CN114997245A - Artificial intelligence-based anti-falling method and device - Google Patents
Artificial intelligence-based anti-falling method and device Download PDFInfo
- Publication number
- CN114997245A CN114997245A CN202210845138.6A CN202210845138A CN114997245A CN 114997245 A CN114997245 A CN 114997245A CN 202210845138 A CN202210845138 A CN 202210845138A CN 114997245 A CN114997245 A CN 114997245A
- Authority
- CN
- China
- Prior art keywords
- target
- falling
- time
- action
- real
- 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.)
- Pending
Links
Images
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/103—Detecting, measuring or recording devices for testing the shape, pattern, colour, size or movement of the body or parts thereof, for diagnostic purposes
- A61B5/11—Measuring movement of the entire body or parts thereof, e.g. head or hand tremor, mobility of a limb
- A61B5/1116—Determining posture transitions
- A61B5/1117—Fall detection
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/74—Details of notification to user or communication with user or patient ; user input means
- A61B5/746—Alarms related to a physiological condition, e.g. details of setting alarm thresholds or avoiding false alarms
-
- G—PHYSICS
- G08—SIGNALLING
- G08B—SIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
- G08B21/00—Alarms responsive to a single specified undesired or abnormal condition and not otherwise provided for
- G08B21/02—Alarms for ensuring the safety of persons
- G08B21/04—Alarms for ensuring the safety of persons responsive to non-activity, e.g. of elderly persons
- G08B21/0407—Alarms for ensuring the safety of persons responsive to non-activity, e.g. of elderly persons based on behaviour analysis
- G08B21/043—Alarms 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
Landscapes
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Animal Behavior & Ethology (AREA)
- Public Health (AREA)
- Pathology (AREA)
- Veterinary Medicine (AREA)
- Biomedical Technology (AREA)
- Heart & Thoracic Surgery (AREA)
- Medical Informatics (AREA)
- Molecular Biology (AREA)
- Surgery (AREA)
- Biophysics (AREA)
- Physiology (AREA)
- Psychiatry (AREA)
- Business, Economics & Management (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Psychology (AREA)
- Social Psychology (AREA)
- Gerontology & Geriatric Medicine (AREA)
- Dentistry (AREA)
- Emergency Management (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Signal Processing (AREA)
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
Abstract
The invention discloses an artificial intelligence-based anti-falling method and device, and relates to the field of data processing, wherein the method comprises the following steps: collecting target motion data records of a target user in real time; extracting a target resultant acceleration record, and drawing a target resultant acceleration waveform in real time; obtaining a target action characteristic set according to the characteristic analysis result; constructing an action recognition model according to the historical motion data set; analyzing the target action characteristic set through the action recognition model to obtain a target action type of a target user; and judging whether the target action type meets a preset type list or not, and performing real-time falling prevention early warning according to a judgment result. The problem of among the prior art to the accuracy of preventing the early warning of tumbleing low, and then cause to prevent falling the not good technique of protection effect is solved. The accuracy and the real-time performance of the anti-falling early warning are improved, the anti-falling early warning effect is further improved, and the technical effects of high-quality anti-falling protection and the like are achieved.
Description
Technical Field
The invention relates to the field of data processing, in particular to an artificial intelligence-based anti-falling method and device.
Background
A fall can occur when a person's body falls suddenly and involuntarily onto the ground or other surface. After the person falls down, the person can fear the movement in serious time, fear the movement, subconsciously reduce the movement times, lead to the decline of physical mechanism, and bring serious economic burden for the family. In addition, falls are one of the important factors in the accidental death of people, and falls dominate the factors in the accidental death of the elderly. How to effectively prevent falls has become a common concern for all members of society.
In the prior art, the early warning method for preventing falling has the technical problem that the early warning for preventing falling is low in accuracy, and therefore the falling prevention protection effect is poor.
Disclosure of Invention
The application provides a method and a device for preventing falling based on artificial intelligence, and solves the technical problem that in the prior art, the accuracy of early warning for preventing falling is low, and the falling prevention protection effect is poor.
In view of the above problems, the present application provides an artificial intelligence based fall prevention method and apparatus.
In a first aspect, the present application provides an artificial intelligence-based fall prevention method, where the method is applied to an artificial intelligence-based fall prevention device, and the method includes: acquiring a target motion data record of a target user in real time by using an inertial sensor; extracting a target resultant acceleration record in the target motion data record, wherein the target resultant acceleration record is record data with a time identifier; according to the target resultant acceleration record with the time identification, a target resultant acceleration waveform is drawn in real time; performing characteristic analysis on the target combined acceleration waveform, and obtaining a target action characteristic set according to a characteristic analysis result; acquiring a historical motion data set based on big data acquisition, and constructing an action recognition model according to the historical motion data set;
analyzing the target action characteristic set through the action recognition model to obtain a target action type of the target user; and judging whether the target action type meets a preset type list or not, and performing real-time falling prevention early warning according to a judgment result.
