision characteristic threshold value parameter set convolutional neural network learning model for learning, training, analyzing and modifying according to own characteristics to obtain a user falling decision real-time characteristic threshold value parameter set and replace a human body falling decision characteristic threshold value initial parameter set; the user fall decision real-time characteristic threshold parameter set comprises' delta g x Fall-TH Characteristic threshold parameter,' Δ g y Fall-TH Characteristic threshold parameter,' Δ g z Fall-TH Characteristic threshold parameter,' Delta theta Fall-TH Characteristic threshold parameter,' SVM k Fall-TH Characteristic threshold parameter,' Δ SVM R Fall-TH Characteristic threshold parameter,' DSVM Fall-TH Characteristic threshold parameter,' SVM sd Fall-TH Characteristic threshold parameter,' Δ HRV Fall-TH A characteristic threshold parameter.
8. The dynamic human body fall monitoring method based on the multidimensional characteristic parameters as claimed in claim 1, wherein the method for judging and deciding whether the user falls in real time according to the real-time characteristic threshold parameter set for the fall decision of the user comprises the following steps:
a. the characteristic parameter delta g is obtained by the difference value of the moving acceleration components in the characteristic parameter set x And Δ g y And a characteristic parameter Δ g of the gravity acceleration component difference z Judging the falling state of the human body;
b. further judging the falling state of the human body according to the attitude angular velocity difference characteristic parameter delta theta in the characteristic parameter set;
c. further judging the falling state of the human body through a real-time heart rate difference value characteristic parameter delta HRV in the characteristic parameter set;
d. SVM (support vector machine) by using synthesized acceleration vector amplitude characteristic parameters in characteristic parameter set k And synthesizing the extreme difference characteristic parameter delta SVM of the acceleration vector amplitude R And the differential acceleration vector amplitude absolute average characteristic parameter DSVM further judges the falling state of the human body;
e. and further judging the falling state of the human body through a synthetic acceleration vector amplitude standard deviation characteristic parameter SVM sd in the characteristic parameter set to obtain whether the human body is in the falling state at present.
9. A human body falling dynamic monitoring system based on multi-dimensional characteristic parameters is used for realizing the human body falling dynamic monitoring method based on the multi-dimensional characteristic parameters, which is characterized by comprising a human body falling monitoring mechanism building module, a human body falling decision characteristic threshold value model building module, a human body falling decision characteristic threshold value correction module and a human body falling judgment and decision module, wherein the human body falling monitoring mechanism building module is used for building a human body dynamic three-dimensional space monitoring mechanism and a biomedical vital sign heart rate monitoring mechanism; the human body falling decision characteristic threshold value model building module is used for building a human body falling decision characteristic threshold value parameter set convolutional neural network learning model, extracting characteristic parameters of data acquired by an inertial sensor and a biosensor which are composed of an acceleration and a gyroscope to form a characteristic parameter set, and introducing the characteristic parameter set into the human body falling decision characteristic threshold value parameter set convolutional neural network learning model for deep learning, training, analyzing and deciding to obtain a human body falling decision characteristic threshold value initial parameter set; the human body falling decision characteristic threshold value correcting module is used for correcting an initial parameter set of a human body falling decision characteristic threshold value through a human body falling decision characteristic threshold value parameter set convolutional neural network learning model according to the self characteristics of different users to obtain a user falling decision real-time characteristic threshold value parameter set; the human body falling judgment and decision module is used for judging and deciding whether the activity behavior state of the user falls in real time according to the real-time characteristic threshold value parameter set for the falling decision of the user, and outputting a falling state result.
CN202211085849.4A 2022-09-06 2022-09-06 Human body falling dynamic monitoring method and system based on multidimensional characteristic parameters Active CN115462782B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202211085849.4A CN115462782B (en) 2022-09-06 2022-09-06 Human body falling dynamic monitoring method and system based on multidimensional characteristic parameters

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202211085849.4A CN115462782B (en) 2022-09-06 2022-09-06 Human body falling dynamic monitoring method and system based on multidimensional characteristic parameters

Publications (2)

Publication Number Publication Date
CN115462782A true CN115462782A (en) 2022-12-13
CN115462782B CN115462782B (en) 2024-06-11

Family

ID=84370432

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202211085849.4A Active CN115462782B (en) 2022-09-06 2022-09-06 Human body falling dynamic monitoring method and system based on multidimensional characteristic parameters

Country Status (1)

