CN109166275A - A kind of tumble detection method for human body based on acceleration transducer - Google Patents
A kind of tumble detection method for human body based on acceleration transducer Download PDFInfo
- Publication number
- CN109166275A CN109166275A CN201811112919.4A CN201811112919A CN109166275A CN 109166275 A CN109166275 A CN 109166275A CN 201811112919 A CN201811112919 A CN 201811112919A CN 109166275 A CN109166275 A CN 109166275A
- Authority
- CN
- China
- Prior art keywords
- tumble
- acceleration
- data
- human body
- behavior
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
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
Landscapes
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Gerontology & Geriatric Medicine (AREA)
- Business, Economics & Management (AREA)
- Emergency Management (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Psychiatry (AREA)
- Psychology (AREA)
- Social Psychology (AREA)
- Emergency Alarm Devices (AREA)
- Alarm Systems (AREA)
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
Abstract
The invention discloses a kind of tumble detection method for human body based on acceleration transducer, are related to processing of biomedical signals technical field.Human body fall detection algorithm of this method based on dual confirmation, level-one fall detection therein act (ADL) and tumble acceleration samples data according to daily behavior gathered in advance, extract threshold set；Then the ADL acceleration information of acquisition human body, the variance for extracting this group of data update the pre- tumble behavior asset pricing for being directed to the user as dynamic threshold part in real time；Set the level-one detection that human body tumble behavior is carried out using upper threshold value as human body tumble judgment criteria；During level-one fall detection, pre- tumble behavior occurs once detecting, sends the human body behavioral data started of ts before pre- tumble behavior to the second level fall detection judgement carried out on the server of nearly data source based on SVM；The threshold set for finally determining human body tumble event according to dual tumble judging result, and dynamically updating fall detection algorithm is used for the fall detection of the subsequent user.
Description
Technical field
The present invention relates to processing of biomedical signals technical fields, and in particular to a kind of human body based on acceleration transducer
Fall detection method.
Background technique
With the sustainable growth of world population, medical system is constantly improve, and pace of population aging constantly speeds, China
It is equally faced with aging outstanding problem, old solitary people is more and more, and empty nest phenomenon is also more and more obvious.However, old solitary people
Health and safety problem gradually become severe social concern.According to statistics, the whole world is more than every year 65 there are about one third
The elderly in year once fell, and fell and be usually associated with serious body and psychological injury, and then give family and society
It increases burden.If in time, the tumble event of accurate detection the elderly and issuing alarm and can reduce tumble pair to greatest extent
The harm of the elderly's bring, has very important effect to living on one's own life for the elderly.
In human body fall detection research, the fall detection method based on acceleration transducer is most commonly seen detection hand
One of section.During human body is fallen, significant changes can occur for athletic posture, for example, human body state of weightlessness, acutely
Hit etc., acceleration transducer equipment is worn by human body, acquires body motion information in real time, is calculated in conjunction with specific fall detection
Method can realize the detection to human body tumble state.For example, in Patent No. CN201710796479.8 Chinese patent, by adopting
Collect physical activity acceleration information, converts thereof into angle value to calculate angle gradient data, and then extract inclination angle gradient variance
As characteristic value, realization is compared based on dual threshold with the threshold value that latter two moment occurs of falling by choosing to fall
The fall detection algorithm of human body.
In the Chinese patent of Patent No. CN201711268665.0, is extracted and accelerated using training sample acceleration information
Spend signal vector amplitude peak, the most value difference value of acceleration signal vector magnitude, acceleration signal vector magnitude standard deviation and
Relative angle changing value, in conjunction with two points of decision thresholds of K-means clustering method training characteristic value, by real-time inspection of falling
Raw information when survey is compared the judgement realized and fallen to human body with characteristic threshold value after carrying out feature extraction.
In the Chinese patent of Patent No. CN201711489128.9, pass through acceleration transducer, gyroscope, magnetometer
The acceleration information that human hands are swung in device and baroceptor acquisition walking, the valley value by extracting acceleration are averaged
It is worth the average value of the time between the valley value a1 of average value acceleration and the crest value a2 of acceleration of the crest value of acceleration
The average value of time between the crest value a2 of acceleration and the valley value a ' 1 of next acceleration is constructed as characteristic parameter
SVM classifier is trained, and then realizes the fall detection of human body.
In the Chinese patent of Patent No. CN201210586385.5, entire tumble process is divided into 4 stages: perpendicular
It encounter stage and lies low after falling and nearly quiescent phase during straight stance, the fall stage fallen early period, tumble, wherein
It falls and encounter stage uses setting acceleration rate threshold method to realize judgement, and combine the standing of angle analysis human body and lying status
Realize fall detection.
Although the above fall detection method is able to detect that the tumble behavior of most of human body, but due to the individual difference of user
Different and current true tumble data deficiencies, so that the accuracy of fall detection is difficult to ensure.So although existing tumble is examined
There are many survey method, but are not well positioned to meet the requirement of the pinpoint accuracy of fall detection.
Summary of the invention
The purpose of the present invention is in view of the above deficiencies, propose a kind of human body fall detection side based on acceleration transducer
Method drops in detection report by mistake caused by individual difference due to ignoring, fail to report problem and really fall this method solve current human
Insufficient problem reciprocal.
The present invention specifically adopts the following technical scheme that
A kind of tumble detection method for human body based on acceleration transducer, this method are based on detection system, detection system packet
Include data acquisition module interconnected, threshold value extraction module, fall detection module and alarm and threshold value update module, alarm with
Threshold value update module is connected with the fall detection module constitutes a circuit, specifically includes:
Step 1: threshold set is extracted；The acceleration samples data of ADL gathered in advance and tumble are pre-processed and mentioned
Pre- tumble behavior asset pricing TH1, tumble collision threshold TH2 are taken, fall posture threshold value TH4 after restoring state threshold TH3 and falling；It is logical
Cross calculate in acceleration samples data gathered in advance the sum of the mean value of all ADL data and standard deviation judge as pre- tumble it is quiet
State threshold portion；The tumble data set in acceleration samples data gathered in advance is analyzed, is calculated separately from acceleration information paddy
It is worth difference, relative acceleration value and the final angle value of peak value as feature, extracts the tumble collision threshold TH2 of human body, fall
Posture threshold value TH4 after restoring state threshold TH3 and falling；
Step 2: the update of threshold value TH1 is carried out for the tumble behavior of user；By acquiring actual user in real time specified
Acceleration information under ADL movement, and calculate this group of daily behavior and act the standard deviation of lower data as dynamic threshold part, knot
It further extracts and updates pre- tumble behavior asset pricing TH1 in the static threshold part closed in step 1；
Step 3: the judgement of level-one lightweight fall detection is carried out；The real time acceleration data of user are acquired and calculate in real time,
Gradually judge pre- tumble behavior, tumble collision behavior, tumble recovery behavior and the final carriage of human body, and then whether determines human body
It falls, while detecting human body generation tumble behavior moment, starting wireless transmission；The ts moment before the moment is started directly
Data are sent to the server at nearly data source in real time to be further processed in the algorithm finish time, meanwhile, level-one light weight
The corresponding alarm signal of grade fall detection is sent to server end together；
Step 4: the fall detection judgement based on SVM, the acceleration using trained SVM classifier to receiving are carried out
Degree is according to progress fall detection；If result is non-tumble behavior, do not alarm, if result is tumble behavior, alarms；
Step 5: carrying out dual confirmation and threshold value updates, and is judged according to the two-stage fall detection in step 3 and step 4
As a result Comprehensive affirming is carried out, hierarchical detection is alarmed, then confirms that human body is fallen, while by the number in this section of period
According to calculating and updating TH2, TH3, TH4 again as tumble data, for the fall detection of the subsequent user, if human body behavior is
Daily behavior movement, which is calculated again and updates TH1.
Preferably, acceleration samples data gathered in advance described in the step 1 are by acquiring not the same year in advance
User under the requirements such as age, gender, height and body wear include acceleration transducer equipment according to specified ADL movement and
The acceleration information or existing sample database of tumble movement.
Preferably, the pre- tumble behavior asset pricing TH1 is to calculate ADL data in acceleration samples data gathered in advance
The static threshold of acquisition and the dynamic threshold that the acceleration information of acquisition actual user ADL obtains in real time, and sum of the two is taken to obtain
?.
Preferably, the level-one lightweight fall detection in the step three uses level Four state judgment mode step by step, leads to
It crosses whether the acceleration samples data that judgement acquires in real time are greater than TH1, if it is not, then currently tumble behavior does not occur for judgement, lays equal stress on
It is new to execute lightweight fall detection；If so, pre- tumble behavior currently has occurred in judgement, continue to judge acceleration information from valley
Whether the difference to peak value is greater than TH2, if it is not, then currently tumble behavior does not occur for judgement, lightweight fall detection again；If so,
Then judge that hard hit has occurred in human body, continue to judge whether relative acceleration value is less than TH3, if it is not, then judgement is not sent out currently
Raw tumble behavior, re-executes lightweight fall detection；If so, judging that human body is in metastable state, continue judgement most
Whether whole angle is less than TH4, if it is not, then currently tumble behavior does not occur for judgement, re-executes lightweight fall detection；If so,
Then judge that tumble behavior occurs for human body.
Preferably, trained SVM classifier is to utilize above-mentioned acceleration samples number gathered in advance in the step 4
It is obtained according to training, by searching for the period of each group of training data, and calculates the mean value of each cycle data, standard deviation, acceleration
Difference of the valley to peak value, acceleration trough to peak time-interval, the difference of acceleration peak value to valley, acceleration wave crest to trough
Time interval, angle behind acceleration mean value and standard deviation and specified time interval in specified time interval after second trough
Value is trained extracted feature as characteristic value collection construction SVM classifier.
Preferably, the threshold value update of TH2 in the step 5, TH3, TH4 are when dual confirmation result is row of falling
For when, recalculate using data at this time as tumble data to update and realize, the update of the threshold value of TH1 is when dual confirmation knot
When fruit is ADL movement, data at this time are recalculated into update as ADL data and are realized.
Preferably, the threshold value of the TH2, TH3, TH4 update, and are the tumble numbers before being fused to the new data of addition
According to collection, TH2 is equally extracted using confidence interval Mathematical Method again, TH3, TH4 are as user's fall detection threshold value；
The threshold value of TH1, which updates, to be realized, is the ADL data set before the data fusion that will be newly added arrives, is recalculated the static state of data set
Threshold value and dynamic threshold method are extracted again to be obtained.
Preferably, the data acquisition module specifies the acceleration samples of ADL movement and tumble movement for acquiring user
Data, and the real-time acquisition of the acceleration information of user action in actual use；
The threshold value extraction module, connect with acquisition module, for analyzing and extracting acquisition ADL acceleration information collection in advance
Static threshold and the real-time ADL acceleration information of user dynamic threshold part, and then extract TH1, while analyze in advance acquisition
Tumble action acceleration data set extracts TH2, TH3, TH4；
The fall detection module, connect with data acquisition module and threshold value extraction module, using based on dual confirmation
Two-stage tumble judgment mode realizes the real-time detection of the final tumble behavior of user；
It is described alarm be connected with threshold value update module and fall detection module constitute a circuit, for tumble alarm and
The update of TH1, TH2, TH3, TH4 threshold set.
The invention has the following beneficial effects:
Thought is updated using dynamic threshold value, is solved because individual difference causes fall detection accuracy problem；
It is the dual determination of lightweight tumble judgement and the second level fall detection judgement based on SVM using level-one fall detection
Fall monitoring method, update given threshold using true tumble data, solve that current true tumble data difficulty obtains asks
Topic, and the accuracy of fall detection is improved, meet the high-precision requirement of human body fall detection.
Real-time, pinpoint accuracy fall detection is carried out to user in the case where human body is fallen, it can be to user reality
Timely, effective relief is applied, and then can guarantee the personal safety of user.
Detailed description of the invention
Fig. 1 is the flow chart of the tumble detection method for human body based on acceleration transducer；
Fig. 2 is the specific flow chart of step S3 in the tumble detection method for human body based on acceleration transducer；
Fig. 3 is the human body fall detection system theory structure schematic diagram based on acceleration transducer.
Specific embodiment
ADL: daily behavior movement；
SVM classifier: being the identification and classification device defined by Optimal Separating Hyperplane, i.e., the training sample of given one group of tape label,
Algorithm will export an optimal hyperlane and classify to new samples (test sample).
A specific embodiment of the invention is described further in the following with reference to the drawings and specific embodiments:
As shown in Figure 1-Figure 3, a kind of tumble detection method for human body based on acceleration transducer, this method are based on detection system
System, detection system include data acquisition module interconnected, threshold value extraction module, fall detection module and alarm with threshold value more
New module, alarm is connected with threshold value update module with the fall detection module constitutes a circuit, specifically includes:
Step 1: threshold set is extracted；The acceleration samples data of ADL gathered in advance and tumble are pre-processed and mentioned
Pre- tumble behavior asset pricing TH1, tumble collision threshold TH2 are taken, fall posture threshold value TH4 after restoring state threshold TH3 and falling；It is logical
Cross calculate in acceleration samples data gathered in advance the sum of the mean value of all ADL data and standard deviation judge as pre- tumble it is quiet
State threshold portion；The tumble data set in acceleration samples data gathered in advance is analyzed, is calculated separately from acceleration information paddy
It is worth difference, relative acceleration value and the final angle value of peak value as feature, extracts the tumble collision threshold TH2 of human body, fall
Posture threshold value TH4 after restoring state threshold TH3 and falling；
The main acquisition modes of acceleration samples data gathered in advance are as follows: enabling each user wear in waist includes to add
The equipment of velocity sensor completes specified ADL and tumble movement, acquires the user's during completing required movement
Acceleration information, wherein the user for participating in data acquisition mainly includes all ages and classes, gender, height, the use under the requirements such as weight
Family.Specified ADL movement mainly includes standing, and walks, sits, and jump squats down, lies down, walk-sit, walks-lie, and crouching-is stood, stair climbing etc.,
Specified tumble movement mainly includes falling forward, falls back, falls, trip to side.
Step 2: the update of threshold value TH1 is carried out for the tumble behavior of user；By acquiring actual user in real time specified
Acceleration information under ADL movement, and calculate this group of daily behavior and act the standard deviation of lower data as dynamic threshold part, knot
It further extracts and updates pre- tumble behavior asset pricing TH1 in the static threshold part closed in step 1；
Step 3: the judgement of level-one lightweight fall detection is carried out；The real time acceleration data of user are acquired and calculate in real time,
Gradually judge pre- tumble behavior, tumble collision behavior, tumble recovery behavior and the final carriage of human body, and then whether determines human body
It falls, while detecting human body generation tumble behavior moment, starting wireless transmission；The ts moment before the moment is started directly
Data are sent to the server at nearly data source in real time to be further processed in the algorithm finish time, meanwhile, level-one light weight
The corresponding alarm signal of grade fall detection is sent to server end together；
Step 4: the fall detection judgement based on SVM, the acceleration using trained SVM classifier to receiving are carried out
Degree is according to progress fall detection；If result is non-tumble behavior, do not alarm, if result is tumble behavior, alarms；
Step 5: carrying out dual confirmation and threshold value updates, and is judged according to the two-stage fall detection in step 3 and step 4
As a result Comprehensive affirming is carried out, hierarchical detection judges to alarm, then confirms that human body is fallen, while by the number in this section of period
According to calculating and updating TH2, TH3, TH4 again as tumble data, for the fall detection of the subsequent user, if human body behavior is
Daily behavior movement, which is calculated again and updates TH1.
Acceleration samples data gathered in advance described in step 1 are by acquiring all ages and classes, gender, height in advance
Worn with the user under the requirements such as body include acceleration transducer equipment according to specified ADL movement and tumble movement plus
Speed data or existing sample database.