In a second aspect, the present application further provides an artificial intelligence based fall prevention device, wherein the device comprises: the acquisition module is used for acquiring a target motion data record of a target user in real time by using the inertial sensor; the extraction module is used for extracting a target resultant acceleration record in the target motion data record, wherein the target resultant acceleration record is record data with a time identifier; the drawing module is used for drawing a target combined acceleration waveform in real time according to the target combined acceleration record with the time identification; the characteristic analysis module is used for carrying out characteristic analysis on the target combined acceleration waveform and obtaining a target action characteristic set according to a characteristic analysis result; the building module is used for acquiring a historical motion data set based on big data and building an action recognition model according to the historical motion data set; the action analysis module is used for analyzing the target action characteristic set through the action recognition model to obtain a target action type of the target user; and the judging and early warning module is used for judging whether the target action type meets a preset type list or not and carrying out real-time falling prevention early warning according to a judgment result.
One or more technical solutions provided in the present application have at least the following technical effects or advantages:
acquiring a target motion data record of a target user in real time through an inertial sensor; extracting the target motion data record to obtain a target resultant acceleration record, wherein the target resultant acceleration record has a time identifier; according to the target resultant acceleration record with the time identification, drawing a target resultant acceleration waveform in real time; performing characteristic analysis on the target combined acceleration waveform to obtain a characteristic analysis result, and determining a target action characteristic set according to the characteristic analysis result; acquiring a historical motion data set through big data acquisition, and further constructing an action recognition model; analyzing the target action characteristic set through the action recognition model to obtain a target action type of the target user; and judging whether the target action type meets a preset type list or not, and performing real-time falling prevention early warning according to a judgment result. The accuracy and the real-time performance of the anti-falling early warning are improved, the anti-falling early warning effect is further improved, and high-quality anti-falling protection is realized; meanwhile, the intelligence of the anti-falling early warning is improved, and the scientific anti-falling technical effect is achieved.
Drawings
Fig. 1 is a schematic flow chart of an artificial intelligence-based fall prevention method according to the present application;
fig. 2 is a schematic flow chart of a real-time drawing of a target combined acceleration waveform in the artificial intelligence-based anti-falling method according to the present application;
fig. 3 is a schematic flow chart of fall prevention intervention in the artificial intelligence-based fall prevention method according to the present application;
fig. 4 is a schematic structural diagram of an anti-fall device based on artificial intelligence.
Description of reference numerals: the system comprises an acquisition module 11, an extraction module 12, a drawing module 13, a feature analysis module 14, a construction module 15, an action analysis module 16 and a judgment and early warning module 17.
Detailed Description
The application provides an artificial intelligence-based anti-falling method and device. The problem of among the prior art to the accuracy of preventing the early warning of tumbleing low, and then cause to prevent falling the not good technical problem of protection effect. The accuracy and the real-time performance of the anti-falling early warning are improved, the anti-falling early warning effect is further improved, and high-quality anti-falling protection is realized; meanwhile, the intelligence of the anti-falling early warning is improved, and the scientific anti-falling technical effect is achieved.
Example one
Referring to fig. 1, the present application provides an artificial intelligence based fall prevention method, where the method is applied to an artificial intelligence based fall prevention device, and the method specifically includes the following steps:
step S100: acquiring a target motion data record of a target user in real time by using an inertial sensor;
step S200: extracting a target resultant acceleration record in the target motion data record, wherein the target resultant acceleration record is record data with a time identifier;
specifically, real-time motion data acquisition is carried out on a target user through an inertial sensor, a target motion data record is obtained, information extraction is carried out on the target motion data record, and a target resultant acceleration record is obtained. The inertial sensor can be an inertial sensing device such as a six-axis inertial sensor, a nine-axis inertial sensor and the like in the prior art. The inertial sensor has the functions of acquiring real-time motion data of a target user and the like. The target user is any user who uses the artificial intelligence-based fall prevention device for intelligent fall prevention protection. For example, the target user may be an elderly person, a child, or the like. The target motion data records comprise data information such as motion time, X-axis acceleration, Y-axis acceleration, Z-axis acceleration and resultant acceleration, X-axis angular velocity, Y-axis angular velocity, Z-axis angular velocity, resultant angular velocity and attitude angle of a target user in a real-time motion process. The target resultant acceleration record comprises resultant acceleration data information in the target motion data record, and the target resultant acceleration record has a time identifier. Namely, the target resultant acceleration record comprises a time identifier and a target resultant acceleration. The target motion data record of the target user is collected, and the accurate target combined acceleration record is extracted by using the target motion data record, so that the technical effect of laying a foundation for subsequent intelligent falling prevention protection is achieved.
Step S300: according to the target resultant acceleration record with the time identification, a target resultant acceleration waveform is drawn in real time;
further, as shown in fig. 2, step S300 of the present application further includes:
step S310: taking the time mark in the target combined acceleration record as an abscissa and the target combined acceleration in the target combined acceleration record as an ordinate to obtain a target combined acceleration initial waveform;
step S320: and performing elongation processing of preset times on the ordinate in the target combined acceleration initial waveform to obtain the target combined acceleration waveform.
Specifically, the time mark and the target combined acceleration in the target combined acceleration record are input into drawing software such as MATLAB, and an initial waveform of the target combined acceleration with the time mark as an abscissa and the target combined acceleration as an ordinate is obtained. Further, the ordinate of the target combined acceleration initial waveform is subjected to stretching processing of preset times to obtain the target combined acceleration waveform. Wherein, the preset multiple is preset and determined by the fall prevention device based on artificial intelligence. For example, if the preset multiple may be 3 times, the ordinate of the target combined acceleration initial waveform is elongated by 3 times. The method achieves the technical effects that the target combined acceleration waveform which is easy to observe and high in visualization degree is determined by processing the target combined acceleration initial waveform elongation, and the accuracy of subsequent characteristic analysis on the target combined acceleration waveform is improved.
Step S400: performing characteristic analysis on the target combined acceleration waveform, and obtaining a target action characteristic set according to a characteristic analysis result;
specifically, a characteristic analysis result is obtained by performing characteristic analysis on the target combined acceleration waveform, and a target action characteristic set is determined according to the characteristic analysis result. The characteristic analysis result comprises waveform characteristic information of instantaneous values, wave crests, wave troughs and the like of the target combined acceleration waveform. And the target action characteristic set is a characteristic analysis result. The technical effects of determining the reliable target action characteristic set and providing data support for subsequently obtaining the target action type of the target user are achieved.
Step S500: acquiring a historical motion data set based on big data acquisition, and constructing an action recognition model according to the historical motion data set;
further, step S500 of the present application further includes:
step S510: dividing the historical movement data set into a falling movement data set and a daily movement data set;
step S520: obtaining motion data of a plurality of falling events according to the falling motion data set;
step S530: analyzing the motion data of the plurality of falling events in sequence to obtain a plurality of falling data analysis results;
step S540: establishing a falling feature-type mapping list between falling features and falling types according to the analysis results of the plurality of falling data;
step S550: constructing the action recognition model according to the falling feature-type mapping list;
step S560: and carrying out a verification experiment on the action recognition model according to the daily movement data set, and adjusting and optimizing the action recognition model according to an experiment result.
Specifically, a historical movement data set is obtained through big data acquisition, and data division of falling movement and daily movement is carried out on the historical movement data set to obtain a falling movement data set and a daily movement data set. Further, the movement data of a plurality of falling events are analyzed in sequence, a plurality of falling data analysis results are determined, a falling feature-type mapping list between falling features and falling types is built according to the falling data analysis results, and then an action recognition model is determined. And then, inputting the daily exercise data set into the action recognition model to perform a verification experiment, preventing the daily exercise from being mistaken for a falling action, obtaining an experiment result, and correcting, adjusting and optimizing the action recognition model by using the experiment result. Wherein the historical movement data set comprises historical movement data information of a plurality of target users. The falling motion data set comprises data information of falling motions such as forward falling, backward falling, and side-tipping falling in the historical motion data set. The motion data of the multiple falling events comprise data information of falling time, falling condition, falling degree and the like of the multiple falling events in the falling motion data set. The multiple falling data analysis results comprise falling instantaneous values such as falling types, falling combined accelerations, angular velocities and attitude angles corresponding to the motion data of multiple falling events, and falling characteristic information such as troughs and peaks of the falling combined accelerations. The falling feature-type mapping list is data information used for representing falling instantaneous values such as the combined acceleration, the angular velocity and the attitude angle of falling and corresponding relations between falling features such as troughs and wave crests of the combined acceleration of falling and falling types. The motion recognition model is a falling feature-type mapping list. The daily movement data set comprises data information of daily movements such as walking, jogging, jumping, going up and down stairs and the like in the historical movement data set. The experimental results include data information in the daily movement data set that was mistaken for a fall action. The action recognition model with high accuracy is obtained, the accuracy of the target action type of the target user obtained subsequently is improved, and the technical effect of high-quality anti-falling early warning is achieved.
Step S600: analyzing the target action characteristic set through the action recognition model to obtain a target action type of the target user;
further, step S600 of the present application further includes:
step S610: sequentially obtaining a plurality of target wave troughs within a preset time threshold according to the target resultant acceleration waveform, wherein the target wave troughs comprise real-time wave troughs;
step S620: comparing the multiple target wave troughs to obtain a target wave trough minimum value;
step S630: judging whether the minimum value of the target wave trough is the real-time wave trough or not through the action recognition model;
step S640: and if the minimum value of the target wave trough is not the real-time wave trough, the target action type is daily action.