Country Link
CN (1) CN115462782B (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117333929A (en) * 2023-12-01 2024-01-02 贵州省公路建设养护集团有限公司 Method and system for identifying abnormal personnel under road construction based on deep learning
CN118021294A (en) * 2024-04-11 2024-05-14 四川省铁路建设有限公司 Fall detection method and system based on multiple sensors

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2015009951A1 (en) * 2013-07-18 2015-01-22 Vital Connect, Inc. Fall detection using machine learning
CN105708470A (en) * 2016-01-21 2016-06-29 山东大学 Falling detection system and method based on combination of Doppler detector and sensor
CN106781278A (en) * 2017-02-04 2017-05-31 华南理工大学 A kind of fall detection method and device based on Fusion
CN108021888A (en) * 2017-12-05 2018-05-11 电子科技大学 A kind of fall detection method
CN111693024A (en) * 2020-06-11 2020-09-22 秦奕天 Wearable human body sensing monitoring equipment based on nine-axis inertia measurement unit
CN111931568A (en) * 2020-06-30 2020-11-13 青岛科技大学 Human body falling detection method and system based on enhanced learning

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2015009951A1 (en) * 2013-07-18 2015-01-22 Vital Connect, Inc. Fall detection using machine learning
CN105708470A (en) * 2016-01-21 2016-06-29 山东大学 Falling detection system and method based on combination of Doppler detector and sensor
CN106781278A (en) * 2017-02-04 2017-05-31 华南理工大学 A kind of fall detection method and device based on Fusion
CN108021888A (en) * 2017-12-05 2018-05-11 电子科技大学 A kind of fall detection method
CN111693024A (en) * 2020-06-11 2020-09-22 秦奕天 Wearable human body sensing monitoring equipment based on nine-axis inertia measurement unit
CN111931568A (en) * 2020-06-30 2020-11-13 青岛科技大学 Human body falling detection method and system based on enhanced learning

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117333929A (en) * 2023-12-01 2024-01-02 贵州省公路建设养护集团有限公司 Method and system for identifying abnormal personnel under road construction based on deep learning
CN117333929B (en) * 2023-12-01 2024-02-09 贵州省公路建设养护集团有限公司 Method and system for identifying abnormal personnel under road construction based on deep learning
CN118021294A (en) * 2024-04-11 2024-05-14 四川省铁路建设有限公司 Fall detection method and system based on multiple sensors

Also Published As

Publication number Publication date
CN115462782B (en) 2024-06-11

Similar Documents

Publication Publication Date Title
CN115462782B (en) Human body falling dynamic monitoring method and system based on multidimensional characteristic parameters
Dian et al. Wearables and the Internet of Things (IoT), applications, opportunities, and challenges: A Survey
US10319209B2 (en) Method and system for motion analysis and fall prevention
de la Concepción et al. Mobile activity recognition and fall detection system for elderly people using Ameva algorithm
Wang et al. Accelerometry based classification of walking patterns using time-frequency analysis
Nandy et al. Detailed human activity recognition using wearable sensor and smartphones
Scheurer et al. Human activity recognition for emergency first responders via body-worn inertial sensors
JP6268193B2 (en) Pulse wave measuring device, portable device, medical device system, and biological information communication system
Jafari et al. Physical activity monitoring for assisted living at home
JP5120795B2 (en) Human posture motion discrimination device and energy consumption calculation device
Jian et al. A portable fall detection and alerting system based on k-NN algorithm and remote medicine
KR101797854B1 (en) System and method for fall detection using smart band
JP7531917B2 (en) Heart Rate Measurement System
Shi et al. Fall Detection Algorithm Based on Triaxial Accelerometer and Magnetometer.
US20160242680A1 (en) Intelligent comfort level monitoring system
Rasheed et al. Evaluation of human activity recognition and fall detection using android phone
Zhao et al. Recognition of human fall events based on single tri-axial gyroscope
CN111183460A (en) Fall detector and improvement of fall detection
Nguyen et al. IMU-based spectrogram approach with deep convolutional neural networks for gait classification
US20240099639A1 (en) System for measuring heart rate
CN111166340A (en) Human body posture real-time identification method based on self-adaptive acceleration signal segmentation
Wójtowicz et al. Fall detector using discrete wavelet decomposition and SVM classifier
Cao et al. ActiRecognizer: Design and implementation of a real-time human activity recognition system
Chen et al. A Novel CNN-BiLSTM Ensemble Model With Attention Mechanism for Sit-to-Stand Phase Identification Using Wearable Inertial Sensors
CN111028488A (en) Intelligent sedentary reminding method based on PSO-SVM

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant