CN109166275B - Human body falling detection method based on acceleration sensor - Google Patents
Human body falling detection method based on acceleration sensor Download PDFInfo
- Publication number
- CN109166275B CN109166275B CN201811112919.4A CN201811112919A CN109166275B CN 109166275 B CN109166275 B CN 109166275B CN 201811112919 A CN201811112919 A CN 201811112919A CN 109166275 B CN109166275 B CN 109166275B
- Authority
- CN
- China
- Prior art keywords
- falling
- data
- threshold
- acceleration
- human body
- 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.)
- Active
Links
Images
Classifications
-
- 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
-
- 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
- G08B21/0438—Sensor means for detecting
- G08B21/0446—Sensor means for detecting worn on the body to detect changes of posture, e.g. a fall, inclination, acceleration, gait
Abstract
The invention discloses a human body falling detection method based on an acceleration sensor, and relates to the technical field of biomedical signal processing. The method is based on a double-confirmation human body falling detection algorithm, wherein primary falling detection extracts a threshold set according to pre-acquired daily behavior and Action (ADL) and falling acceleration sample data; then, ADL acceleration data of the human body are collected in real time, and the variance of the group of data is extracted to serve as a dynamic threshold part to update a pre-falling behavior threshold value for the user; setting the threshold as a human body falling judgment standard to perform primary detection of human body falling behaviors; in the primary falling detection process, once the pre-falling behavior is detected, the human body behavior data starting from the ts before the pre-falling behavior is sent to a server close to a data source for secondary falling detection and judgment based on the SVM; and finally, determining a human body falling event according to the double falling judgment result, and dynamically updating a threshold set of a falling detection algorithm for subsequent falling detection of the user.
Description
Technical Field
The invention relates to the technical field of biomedical signal processing, in particular to a human body falling detection method based on an acceleration sensor.
Background
With the continuous increase of the world population, the continuous perfection of medical systems and the continuous increase of the aging speed of the population, China also faces the prominent problem of aging, more and more elderly people live alone are in the elderly, and the empty nest phenomenon is more and more obvious. However, the health and safety problems of elderly people living alone are becoming serious social problems. Statistically, about one third of the elderly over 65 years of age have fallen worldwide every year, and falls are often accompanied by serious physical and psychological injuries, thereby putting a burden on the family and society. If the falling event of the old is detected timely and accurately and the alarm is given, the damage of falling to the old can be reduced to the maximum extent, and the method plays an important role in the independent life of the old.
In the human body fall detection research, a fall detection method based on an acceleration sensor is one of the most common detection means. In the falling process of a human body, the motion posture of the human body can be changed remarkably, for example, the weightlessness state of the human body, violent impact and the like, the human body wears acceleration sensor equipment to acquire the motion information of the human body in real time, and the falling state of the human body can be detected by combining a specific falling detection algorithm. For example, in chinese patent No. CN201710796479.8, the fall detection algorithm for a human body based on dual thresholds is implemented by collecting acceleration data of human body movement, converting the acceleration data into angle values to calculate angle gradient data, further extracting an inclination gradient variance as a feature value, and selecting thresholds at two moments after a fall occurs and comparing the thresholds.
In chinese patent No. CN201711268665.0, training sample acceleration data is used to extract the peak value of acceleration signal vector magnitude, the most significant difference value of acceleration signal vector magnitude, the standard deviation of acceleration signal vector magnitude, and the relative angle variation value, and a K-means clustering method is combined to train a binary decision threshold of each feature value, and the determination of human body falling is realized by performing feature extraction on the original information during real-time falling detection and then comparing the extracted original information with the feature threshold.
In chinese patent No. CN201711489128.9, acceleration data of hand swing of a human body during walking is collected by an acceleration sensor, a gyroscope, a geomagnetic instrument, and an air pressure sensor, and training is performed by extracting an average value of time between a peak value a2 of an average acceleration and a valley value a' 1 of a next acceleration between a peak value a1 of an average value of an average acceleration of a valley value of an acceleration and a peak value a2 of an acceleration as a feature parameter structure SVM classifier, thereby realizing fall detection of the human body.
In chinese patent No. CN201210586385.5, the whole fall process is divided into 4 stages: the falling detection method comprises a vertical standing stage, a falling stage in the early stage of falling, a collision stage in the falling process and a lying and near-static stage after falling, wherein the falling and collision stages are judged by adopting a method of setting an acceleration threshold value, and fall detection is realized by analyzing the standing and lying states of a human body in combination with an angle.
Although the fall detection method can detect fall behaviors of most human bodies, the accuracy of fall detection is difficult to guarantee due to individual differences of users and insufficient current real fall data. Therefore, although many existing fall detection methods are available, the requirements for high accuracy of fall detection cannot be met well.
Disclosure of Invention
The invention aims to provide a human body falling detection method based on an acceleration sensor, which solves the problems of false alarm and missing report caused by neglecting individual difference in the current human body falling detection and the problem of insufficient real falling number.
The invention specifically adopts the following technical scheme:
a human body falling detection method based on an acceleration sensor is based on a detection system, the detection system comprises a data acquisition module, a threshold extraction module, a falling detection module and an alarm and threshold updating module which are connected with each other, the alarm and threshold updating module is connected with the falling detection module to form a loop, and the method specifically comprises the following steps:
the method comprises the following steps: extracting a threshold value set; preprocessing ADL and tumble acceleration sample data which are collected in advance, and extracting a pre-tumble behavior threshold TH1, a tumble collision threshold TH2, a tumble recovery state threshold TH3 and a tumble posture threshold TH 4; calculating the sum of the mean value and the standard deviation of all ADL data in acceleration sample data acquired in advance to serve as a pre-falling judgment static threshold part; analyzing a falling data set in acceleration sample data collected in advance, respectively calculating the difference from a valley value to a peak value of the acceleration data, a relative acceleration value and a final angle value as characteristics, and extracting a falling collision threshold TH2, a falling recovery state threshold TH3 and a falling posture threshold TH4 of a human body;
step two: updating the threshold TH1 for fall behaviour of the user; acquiring acceleration data of an actual user under an appointed ADL action in real time, calculating a standard deviation of the data under the action of the daily action as a dynamic threshold part, and further extracting and updating a pre-falling action threshold TH1 by combining the static threshold part in the step one;
step three: performing primary lightweight fall detection and judgment; acquiring and calculating real-time acceleration data of a user in real time, gradually judging pre-falling behavior, falling collision behavior, falling recovery behavior and final posture of a human body, further determining whether the human body falls, and starting wireless transmission at the moment when the falling behavior of the human body is detected; sending the data from the ts moment before the moment to the algorithm ending moment to a server near a data source in real time for further processing, and simultaneously sending alarm signals corresponding to the first-level lightweight fall detection to the server end;
step four: carrying out falling detection judgment based on the SVM, and carrying out falling detection on the received acceleration data by using a trained SVM classifier; if the result is the non-falling behavior, no alarm is given, and if the result is the falling behavior, an alarm is given;
step five: and performing double confirmation and threshold updating, performing comprehensive confirmation according to the judgment results of the two-stage fall detection in the third step and the fourth step, alarming when the two-stage detection occurs, confirming that the human body falls, calculating and updating TH2, TH3 and TH4 again by taking the data in the period of time as fall data, and performing subsequent fall detection on the user, and calculating and updating TH1 again by using the data if the human behavior is a daily behavior.
Preferably, the acceleration sample data collected in advance in the first step is acceleration data of actions and falls of a user wearing equipment including an acceleration sensor according to a specified ADL by collecting in advance different ages, sexes, heights, bodies and other requirements, or an existing sample database.
Preferably, the pre-fall behavior threshold TH1 is obtained by calculating a static threshold obtained by acquiring ADL data in acceleration sample data acquired in advance and a dynamic threshold obtained by acquiring acceleration data of an actual user ADL in real time, and taking the sum of the two thresholds.
Preferably, the first-stage lightweight fall detection in the third step adopts a four-stage state step-by-step judgment mode, and judges whether the acceleration sample data acquired in real time is greater than TH1, if not, judges that no fall behavior occurs currently, and executes the lightweight fall detection again; if so, judging that the pre-falling behavior occurs currently, continuously judging whether the difference of the acceleration data from the valley value to the peak value is greater than TH2, if not, judging that the falling behavior does not occur currently, and performing light-weight falling detection again; if so, judging that the human body is violently impacted, continuously judging whether the relative acceleration value is smaller than TH3, if not, judging that no falling behavior occurs currently, and executing the lightweight falling detection again; if so, judging that the human body is in a relatively stable state, continuously judging whether the final angle is smaller than TH4, if not, judging that no falling behavior occurs currently, and executing the lightweight falling detection again; if so, judging that the human body has a falling behavior.
Preferably, the SVM classifier trained in the fourth step is obtained by training with the acceleration sample data collected in advance, and is trained by searching a period of each group of training data, and calculating a mean value, a standard deviation, a difference between an acceleration valley and a peak value, a time interval between an acceleration valley and a peak value, a difference between an acceleration peak value and a valley value, a time interval between an acceleration peak value and a valley value, an acceleration mean value and a standard deviation in a specified time interval after a second valley, and a rear angle value in the specified time interval, and using the extracted features as a feature value set to construct an SVM classifier for training.
Preferably, in the fifth step, the threshold updating of TH2, TH3 and TH4 is realized by recalculating and updating the data at this time as fall data when the double confirmation result indicates that both the persons are falling behaviors, and the threshold updating of TH1 is realized by recalculating and updating the data at this time as ADL data when both the double confirmation result indicates that both the persons are ADL actions.
Preferably, the threshold updating of TH2, TH3 and TH4 is to fuse the added new data into the previous fall data set, and also to extract the threshold of TH2, TH3 and TH4 again as the fall detection threshold of the user by using a confidence interval mathematical analysis method; the threshold updating of TH1 is achieved by fusing newly added data to the previous ADL data set, recalculating the static thresholds of the data set, and re-extracting by the dynamic threshold method.
Preferably, the data acquisition module is configured to acquire acceleration sample data of an ADL action and a fall action specified by a user and acquire the acceleration data of the user action in real time in an actual use process;
the threshold extraction module is connected with the acquisition module and is used for analyzing and extracting static thresholds of an ADL acceleration data set acquired in advance and dynamic threshold parts of real-time ADL acceleration data of a user so as to extract TH1, analyzing an acceleration data set of tumble actions acquired in advance, and extracting TH2, TH3 and TH 4;
the fall detection module is connected with the data acquisition module and the threshold extraction module, and realizes real-time detection of the final fall behavior of the user by adopting a two-stage fall judgment mode based on double confirmation;
the alarm and threshold updating module and the fall detection module are connected to form a loop for fall alarm and updating of threshold sets TH1, TH2, TH3 and TH 4.
The invention has the following beneficial effects:
the problem of falling detection accuracy caused by individual difference is solved by adopting a dynamic threshold updating idea;
the fall monitoring method which utilizes dual determination of light-weight fall judgment and SVM-based two-stage fall detection judgment by utilizing one-stage fall detection is adopted, the set threshold value is updated by utilizing real fall data, the problem that the current real fall data is difficult to obtain is solved, the fall detection accuracy is improved, and the high-precision requirement of human body fall detection is met.
The system can detect the falling of the user in real time and with high accuracy under the condition that the human body falls, and can timely and effectively rescue the user, thereby ensuring the personal safety of the user.
Drawings
Fig. 1 is a flow chart of a human fall detection method based on an acceleration sensor;
fig. 2 is a detailed flowchart of step S3 in the method for detecting a human fall based on an acceleration sensor;
fig. 3 is a schematic structural diagram of a human body fall detection system based on an acceleration sensor.
Detailed Description
ADL: a daily behavior act;
SVM classifier: the discriminant classifier is defined by a classification hyperplane, i.e., given a set of labeled training samples, the algorithm will output an optimal hyperplane to classify new samples (test samples).
The following description of the embodiments of the present invention will be made with reference to the accompanying drawings:
as shown in fig. 1 to 3, a human body fall detection method based on an acceleration sensor is based on a detection system, the detection system includes a data acquisition module, a threshold extraction module, a fall detection module and an alarm and threshold update module, which are connected to each other, the alarm and threshold update module is connected to the fall detection module to form a loop, and the method specifically includes:
the method comprises the following steps: extracting a threshold value set; preprocessing ADL and tumble acceleration sample data which are collected in advance, and extracting a pre-tumble behavior threshold TH1, a tumble collision threshold TH2, a tumble recovery state threshold TH3 and a tumble posture threshold TH 4; calculating the sum of the mean value and the standard deviation of all ADL data in acceleration sample data acquired in advance to serve as a pre-falling judgment static threshold part; analyzing a falling data set in acceleration sample data collected in advance, respectively calculating the difference from a valley value to a peak value of the acceleration data, a relative acceleration value and a final angle value as characteristics, and extracting a falling collision threshold TH2, a falling recovery state threshold TH3 and a falling posture threshold TH4 of a human body;
the main acquisition mode of the acceleration sample data collected in advance is as follows: the method comprises the steps that each user wears equipment comprising an acceleration sensor on the waist to finish specified ADL and fall actions, and acceleration data of the user are collected in the process of finishing the specified actions, wherein the users participating in data collection mainly comprise users of different ages, sexes, heights, weights and the like. The designated ADL actions mainly include standing, walking, sitting, jumping, squatting, lying, walk-sitting, walk-lying, squatting-standing, climbing stairs, etc., and the designated fall actions mainly include falling forward, falling backward, falling sideways, tripping, etc.
Step two: updating the threshold TH1 for fall behaviour of the user; acquiring acceleration data of an actual user under an appointed ADL action in real time, calculating a standard deviation of the data under the action of the daily action as a dynamic threshold part, and further extracting and updating a pre-falling action threshold TH1 by combining the static threshold part in the step one;
step three: performing primary lightweight fall detection and judgment; acquiring and calculating real-time acceleration data of a user in real time, gradually judging pre-falling behavior, falling collision behavior, falling recovery behavior and final posture of a human body, further determining whether the human body falls, and starting wireless transmission at the moment when the falling behavior of the human body is detected; sending the data from the ts moment before the moment to the algorithm ending moment to a server near a data source in real time for further processing, and simultaneously sending alarm signals corresponding to the first-level lightweight fall detection to the server end;
step four: carrying out falling detection judgment based on the SVM, and carrying out falling detection on the received acceleration data by using a trained SVM classifier; if the result is the non-falling behavior, no alarm is given, and if the result is the falling behavior, an alarm is given;
step five: and performing double confirmation and threshold updating, performing comprehensive confirmation according to the judgment results of the two-stage fall detection in the third step and the fourth step, performing judgment and alarm in both the two-stage detection, confirming that the human body falls, simultaneously calculating and updating TH2, TH3 and TH4 by taking the data in the period of time as fall data, and performing subsequent fall detection on the user, and calculating and updating TH1 the data again if the human behavior is a daily behavior.
The acceleration sample data collected in advance in the first step is acceleration data of a user wearing equipment including an acceleration sensor according to a specified ADL action and a fall action under the requirements of different ages, sexes, heights, bodies and the like, or is an existing sample database.
The pre-fall behavior threshold TH1 is obtained by calculating a static threshold obtained by acquiring ADL data in acceleration sample data acquired in advance and a dynamic threshold obtained by acquiring acceleration data of an actual user ADL in real time, and taking the sum of the two thresholds.
In the third step, the first-level lightweight fall detection adopts a four-level state step-by-step judgment mode, whether the acceleration sample data acquired in real time is greater than TH1 is judged, if not, the current fall behavior is judged, and the lightweight fall detection is executed again; if so, judging that the pre-falling behavior occurs currently, continuously judging whether the difference of the acceleration data from the valley value to the peak value is greater than TH2, if not, judging that the falling behavior does not occur currently, and performing light-weight falling detection again; if so, judging that the human body is violently impacted, continuously judging whether the relative acceleration value is smaller than TH3, if not, judging that no falling behavior occurs currently, and executing the lightweight falling detection again; if so, judging that the human body is in a relatively stable state, continuously judging whether the final angle is smaller than TH4, if not, judging that no falling behavior occurs currently, and executing the lightweight falling detection again; if so, judging that the human body has a falling behavior.
The SVM classifier trained in the fourth step is obtained by utilizing the acceleration sample data acquired in advance, the period of each group of training data is searched, the mean value, the standard deviation, the difference between the acceleration valley value and the peak value, the time interval from the acceleration trough to the acceleration peak value, the difference between the acceleration peak value and the acceleration valley value, the time interval from the acceleration peak value to the valley value, the mean value and the standard deviation of the acceleration in the specified time interval after the second valley and the rear angle value of the specified time interval are calculated, and the extracted 9 features are used as the feature value set to construct the SVM classifier to be trained.
In the fifth step, the threshold updating of TH2, TH3 and TH4 is realized by recalculating and updating the data at this time as fall data when the double confirmation result indicates that the fall is a fall action, and the threshold updating of TH1 is realized by recalculating and updating the data at this time as ADL data when the double confirmation result indicates that the ADL action is a double action.
Threshold updating of TH2, TH3 and TH4, namely, fusing the added new data to a previous falling data set, and also adopting a confidence interval mathematical analysis method to extract TH2, TH3 and TH4 as the falling detection threshold of the user; the threshold updating of TH1 is achieved by fusing newly added data to the previous ADL data set, recalculating the static thresholds of the data set, and re-extracting by the dynamic threshold method.
The data acquisition module is used for acquiring acceleration sample data of ADL actions and falling actions appointed by a user and acquiring the acceleration data of the user actions in real time in the actual use process;
the threshold extraction module is connected with the acquisition module and is used for analyzing and extracting static thresholds (TH1 components) of the action acceleration data set of daily behaviors and dynamic threshold parts of real-time ADL action acceleration data of the user in advance so as to extract TH1, analyzing and acquiring a fall action acceleration data set in advance, and extracting TH2, TH3 and TH 4;
the falling detection module is connected with the data acquisition module and the threshold extraction module and adopts a two-stage falling judgment mode based on double confirmation to realize the real-time detection of the final falling behavior of the user;
the alarm and threshold updating module and the fall detection module are connected to form a loop for fall alarm and updating of threshold sets TH1, TH2, TH3 and TH 4.
It is to be understood that the above description is not intended to limit the present invention, and the present invention is not limited to the above examples, and those skilled in the art may make modifications, alterations, additions or substitutions within the spirit and scope of the present invention.
Claims (8)
1. A human body falling detection method based on an acceleration sensor is based on a detection system, the detection system comprises a data acquisition module, a threshold extraction module, a falling detection module and an alarm and threshold updating module which are connected with each other, the alarm and threshold updating module is connected with the falling detection module to form a loop, and the human body falling detection method is characterized by specifically comprising the following steps:
the method comprises the following steps: extracting a threshold value set; preprocessing pre-collected daily behavior and Action (ADL) and falling acceleration sample data and extracting a pre-falling behavior threshold TH1, a falling collision threshold TH2, a falling recovery state threshold TH3 and a falling posture threshold TH 4; calculating the sum of the mean value and the standard deviation of all ADL data in acceleration sample data acquired in advance to serve as a pre-falling judgment static threshold part; analyzing a falling data set in acceleration sample data collected in advance, respectively calculating the difference from a valley value to a peak value of the acceleration data, a relative acceleration value and a final angle value as characteristics, and extracting a falling collision threshold TH2, a falling recovery state threshold TH3 and a falling posture threshold TH4 of a human body;
step two: updating the threshold TH1 for fall behaviour of the user; acquiring acceleration data of an actual user under an appointed ADL action in real time, calculating a standard deviation of the data of the actual user under the appointed ADL action in real time as a dynamic threshold part, and further extracting and updating a pre-falling behavior threshold TH1 by combining a static threshold part in the step one;
step three: performing primary lightweight fall detection and judgment; acquiring and calculating real-time acceleration data of a user in real time, gradually judging pre-falling behavior, falling collision behavior, falling recovery behavior and final posture of a human body, further determining whether the human body falls, and starting wireless transmission at the moment when the falling behavior of the human body is detected; sending the data from the ts moment before the moment to the moment when the primary lightweight fall detection is finished to a server near a data source in real time for further processing, and simultaneously sending alarm signals corresponding to the primary lightweight fall detection to the server end;
step four: carrying out falling detection judgment based on the SVM, and carrying out falling detection on the received acceleration data by using a trained SVM classifier; if the result is the non-falling behavior, no alarm is given, and if the result is the falling behavior, an alarm is given;
step five: and performing double confirmation and threshold updating, performing comprehensive confirmation according to the judgment results of the two-stage fall detection in the third step and the fourth step, alarming when the two-stage detection occurs, confirming that the human body falls, calculating and updating TH2, TH3 and TH4 again by taking the data in the time period as fall data, and performing subsequent fall detection on the user, and calculating and updating TH1 again the data in the time period if the human behavior is daily behavior.
2. The method for detecting human body fall based on acceleration sensor as claimed in claim 1, wherein the pre-collected acceleration sample data in the step one is obtained by pre-collecting acceleration data of user wearing equipment including acceleration sensor according to the assigned daily activity and fall (ADL) under different age, sex, height and weight requirements, or is an existing sample database.
3. An acceleration sensor based human fall detection method as claimed in claim 1, wherein the pre-fall behavior threshold TH1 is obtained by calculating a static threshold obtained by acquiring ADL data from pre-collected acceleration sample data and a dynamic threshold obtained by collecting acceleration data of an actual user ADL in real time, and taking the sum of the two thresholds.
4. The human body falling detection method based on the acceleration sensor as claimed in claim 1, wherein the first-level lightweight falling detection in the third step adopts a four-level state step-by-step judgment mode, by judging whether the acceleration sample data acquired in real time is greater than TH1, if not, judging that no falling behavior occurs currently, and re-executing the lightweight falling detection; if so, judging that the pre-falling behavior occurs currently, continuously judging whether the difference of the acceleration data from the valley value to the peak value is greater than TH2, if not, judging that the falling behavior does not occur currently, and performing light-weight falling detection again; if so, judging that the human body is violently impacted, continuously judging whether the relative acceleration value is smaller than TH3, if not, judging that no falling behavior occurs currently, and executing the lightweight falling detection again; if so, judging that the human body is in a relatively stable state, continuously judging whether the final angle is smaller than TH4, if not, judging that no falling behavior occurs currently, and executing the lightweight falling detection again; if so, judging that the human body has a falling behavior.
5. The method as claimed in claim 1, wherein the SVM classifier trained in step four is obtained by training with the pre-collected acceleration sample data, and the training is performed by searching the period of each training data set, and calculating the mean, standard deviation, acceleration valley to peak difference, acceleration valley to peak time interval, acceleration peak to valley difference, acceleration peak to valley time interval, acceleration mean and standard deviation within a specified time interval after the second valley, and rear angle value within a specified time interval, and constructing the SVM classifier using the extracted features as the feature value set for training.
6. The human body fall detection method based on the acceleration sensor as claimed in claim 1, wherein in the fifth step, the threshold updating of TH2, TH3 and TH4 is implemented by recalculating and updating the data at this time as fall data when the double confirmation result indicates that both the two persons fall, and the threshold updating of TH1 is implemented by recalculating and updating the data at this time as ADL data when the double confirmation result indicates that both the two persons fall.
7. The human body fall detection method based on the acceleration sensor as claimed in claim 6, characterized in that, the threshold values of TH2, TH3 and TH4 are updated by fusing new data added to the previous fall data set, and extracting again the threshold values of TH2, TH3 and TH4 as the fall detection threshold values of the user by using a confidence interval mathematical analysis method; the threshold updating of TH1 is achieved by fusing newly added data to the previous ADL data set, recalculating the static thresholds of the data set, and re-extracting by the dynamic threshold method.
8. The human body fall detection method based on the acceleration sensor as claimed in claim 1, wherein the data acquisition module is configured to acquire acceleration sample data of a user specified ADL action and fall action and real-time acquisition of acceleration data of the user action during actual use;
the threshold extraction module is connected with the acquisition module and is used for analyzing and extracting static thresholds of an ADL acceleration data set acquired in advance and dynamic threshold parts of real-time ADL acceleration data of a user so as to extract TH1, analyzing an acceleration data set of tumble actions acquired in advance, and extracting TH2, TH3 and TH 4;
the fall detection module is connected with the data acquisition module and the threshold extraction module, and realizes real-time detection of the final fall behavior of the user by adopting a two-stage fall judgment mode based on double confirmation;
the alarm and threshold updating module and the fall detection module are connected to form a loop for fall alarm and updating of threshold sets TH1, TH2, TH3 and TH 4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811112919.4A CN109166275B (en) | 2018-09-25 | 2018-09-25 | Human body falling detection method based on acceleration sensor |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811112919.4A CN109166275B (en) | 2018-09-25 | 2018-09-25 | Human body falling detection method based on acceleration sensor |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109166275A CN109166275A (en) | 2019-01-08 |
| CN109166275B true CN109166275B (en) | 2020-08-11 |
Family
ID=64880163