Pre- tumble behavior asset pricing TH1 is to calculate the static threshold that ADL data obtain in acceleration samples data gathered in advance
The dynamic threshold that the acceleration information of value and acquisition actual user ADL in real time obtain, and sum of the two is taken to obtain.
Level-one lightweight fall detection in step 3 uses level Four state judgment mode step by step, passes through and judges acquisition in real time
Acceleration samples data whether be greater than TH1, if it is not, then currently tumble behavior does not occur for judgement, and re-execute lightweight and fall
It detects；If so, pre- tumble behavior currently has occurred in judgement, continue whether to judge difference of the acceleration information from valley to peak value
Greater than TH2, if it is not, then currently tumble behavior does not occur for judgement, lightweight fall detection again；If so, judging human body
Hard hit, continues to judge whether relative acceleration value is less than TH3, if it is not, then currently tumble behavior does not occur for judgement, again
Execute lightweight fall detection；If so, judging that human body is in metastable state, continue to judge whether final angle is less than
TH4 re-executes lightweight fall detection if it is not, then currently tumble behavior does not occur for judgement；If so, judging human body
Tumble behavior.
Trained SVM classifier is obtained using above-mentioned acceleration samples data training gathered in advance in step 4,
By searching for the period of each group of training data, and calculate the mean value of each cycle data, standard deviation, acceleration valley to peak value it
Difference, acceleration trough to peak time-interval, the difference of acceleration peak value to valley, acceleration wave crest to decrease amount interval, the
Angle value behind acceleration mean value and standard deviation and specified time interval in specified time interval after two troughs, will be extracted
9 features are trained as characteristic value collection construction SVM classifier.
TH2 in step 5, TH3, the threshold value update of TH4 be when dual confirmation result is tumble behavior, will at this time
Data recalculate to update as tumble data and realize, the update of the threshold value of TH1 is when dual confirmation result is ADL dynamic
When making, data at this time are recalculated into update as ADL data and are realized.
The threshold value of TH2, TH3, TH4 update, and are the tumble data sets before being fused to the new data of addition, same to use
Confidence interval Mathematical Method extracts TH2 again, and TH3, TH4 are as user's fall detection threshold value；The threshold value of TH1 updates real
It is existing, it is the ADL data set before the data fusion that will be newly added arrives, recalculates static threshold and the dynamic threshold side of data set
Method is extracted again and is obtained.
Data acquisition module specifies the acceleration samples data and reality of ADL movement and tumble movement for acquiring user
The real-time acquisition of the acceleration information of user action in the use process of border；
Threshold value extraction module, connect with acquisition module, for analyzing and extracting acquisition daily behavior action acceleration in advance
The static threshold (TH1 component part) of data set and the dynamic threshold part of the real-time ADL action acceleration data of user, Jin Erti
TH1 is taken, while analyzing acquisition tumble action acceleration data set in advance, extracts TH2, TH3, TH4；
Fall detection module is connect with data acquisition module and threshold value extraction module, using the two-stage based on dual confirmation
Tumble judgment mode realizes the real-time detection of the final tumble behavior of user；
Alarm is connected with threshold value update module and fall detection module constitutes a circuit, is used for fall alarm and TH1,
The update of TH2, TH3, TH4 threshold set.
Certainly, the above description is not a limitation of the present invention, and the present invention is also not limited to the example above, this technology neck
The variations, modifications, additions or substitutions that the technical staff in domain is made within the essential scope of the present invention also should belong to of the invention
Protection scope.