Specifically, a plurality of target troughs within a preset time threshold are obtained based on the target combined acceleration waveform and the preset time threshold. Further, comparing a plurality of target wave troughs within a preset time threshold value to obtain the minimum value of the target wave troughs. And then inputting the minimum value of the target trough into the motion recognition model, judging whether the minimum value of the target trough is a real-time trough through the motion recognition model, and if the minimum value of the target trough is not the real-time trough, determining that the target motion type is daily motion. Wherein, the preset time threshold value can be determined by the user-defined setting of the falling prevention device based on artificial intelligence. The multiple target wave troughs comprise multiple wave troughs in a preset time threshold in a target combined acceleration waveform. And, the plurality of target troughs includes real-time troughs. The real-time wave trough refers to the wave trough closest to the real-time resultant acceleration. The minimum value of the target wave troughs is the deepest wave trough in the plurality of target wave troughs. In the target resultant acceleration waveform, a plurality of target wave troughs are a plurality of wave troughs appearing within a preset time threshold, the plurality of target wave troughs can reflect the weightlessness degree of a target user, the target user can have weightlessness with different degrees in the motion process, the deeper the wave trough is, namely the smaller the target resultant acceleration is, the more serious the weightlessness of the target user is, and the motion state of the target user tends to fall down. The method achieves the technical effects of obtaining the accurate target action type of the target user by utilizing the action recognition model and improving the accuracy of the follow-up real-time fall prevention early warning.
Further, step S640 of the present application further includes:
step S641: if the minimum value of the target wave trough is the real-time wave trough, the target action type is a fall-like action;
step S642: extracting a target attitude angle record in the target motion data record, wherein the target attitude angle record is record data with a time identifier;
step S643: sequentially obtaining an initial attitude angle and a real-time attitude angle of the target user according to the target attitude angle record with the time identification;
step S644: calculating the difference value between the initial attitude angle and the real-time attitude angle of the target user to obtain an attitude angle real-time difference value;
step S645: judging whether the real-time attitude angle difference value meets a preset attitude angle difference value threshold value or not;
step S646: and if the real-time attitude angle difference value meets a preset attitude angle difference value threshold, the falling-like action is a real falling action.
Specifically, if the target trough minimum value is a real-time trough, the motion state of the target user approaches falling, and the target action type is fall-like action. Further, the falling-like action is further judged through the attitude angle, information extraction is carried out on the target motion data record, a target attitude angle record is obtained, and the initial attitude angle and the real-time attitude angle of the target user are determined according to the target attitude angle record. And then, judging whether the real-time difference value of the attitude angles meets a preset attitude angle difference value threshold value, and if the real-time difference value of the attitude angles meets the preset attitude angle difference value threshold value, determining that the tumble-like action is a real tumble action. And if the real-time difference value of the attitude angles does not meet the preset attitude angle difference value threshold, the similar falling action is a false falling action. Wherein the fall-like actions comprise real fall actions and false fall actions. The target attitude angle record comprises an initial attitude angle and a real-time attitude angle of a target user. And the target attitude angle record has a time mark. The initial attitude angle can be used for representing initial attitude information of a target user corresponding to the real-time trough. The real-time attitude angle can be used for representing real-time attitude information of a target user corresponding to the real-time trough. The real-time attitude angle difference comprises a difference between an initial attitude angle and a real-time attitude angle of the target user. The preset attitude angle difference value threshold is determined by the artificial intelligence-based falling prevention device according to the self-adaptive setting of the actual situation. When the type of the target action is the falling-like action, whether the falling-like action is the real falling action or not is accurately judged through the attitude angle real-time difference value and the preset attitude angle difference threshold value, and then the technical effect of improving the accuracy of real-time falling-preventing early warning is achieved.
Step S700: and judging whether the target action type meets a preset type list or not, and performing real-time falling prevention early warning according to a judgment result.
Specifically, whether the target action type meets a preset type list or not is judged, a judgment result is obtained, and real-time falling prevention early warning is carried out according to the judgment result. Wherein the target action types comprise daily action, real fall action and false fall action. The preset type list comprises a plurality of real falling action types such as forward falling, backward falling, left falling, right falling and the like, and data information such as a resultant acceleration change condition, an angular velocity change condition, an attitude angle change condition and the like corresponding to the plurality of real falling action types. The judgment result comprises a list that the target action type meets the preset type and a list that the target action type does not meet the preset type. Illustratively, when the target action type is a real falling action, the target motion data record corresponding to the real falling action is compared with a preset type list, if the real falling action meets the preset type list, a real-time falling prevention early warning signal is sent out, real-time rescue measures of people nearby the target user are timely sought, and falling of the target user is avoided as much as possible. The anti-falling early warning method and the system achieve the technical effects of effectively performing real-time anti-falling early warning on the target user and improving the accuracy and the real-time performance of the anti-falling early warning.