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811112919.4A Active CN109166275B (en) | 2018-09-25 | 2018-09-25 | Human body falling detection method based on acceleration sensor |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109166275B (en) |
Families Citing this family (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110017834B (en) * | 2019-04-15 | 2021-12-24 | 歌尔科技有限公司 | Usage object determination method, usage object determination apparatus, and storage medium |
| CN110322662B (en) * | 2019-05-05 | 2021-04-30 | 北京理工大学 | Posture monitoring and falling alarm bracelet based on physiological and behavioral characteristics |
| CN110738822B (en) * | 2019-10-24 | 2022-09-20 | 合肥盛东信息科技有限公司 | Tumble early warning system |
| CN110769068A (en) * | 2019-10-30 | 2020-02-07 | 软通动力信息技术有限公司 | Method, device, server and storage medium for generating alarm message |
| EP3828855A1 (en) | 2019-11-29 | 2021-06-02 | Koninklijke Philips N.V. | Personalized fall detector |
| CN111998829B (en) * | 2020-10-30 | 2021-01-29 | 四川写正智能科技有限公司 | Method for judging read-write posture based on sensor |
| CN113358111A (en) * | 2021-05-28 | 2021-09-07 | 北京金山云网络技术有限公司 | Fall detection method and device and electronic equipment |
| CN114034821A (en) * | 2021-11-10 | 2022-02-11 | 国家电网有限公司 | System and method for monitoring and alarming gas of power line deep foundation pit |
| CN114926959A (en) * | 2022-06-16 | 2022-08-19 | 王思杨 | Self-power-generation old man falling alarm device and falling judgment method thereof |
| CN114997245A (en) * | 2022-07-19 | 2022-09-02 | 广州帝隆科技股份有限公司 | Artificial intelligence-based anti-falling method and device |
| CN116439694B (en) * | 2023-06-14 | 2023-08-15 | 深圳市魔样科技有限公司 | Intelligent watch dynamic data monitoring method based on motion model training |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8206325B1 (en) * | 2007-10-12 | 2012-06-26 | Biosensics, L.L.C. | Ambulatory system for measuring and monitoring physical activity and risk of falling and for automatic fall detection |
| CN105654666B (en) * | 2016-02-25 | 2018-03-20 | 福州大学 | Falls in Old People monitoring method based on MEMS sensor |
| CN106910314B (en) * | 2017-02-03 | 2019-01-25 | 同济大学 | A kind of personalized fall detection method based on the bodily form |
| 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 |
| CN108257352B (en) * | 2017-12-30 | 2019-12-20 | 广州柏颐信息科技有限公司 | Fall detection and early warning method based on intelligent wearable equipment |
-
2018
- 2018-09-25 CN CN201811112919.4A patent/CN109166275B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN109166275A (en) | 2019-01-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109166275B (en) | Human body falling detection method based on acceleration sensor | |
| CN108256433B (en) | Motion attitude assessment method and system | |
| Wang et al. | Walking pattern classification and walking distance estimation algorithms using gait phase information | |
| Aziz et al. | An analysis of the accuracy of wearable sensors for classifying the causes of falls in humans | |
| CN108509897A (en) | A kind of human posture recognition method and system | |
| CN110287825B (en) | Tumble action detection method based on key skeleton point trajectory analysis | |
| Dubois et al. | Human activities recognition with RGB-Depth camera using HMM | |
| CN104361321B (en) | A kind of method for judging the elderly and falling down behavior and balance ability | |
| Zhao et al. | Recognition of human fall events based on single tri-axial gyroscope | |
| CN103927851B (en) | A kind of individualized multi thresholds fall detection method and system | |
| CN108446733A (en) | A kind of human body behavior monitoring and intelligent identification Method based on multi-sensor data | |
| CN106308809A (en) | Method for recognizing gait of thigh amputation subject | |
| Zhang et al. | Age classification base on gait using HMM | |
| EP3459453B1 (en) | Information processing device, information processing method, and information processing program | |
| CN107609477A (en) | It is a kind of that detection method is fallen down with what Intelligent bracelet was combined based on deep learning | |
| CN103164315A (en) | Computer using time prompting method and system based on intelligent video analysis | |
| Hsieh et al. | Machine learning-based fall characteristics monitoring system for strategic plan of falls prevention | |
| CN108717548B (en) | Behavior recognition model updating method and system for dynamic increase of sensors | |
| Capela et al. | Improving classification of sit, stand, and lie in a smartphone human activity recognition system | |
| CN112115827A (en) | Falling behavior identification method based on human body posture dynamic characteristics | |
| CN104586402B (en) | A kind of feature extracting method of physical activity | |
| Jatesiktat et al. | An elderly fall detection using a wrist-worn accelerometer and barometer | |
| Liu et al. | Validation of an accelerometer-based fall prediction model | |
| CN111931568A (en) | Human body falling detection method and system based on enhanced learning | |
| CN115346272A (en) | Real-time tumble detection method based on depth image sequence |
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 |