Claims (8)
1. a kind of tumble detection method for human body based on acceleration transducer, this method is based on detection system, and detection system includes
Data acquisition module, threshold value extraction module, fall detection module and alarm interconnected and threshold value update module, alarm and threshold
Value update module is connected with the fall detection module constitutes a circuit, which is characterized in that specifically includes:
Step 1: threshold set is extracted；The acceleration samples data of ADL gathered in advance and tumble are pre-processed and extracted pre-
Tumble behavior asset pricing TH1, tumble collision threshold TH2, fall posture threshold value TH4 after restoring state threshold TH3 and falling；Pass through meter
Calculating the sum of the mean value of all ADL data and standard deviation in acceleration samples data gathered in advance is used as pre- fall to judge static threshold
Value part；Analyze the tumble data set in acceleration samples data gathered in advance, calculate separately from acceleration information valley to
Difference, relative acceleration value and the final angle value of peak value extract the tumble collision threshold TH2 of human body as feature, fall extensive
Posture threshold value TH4 after answering state threshold TH3 and falling；
Step 2: the update of threshold value TH1 is carried out for the tumble behavior of user；By acquiring actual user in real time in specified ADL
The lower acceleration information of movement, and calculate the standard deviations of this group of daily behavior movement time data as dynamic threshold part, in conjunction with
It further extracts and updates pre- tumble behavior asset pricing TH1 in static threshold part in step 1；
Step 3: the judgement of level-one lightweight fall detection is carried out；The real time acceleration data of user are acquired and calculate in real time, gradually
Judge pre- tumble behavior, tumble collision behavior, tumble recovery behavior and the final carriage of human body, and then determines whether human body occurs
It falls, while detecting human body generation tumble behavior moment, starting wireless transmission；The ts moment before the moment is started until should
Data are sent to the server at nearly data source in real time to be further processed in algorithm finish time, meanwhile, level-one lightweight falls
It detects corresponding alarm signal and is sent to server end together；
Step 4: the fall detection judgement based on SVM is carried out, using trained SVM classifier to the acceleration degree received
According to progress fall detection；If result is non-tumble behavior, do not alarm, if result is tumble behavior, alarms；
Step 5: carrying out dual confirmation and threshold value updates, according to the two-stage fall detection judging result in step 3 and step 4
Comprehensive affirming is carried out, hierarchical detection is alarmed, then confirms that human body is fallen, while the data in this section of period being made
TH2, TH3, TH4 are calculated and update again for tumble data, for the fall detection of the subsequent user, if human body behavior is daily
The data are calculated again and update TH1 by behavior act.
2. a kind of tumble detection method for human body based on acceleration transducer as described in claim 1, which is characterized in that described
Acceleration samples data gathered in advance described in step 1 are wanted by acquiring all ages and classes, gender, height and body etc. in advance
User's wearing under asking includes that the equipment of acceleration transducer is acted according to specified daily behavior movement (ADL) and tumble
Acceleration information or existing sample database.
3. a kind of tumble detection method for human body based on acceleration transducer as described in claim 1, which is characterized in that described
Pre- tumble behavior asset pricing TH1 be to calculate the static threshold and reality that ADL data obtain in acceleration samples data gathered in advance
When acquisition actual user ADL the dynamic threshold that obtains of acceleration information, and sum of the two is taken to obtain.
4. a kind of tumble detection method for human body based on acceleration transducer as described in claim 1, which is characterized in that described
The step of three in level-one lightweight fall detection using level Four state judgment mode step by step, pass through the acceleration that judgement acquires in real time
Whether degree sample data is greater than TH1, if it is not, then currently tumble behavior does not occur for judgement, and re-executes lightweight fall detection；
If so, pre- tumble behavior currently has occurred in judgement, continue to judge whether difference of the acceleration information from valley to peak value is greater than
TH2, if it is not, then currently tumble behavior does not occur for judgement, lightweight fall detection again；If so, judging that play has occurred in human body
Strong shock continues to judge whether relative acceleration value is less than TH3, if it is not, then currently tumble behavior does not occur for judgement, re-executes
Lightweight fall detection；If so, judging that human body is in metastable state, continue to judge whether final angle is less than TH4, if
No, then currently tumble behavior does not occur for judgement, re-executes lightweight fall detection；If so, judging that row of falling occurs for human body
For.