Further, as shown in fig. 3, after step S700, the method further includes:
step S810: obtaining a user set, and performing fall evaluation on each user in the user set by using a Morse fall evaluation scale to obtain a fall evaluation result;
step S820: screening users with a falling evaluation value meeting a preset evaluation threshold value in the falling evaluation result to obtain a high-risk falling user set;
specifically, fall assessment is carried out on all users in the user set according to a Morse fall assessment scale, and a fall assessment result is obtained. Further, whether the tumble evaluation value in the tumble evaluation result meets a preset evaluation threshold value or not is judged, and the user set is screened according to the tumble evaluation result meeting the preset evaluation threshold value, so that the high-risk tumble user set is obtained. Wherein the user set comprises a plurality of users who use the artificial intelligence based fall prevention device for intelligent fall prevention protection. The Morse falling evaluation gauge has the functions of carrying out falling risk evaluation on all users in a user set according to a preset Morse falling evaluation standard, obtaining a falling evaluation value and the like. The fall evaluation result comprises a fall evaluation value of each user in the user set. The preset evaluation threshold value is preset and determined by the falling prevention device based on artificial intelligence according to the requirements of user set screening. The high-risk falling user set comprises a plurality of corresponding users which meet falling evaluation results of a preset evaluation threshold value in the user set. For example, if the fall evaluation value in the fall evaluation result satisfies the preset evaluation threshold, the user corresponding to the fall evaluation result may be set as a high-risk fall user, and then a high-risk fall user set composed of multiple high-risk fall users is determined. The technical effects of determining the high-risk falling user set and providing data support for subsequent falling prevention intervention on each high-risk falling user in the high-risk falling user set are achieved.
Step S830: and performing fall prevention intervention on all the high-risk fall users in the high-risk fall user set based on the Green mode.
Further, step S830 of the present application further includes:
step S831: carrying out comprehensive falling prevention capability evaluation on the high-risk falling users in the high-risk falling user set by using a falling prevention information credit rating table to obtain a comprehensive falling prevention capability evaluation result;
step S832: extracting a comprehensive falling prevention capability evaluation result of any high-risk falling user in the comprehensive falling prevention capability evaluation result, and individually making a falling prevention intervention scheme based on a Green model;
step S833: and performing fall prevention intervention on any high-risk fall user according to the fall prevention intervention scheme.
Specifically, the high-risk falling users are concentrated and evaluated by the falling prevention known credit rating table, so that the comprehensive falling prevention capability evaluation result is obtained. Further, the comprehensive falling prevention capability evaluation result is randomly extracted to obtain the comprehensive falling prevention capability evaluation result of any high-risk falling user, and a falling prevention intervention scheme is personalized to any high-risk falling user by combining the Green mode. And then carrying out fall prevention intervention on any high-risk fall user through a fall prevention intervention scheme. Wherein, the fall prevention information credit rating table is preset and determined by the artificial intelligence-based fall prevention device through data query. The falling prevention knowledge credit rating table has the function of carrying out comprehensive falling prevention capability evaluation on falling prevention willingness, falling prevention knowledge, falling prevention attitude and the like of all high-risk falling users in a high-risk falling user set. The comprehensive fall prevention capability evaluation result can be used for representing data information such as the fall prevention willingness, the fall prevention knowledge and the enthusiasm of the fall prevention attitude of all the high-risk fall users in the high-risk fall user set. For example, the higher the fall prevention willingness, the more the fall prevention knowledge, and the higher the positivity of the fall prevention attitude of the high-risk fall user, the higher the comprehensive fall prevention ability evaluation result of the high-risk fall user. The Green mode is to intervene in fall prevention willingness, fall prevention knowledge and fall prevention attitude of any high-risk fall user, so that the fall prevention intervention effect of any high-risk fall user is improved. The fall prevention intervention scheme comprises specific methods and steps for improving fall prevention willingness, fall prevention knowledge and fall prevention attitude of any high-risk fall user and other data information. The technical effects that the fall prevention intervention is carried out on any high-risk fall user through the personalized fall prevention intervention scheme, the accuracy of the fall prevention intervention is improved, and the quality of the fall prevention protection is further improved are achieved.
In summary, the anti-falling method based on artificial intelligence provided by the application has the following technical effects:
acquiring a target motion data record of a target user in real time through an inertial sensor; extracting the target motion data record to obtain a target resultant acceleration record, wherein the target resultant acceleration record has a time identifier; according to the target resultant acceleration record with the time identification, drawing a target resultant acceleration waveform in real time; performing characteristic analysis on the target combined acceleration waveform to obtain a characteristic analysis result, and determining a target action characteristic set according to the characteristic analysis result; acquiring a historical motion data set through big data acquisition, and further constructing an action recognition model; analyzing the target action characteristic set through the action recognition model to obtain a target action type of the target user; and judging whether the target action type meets a preset type list or not, and performing real-time falling prevention early warning according to a judgment result. The accuracy and the real-time performance of the anti-falling early warning are improved, the anti-falling early warning effect is further improved, and high-quality anti-falling protection is realized; meanwhile, the intelligence of the anti-falling early warning is improved, and the scientific anti-falling technical effect is achieved.
Example two
Based on the same inventive concept as the fall prevention method based on artificial intelligence in the foregoing embodiment, the present invention further provides a fall prevention device based on artificial intelligence, referring to fig. 4, where the device includes:
the acquisition module 11 is used for acquiring a target motion data record of a target user in real time by using an inertial sensor;
an extraction module 12, where the extraction module 12 is configured to extract a target resultant acceleration record in the target motion data record, where the target resultant acceleration record is recorded data with a time identifier;
the drawing module 13 is used for drawing a target combined acceleration waveform in real time according to the target combined acceleration record with the time identification;
the characteristic analysis module 14 is used for carrying out characteristic analysis on the target combined acceleration waveform and obtaining a target action characteristic set according to a characteristic analysis result;
the building module 15 is used for acquiring a historical motion data set based on big data and building an action recognition model according to the historical motion data set;
the action analysis module 16 is configured to analyze the target action feature set through the action recognition model to obtain a target action type of the target user;
and the judgment and early warning module 17 is used for judging whether the target action type meets a preset type list or not, and performing real-time falling prevention early warning according to a judgment result.
Further, the apparatus further comprises:
the target combined acceleration initial waveform determining module is used for taking a time identifier in the target combined acceleration record as an abscissa and taking the target combined acceleration in the target combined acceleration record as an ordinate to obtain a target combined acceleration initial waveform;
and the elongation processing module is used for performing elongation processing on the ordinate in the target combined acceleration initial waveform by preset times to obtain the target combined acceleration waveform.
Further, the apparatus further comprises:
a data set dividing module for dividing the historical movement data set into a falling movement data set and a daily movement data set;
a fall event determination module for obtaining motion data for a plurality of fall events from the fall motion data set;
the falling data analysis module is used for sequentially analyzing the motion data of the falling events to obtain a plurality of falling data analysis results;
a list establishing module, configured to establish a fall feature-type mapping list between fall features and fall types according to the analysis results of the fall data;
the action recognition model determining module is used for constructing the action recognition model according to the falling feature-type mapping list;
and the adjustment optimization module is used for carrying out a verification experiment on the action recognition model according to the daily movement data set and carrying out adjustment optimization on the action recognition model according to an experiment result.
Further, the apparatus further comprises:
the target wave trough determining module is used for sequentially obtaining a plurality of target wave troughs within a preset time threshold according to the target combined acceleration waveform, wherein the target wave troughs comprise real-time wave troughs;
the comparison module is used for comparing the target wave troughs to obtain a target wave trough minimum value;
the first judgment module is used for judging whether the minimum value of the target wave trough is the real-time wave trough or not through the action recognition model;
and the daily action determining module is used for determining that the target action type is daily action if the minimum value of the target wave trough is not the real-time wave trough.
Further, the apparatus further comprises:
a class falling action determining module, configured to determine that the target action type is a class falling action if the minimum value of the target trough is the real-time trough;
the target attitude angle record determining module is used for extracting a target attitude angle record in the target motion data record, wherein the target attitude angle record is record data with a time mark;
the attitude angle determining module is used for sequentially obtaining an initial attitude angle and a real-time attitude angle of the target user according to the target attitude angle record with the time identification;
the attitude angle real-time difference determining module is used for calculating the difference between the initial attitude angle of the target user and the real-time attitude angle to obtain an attitude angle real-time difference;
the second judgment module is used for judging whether the real-time attitude angle difference value meets a preset attitude angle difference value threshold value or not;
and the real falling action determining module is used for determining that the falling action is a real falling action if the real-time difference value of the attitude angle meets a preset attitude angle difference value threshold.
Further, the apparatus further comprises:
the system comprises a falling evaluation module, a storage module and a processing module, wherein the falling evaluation module is used for obtaining a user set and carrying out falling evaluation on each user in the user set by utilizing a Morse falling evaluation scale to obtain a falling evaluation result;
the high-risk falling user set determining module is used for screening users with falling evaluation values meeting a preset evaluation threshold value in the falling evaluation result to obtain a high-risk falling user set;
the first fall prevention intervention module is used for performing fall prevention intervention on all the high-risk fall users in the high-risk fall user set based on a Green mode.
Further, the apparatus further comprises:
the comprehensive falling prevention capability evaluation module is used for carrying out comprehensive falling prevention capability evaluation on the high-risk falling users in a centralized manner by using a falling prevention knowledge credit rating scale to obtain a comprehensive falling prevention capability evaluation result;
the fall prevention intervention scheme determination module is used for extracting a comprehensive fall prevention capability evaluation result of any high-risk fall user in the comprehensive fall prevention capability evaluation result and formulating a fall prevention intervention scheme based on a Green mode;
and the second fall prevention intervention module is used for performing fall prevention intervention on any high-risk fall user according to the fall prevention intervention scheme.
The application provides an artificial intelligence-based fall prevention method, wherein the method is applied to an artificial intelligence-based fall prevention device, and the method comprises the following steps: acquiring a target motion data record of a target user in real time through an inertial sensor; extracting the target motion data record to obtain a target resultant acceleration record, wherein the target resultant acceleration record has a time identifier; according to the target resultant acceleration record with the time identification, drawing a target resultant acceleration waveform in real time; performing characteristic analysis on the target combined acceleration waveform to obtain a characteristic analysis result, and determining a target action characteristic set according to the characteristic analysis result; acquiring a historical motion data set through big data acquisition, and further constructing an action recognition model; analyzing the target action characteristic set through the action recognition model to obtain a target action type of the target user; and judging whether the target action type meets a preset type list or not, and performing real-time falling prevention early warning according to a judgment result. The problem of among the prior art to the accuracy of preventing the early warning of tumbleing low, and then cause to prevent falling the not good technical problem of protection effect. The accuracy and the real-time performance of the anti-falling early warning are improved, the anti-falling early warning effect is further improved, and high-quality anti-falling protection is realized; meanwhile, the intelligence of the anti-falling early warning is improved, and the scientific anti-falling technical effect is achieved.