5. a kind of tumble detection method for human body based on acceleration transducer as described in claim 1, which is characterized in that described
Trained SVM classifier is obtained using above-mentioned acceleration samples data training gathered in advance in step 4, passes through search
The period of each group of training data, and the mean value of each cycle data is calculated, standard deviation, the difference of acceleration valley to peak value accelerates
Trough is spent to peak time-interval, the difference of acceleration peak value to valley, acceleration wave crest to decrease amount interval, second trough
Angle value behind acceleration mean value and standard deviation and specified time interval in specified time interval afterwards, using extracted feature as
Characteristic value collection construction SVM classifier is trained.
6. a kind of tumble detection method for human body based on acceleration transducer as described in claim 1, which is characterized in that described
TH2 in step 5, TH3, the threshold value update of TH4 are to make data at this time when dual confirmation result is tumble behavior
Recalculate to update for tumble data and realize, the update of the threshold value of TH1 be when dual confirmation result is ADL movement, will
Data at this time recalculate update as ADL data and realize.
7. a kind of tumble detection method for human body based on acceleration transducer as claimed in claim 6, which is characterized in that described
TH2, it is the tumble data set before being fused to the new data of addition that the threshold value of TH3, TH4, which updates, equally uses confidence area
Between Mathematical Method extract TH2 again, TH3, TH4 are as user's fall detection threshold value；The threshold value of TH1, which updates, to be realized, is
ADL data set before the data fusion being newly added is arrived recalculates the static threshold and dynamic threshold method weight of data set
New extract is obtained.
8. a kind of tumble detection method for human body based on acceleration transducer as described in claim 1, which is characterized in that described
Data acquisition module, for acquiring the acceleration samples data and actually used that user specifies ADL movement and tumble to act
The real-time acquisition of the acceleration information of user action in journey；
The threshold value extraction module, connect with acquisition module, acquires the quiet of ADL acceleration information collection in advance for analyzing and extracting
The dynamic threshold part of state threshold value and the real-time ADL acceleration information of user, and then TH1 is extracted, while analyzing acquisition in advance and falling
Action acceleration data set extracts TH2, TH3, TH4；
The fall detection module, connect with data acquisition module and threshold value extraction module, using the two-stage based on dual confirmation
Tumble judgment mode realizes the real-time detection of the final tumble behavior of user；
The alarm is connected with threshold value update module and fall detection module constitutes a circuit, is used for fall alarm and TH1,
The update of TH2, TH3, TH4 threshold set.
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 true CN109166275A (en) | 2019-01-08 |
| CN109166275B 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) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110017834A (en) * | 2019-04-15 | 2019-07-16 | 歌尔科技有限公司 | Method, equipment and storage medium are determined using object |
| CN110322662A (en) * | 2019-05-05 | 2019-10-11 | 北京理工大学 | Based on physiology and the monitoring of the pose of behavioural characteristic and alarm bracelet of falling |
| CN110738822A (en) * | 2019-10-24 | 2020-01-31 | 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 |
| CN111998829A (en) * | 2020-10-30 | 2020-11-27 | 四川写正智能科技有限公司 | Method for judging read-write posture based on sensor |
| WO2021105390A1 (en) * | 2019-11-29 | 2021-06-03 | Koninklijke Philips N.V. | Personalized fall detector |
| 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 |
| CN116439694A (en) * | 2023-06-14 | 2023-07-18 | 深圳市魔样科技有限公司 | Intelligent watch dynamic data monitoring method based on motion model training |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160100776A1 (en) * | 2007-10-12 | 2016-04-14 | BioSensics LLC | Fall detection and fall risk detection systems and methods |
| CN105654666A (en) * | 2016-02-25 | 2016-06-08 | 福州大学 | Old people falling monitoring method based on MEMS sensor |
| CN106781278A (en) * | 2017-02-04 | 2017-05-31 | 华南理工大学 | A kind of fall detection method and device based on Fusion |