The technical features of the above embodiments can be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the above embodiments are not described, but should be considered as the scope of the present specification as long as there is no contradiction between the combinations of the technical features.
The specification and drawings are merely illustrative of the present application, and it is intended that the present invention cover modifications and variations of this invention provided they come within the scope of the invention and their equivalents.
Claims (8)
1. An artificial intelligence-based fall prevention method, comprising:
acquiring a target motion data record of a target user in real time by using an inertial sensor;
extracting a target resultant acceleration record in the target motion data record, wherein the target resultant acceleration record is record data with a time identifier;
according to the target resultant acceleration record with the time identification, a target resultant acceleration waveform is drawn in real time;
performing characteristic analysis on the target combined acceleration waveform, and obtaining a target action characteristic set according to a characteristic analysis result;
acquiring a historical motion data set based on big data acquisition, and constructing an action recognition model according to the historical motion data set;
analyzing the target action characteristic set through the action recognition model to obtain a target action type of the target user;
and judging whether the target action type meets a preset type list or not, and performing real-time falling prevention early warning according to a judgment result.
2. The method of claim 1, wherein the step of plotting the target resultant acceleration waveform in real time according to the target resultant acceleration record with the time stamp comprises:
taking the time mark in the target combined acceleration record as an abscissa and the target combined acceleration in the target combined acceleration record as an ordinate to obtain a target combined acceleration initial waveform;
and performing elongation processing of preset times on the ordinate in the target combined acceleration initial waveform to obtain the target combined acceleration waveform.
3. The method of claim 1, wherein constructing an action recognition model from the historical motion data set comprises:
dividing the historical movement data set into a falling movement data set and a daily movement data set;
obtaining motion data of a plurality of falling events according to the falling motion data set;
analyzing the motion data of the plurality of falling events in sequence to obtain a plurality of falling data analysis results;
establishing a falling feature-type mapping list between falling features and falling types according to the analysis results of the plurality of falling data;
constructing the action recognition model according to the falling feature-type mapping list;
and performing a verification experiment on the action recognition model according to the daily motion data set, and adjusting and optimizing the action recognition model according to an experiment result.
4. The method of claim 3, wherein the analyzing the set of target motion characteristics through the motion recognition model to obtain the target motion type of the target user comprises:
sequentially obtaining a plurality of target wave troughs within a preset time threshold according to the target resultant acceleration waveform, wherein the target wave troughs comprise real-time wave troughs;
comparing the multiple target wave troughs to obtain a target wave trough minimum value;
judging whether the minimum value of the target wave trough is the real-time wave trough or not through the action recognition model;
and if the minimum value of the target wave trough is not the real-time wave trough, the target action type is daily action.
5. The method of claim 4, further comprising:
if the minimum value of the target wave trough is the real-time wave trough, the target action type is a fall-like action;
extracting a target attitude angle record in the target motion data record, wherein the target attitude angle record is record data with a time mark;
sequentially obtaining an initial attitude angle and a real-time attitude angle of the target user according to the target attitude angle record with the time identification;
calculating the difference value between the initial attitude angle and the real-time attitude angle of the target user to obtain an attitude angle real-time difference value;
judging whether the attitude angle real-time difference value meets a preset attitude angle difference value threshold value or not;
and if the real-time difference value of the attitude angles meets a preset attitude angle difference value threshold, the falling-like action is a real falling action.
6. The method of claim 1, further comprising:
obtaining a user set, and performing fall evaluation on each user in the user set by using a Morse fall evaluation scale to obtain a fall evaluation result;
screening users with a falling evaluation value meeting a preset evaluation threshold value in the falling evaluation result to obtain a high-risk falling user set;
and performing fall prevention intervention on all the high-risk fall users in the high-risk fall user set based on the Green mode.
7. The method of claim 6, wherein the performing fall prevention intervention on the high-risk fall users in the high-risk fall user set based on Green's model comprises:
carrying out comprehensive falling prevention capability evaluation on the high-risk falling users in the high-risk falling user set by using a falling prevention information credit rating table to obtain a comprehensive falling prevention capability evaluation result;
extracting the comprehensive falling prevention capability evaluation result of any high-risk falling user in the comprehensive falling prevention capability evaluation result, and formulating a falling prevention intervention scheme based on a Green mode;
and performing fall prevention intervention on any high-risk fall user according to the fall prevention intervention scheme.
8. An artificial intelligence based fall prevention device, the device comprising:
the acquisition module is used for acquiring a target motion data record of a target user in real time by using the inertial sensor;
the extraction module is used for extracting a target resultant acceleration record in the target motion data record, wherein the target resultant acceleration record is record data with a time identifier;
the drawing module is used for drawing a target resultant acceleration waveform in real time according to the target resultant acceleration record with the time identification;
the characteristic analysis module is used for carrying out characteristic analysis on the target combined acceleration waveform and obtaining a target action characteristic set according to a characteristic analysis result;
the building module is used for acquiring a historical motion data set based on big data and building an action recognition model according to the historical motion data set;
the action analysis module is used for analyzing the target action characteristic set through the action recognition model to obtain a target action type of the target user;
and the judging and early warning module is used for judging whether the target action type meets a preset type list or not and carrying out real-time falling prevention early warning according to a judging result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210845138.6A CN114997245A (en) | 2022-07-19 | 2022-07-19 | Artificial intelligence-based anti-falling method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210845138.6A CN114997245A (en) | 2022-07-19 | 2022-07-19 | Artificial intelligence-based anti-falling method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114997245A true CN114997245A (en) | 2022-09-02 |
Family
ID=83022227
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210845138.6A Pending CN114997245A (en) | 2022-07-19 | 2022-07-19 | Artificial intelligence-based anti-falling method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114997245A (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106910314A (en) * | 2017-02-03 | 2017-06-30 | 同济大学 | A kind of personalized fall detection method based on the bodily form |
| CN109087482A (en) * | 2018-09-18 | 2018-12-25 | 西安交通大学 | A kind of falling detection device and method |
| CN109166275A (en) * | 2018-09-25 | 2019-01-08 | 山东科技大学 | A kind of tumble detection method for human body based on acceleration transducer |
| CN112233374A (en) * | 2020-09-21 | 2021-01-15 | 中国科学院深圳先进技术研究院 | Fall detection method, system, terminal and storage medium |
| CN113450539A (en) * | 2021-07-12 | 2021-09-28 | 杭州电子科技大学 | Fall detection method |
-
2022
- 2022-07-19 CN CN202210845138.6A patent/CN114997245A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106910314A (en) * | 2017-02-03 | 2017-06-30 | 同济大学 | A kind of personalized fall detection method based on the bodily form |
| CN109087482A (en) * | 2018-09-18 | 2018-12-25 | 西安交通大学 | A kind of falling detection device and method |
| CN109166275A (en) * | 2018-09-25 | 2019-01-08 | 山东科技大学 | A kind of tumble detection method for human body based on acceleration transducer |
| CN112233374A (en) * | 2020-09-21 | 2021-01-15 | 中国科学院深圳先进技术研究院 | Fall detection method, system, terminal and storage medium |
| CN113450539A (en) * | 2021-07-12 | 2021-09-28 | 杭州电子科技大学 | Fall detection method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Wang et al. | Fall detection based on dual-channel feature integration | |
| CN106022378B (en) | Sitting posture judgment method and based on camera and pressure sensor cervical spondylosis identifying system | |
| CN103699795A (en) | Exercise behavior identification method and device and exercise intensity monitoring system | |
| CN111089604B (en) | Body-building exercise identification method based on wearable sensor | |
| CN112115827B (en) | Falling behavior identification method based on human body posture dynamic characteristics | |
| CN108958482B (en) | Similarity action recognition device and method based on convolutional neural network | |
| CN110418337B (en) | Identity authentication method, electronic device and computer-readable storage medium | |
| CN111709277A (en) | Human body tumbling detection method and device, computer equipment and storage medium | |
| CN110456904B (en) | Augmented reality glasses eye movement interaction method and system without calibration | |
| CN106650300B (en) | Old man monitoring system and method based on extreme learning machine | |
| CN108969980A (en) | Treadmill and step counting method, device and storage medium thereof | |
| KR101898648B1 (en) | Apparatus and Method for Detecting Interacting Groups Between Individuals in an Image | |
| CN111603750A (en) | Motion capture recognition evaluation system and method based on edge calculation | |
| CN113768471B (en) | Parkinson disease auxiliary diagnosis system based on gait analysis | |
| CN113378691B (en) | Intelligent home management system and method based on real-time user behavior analysis | |
| CN110123280A (en) | A kind of finger dexterity detection method based on the identification of intelligent mobile terminal operation behavior | |
| CN111783717B (en) | Biological characteristic movement mode intelligent recognition method and application thereof | |
| CN106971203A (en) | Personal identification method based on characteristic on foot | |
| CN114997245A (en) | Artificial intelligence-based anti-falling method and device | |
| CN111144167A (en) | Gait information identification optimization method, system and storage medium | |
| CN108805037A (en) | It is a kind of to utilize the human body of picture signal and electric signal and equipment matching process | |
| CN114264239B (en) | Motion platform laser calibration system | |
| CN107688828A (en) | A kind of bus degree of crowding estimating and measuring method based on mobile phone sensor | |
| CN112257559A (en) | Identity recognition method based on gait information of biological individual | |
| CN111428690A (en) | Identity authentication method based on gait signal topology analysis |
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 |