| CN106910314A (en) * | 2017-02-03 | 2017-06-30 | 同济大学 | A kind of personalized fall detection method based on the bodily form |
| CN108257352A (en) * | 2017-12-30 | 2018-07-06 | 广州柏颐信息科技有限公司 | A kind of fall detection method for early warning based on intelligent wearable device |
-
2018
- 2018-09-25 CN CN201811112919.4A patent/CN109166275B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160100776A1 (en) * | 2007-10-12 | 2016-04-14 | BioSensics LLC | Fall detection and fall risk detection systems and methods |
| CN105654666A (en) * | 2016-02-25 | 2016-06-08 | 福州大学 | Old people falling monitoring method based on MEMS sensor |
| CN106910314A (en) * | 2017-02-03 | 2017-06-30 | 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 |
| CN108257352A (en) * | 2017-12-30 | 2018-07-06 | 广州柏颐信息科技有限公司 | A kind of fall detection method for early warning based on intelligent wearable device |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110017834A (en) * | 2019-04-15 | 2019-07-16 | 歌尔科技有限公司 | Method, equipment and storage medium are determined using object |
| CN110322662A (en) * | 2019-05-05 | 2019-10-11 | 北京理工大学 | Based on physiology and the monitoring of the pose of behavioural characteristic and alarm bracelet of falling |
| CN110738822A (en) * | 2019-10-24 | 2020-01-31 | 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 |
| WO2021105390A1 (en) * | 2019-11-29 | 2021-06-03 | Koninklijke Philips N.V. | Personalized fall detector |
| US11847903B2 (en) | 2019-11-29 | 2023-12-19 | Lifeline Systems Company | Personalized fall detector |
| CN112762895B (en) * | 2020-10-30 | 2022-07-15 | 四川写正智能科技有限公司 | Method for judging read-write posture based on sensor |
| CN111998829A (en) * | 2020-10-30 | 2020-11-27 | 四川写正智能科技有限公司 | Method for judging read-write posture based on sensor |
| 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 |
| CN112762895A (en) * | 2020-10-30 | 2021-05-07 | 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 |
| CN116439694A (en) * | 2023-06-14 | 2023-07-18 | 深圳市魔样科技有限公司 | Intelligent watch dynamic data monitoring method based on motion model training |
| 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 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109166275B (en) | 2020-08-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109166275A (en) | A kind of tumble detection method for human body based on acceleration transducer | |
| CN103927851B (en) | A kind of individualized multi thresholds fall detection method and system | |
| US11527140B2 (en) | Detecting falls using a mobile device | |
| CN105561567B (en) | A kind of meter step and motion state apparatus for evaluating | |
| US11282361B2 (en) | Detecting falls using a mobile device | |
| US11282362B2 (en) | Detecting falls using a mobile device | |
| US20170188894A1 (en) | System and method for sensing and responding to fatigue during a physical activity | |
| CN106334307B (en) | A kind of swimming monitoring method | |
| US20110246123A1 (en) | Personal status monitoring | |
| Zhao et al. | Recognition of human fall events based on single tri-axial gyroscope | |
| US20200342736A1 (en) | Detecting Falls Using A Mobile Device | |
| Anastasopoulou et al. | Classification of human physical activity and energy expenditure estimation by accelerometry and barometry | |
| CN106725383A (en) | Sleep state judgement system and method based on action and heart rate | |
| CN106456024A (en) | Resting heart rate monitor system | |
| CN108814617A (en) | Freezing of gait recognition methods and device and gait detector | |
| Jatesiktat et al. | An elderly fall detection using a wrist-worn accelerometer and barometer | |
| CN109805935A (en) | A kind of intelligent waistband based on artificial intelligence hierarchical layered motion recognition method | |
| CN108717548B (en) | Behavior recognition model updating method and system for dynamic increase of sensors | |
| CN111931568A (en) | Human body falling detection method and system based on enhanced learning | |
| CN104586402A (en) | Feature extracting method for body activities | |
| CN105551191B (en) | A kind of fall detection method | |
| Aubol et al. | Tibial acceleration reliability and minimal detectable difference during overground and treadmill running | |
| Xie et al. | ART: adaptive and real-time fall detection using COTS smart watch | |
| CN111626273B (en) | Fall behavior recognition system and method based on atomic action time sequence characteristics | |
| Pan et al. | Evaluation of hemiplegic gait based on plantar pressure and inertial sensors |
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 |