CN109035696A - One kind falling down detection method based on acceleration transducer - Google Patents
One kind falling down detection method based on acceleration transducer Download PDFInfo
- Publication number
- CN109035696A CN109035696A CN201810643391.7A CN201810643391A CN109035696A CN 109035696 A CN109035696 A CN 109035696A CN 201810643391 A CN201810643391 A CN 201810643391A CN 109035696 A CN109035696 A CN 109035696A
- Authority
- CN
- China
- Prior art keywords
- data
- acceleration
- axis
- feature
- follows
- 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
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)
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
Abstract
The present invention relates to one kind to fall down detection method based on acceleration transducer.Falling down detection method includes: to acquire data using single acceleration transducer, by Threshold Analysis screen it is potential fall down data, feature extraction is carried out to data, and data are carried out with support vector machines to fall down detection.It is provided by the invention to fall down detection method, it is demonstrate,proved by two steps, filters most of daily behavior data, reduce the calculating consumption of server；This method proposes seven feature extraction equations according to daily behavior and the movement differential for falling down behavior, falls down detection method to the classification performance of new category sample to improve.
Description
Technical field
The present invention relates to one kind to fall down detection method based on acceleration transducer.
Background technique
Falling down detection is to acquire the exercise data of user in real time by sensor, and analyze data by algorithm, judges to use
The current motion state in family falls down behavior or daily behavior.Inertial sensor is that data acquisition relatively conventional at present is set
It is standby, including acceleration transducer, magnetometer, gyroscope etc..Existing acquisition scheme mostly uses greatly multiple groups sensor to cooperate
Mode, merge the characteristic of multiple sensors or sensor be worn on to the different location of human body, obtain more features information,
To improve the accuracy of detection.
Currently, falling down detection algorithm is broadly divided into two major class: detection method based on threshold classification be based on machine
The detection method of learning algorithm.Method based on threshold classification is the characteristic value given threshold to extract, and is carrying out falling down inspection
When survey, it is judged to falling down if feature is greater than threshold value, is otherwise determined as daily behavior.The advantages of threshold classification, is to compare machine
Learning algorithm, it is smaller to the calculated performance requirement for calculating equipment, sorting algorithm is run on smart phone so as to realize,
Without detecting collected data upload server, use cost is substantially reduced.And the disadvantage is that classifies is accurate
Rate is lower, and in certain special scenes, threshold classification algorithm perhaps can obtain good classifying quality, but be difficult to find one kind
The algorithm of threshold classification is tried out in scene used.The above problems, the classification method of machine learning can all be coped with one by one, generation
Valence is that have higher performance requirement to calculating equipment, and debug out suitable sorting algorithm also not a duck soup.In machine learning often
Sorting algorithm has k neighbour to learn (KNN), neural network, support vector machine (SVM) etc..Inspection based on machine learning algorithm
Survey method, in addition to the superiority and inferiority of algorithm itself, the quality of feature extraction is particularly important.
Existing scheme existing defects:
1. existing technical solution mostly uses greatly the mutually matched mode of multiple groups sensor, has ignored sensor and wear just
The loss of the property taken and multisensor bring computing resource.In existing detection method, single-sensor is being used, and
When being classified for the more complicated data set of action classification, functional need is had not yet been reached in nicety of grading.
2. the training set of disaggregated model is built usually only comprising limited kind of motor behavior, and in real life, human motion
Behavior can not be exhaustive, and existing detection scheme is more excellent to action classification detection effect present in training set, and for training set
In the classification that is not present, detection effect is poor.
Summary of the invention
The purpose of the present invention is to provide one kind to fall down detection method based on acceleration transducer, and this method passes through two steps
Card filters most of daily behavior data, reduces the calculating consumption of server；This method is according to daily behavior and falls down behavior
Movement differential proposes seven feature extraction equations, falls down detection method to the classification performance of new category sample to improve.
To achieve the above object, the technical scheme is that a kind of fall down detection method based on acceleration transducer, lead to
Cross single acceleration transducer and acquire initial data in real time, by Threshold Analysis screen it is potential fall down data, data are carried out special
Sign is extracted, and is carried out falling down detection to data with support vector machines.
In an embodiment of the present invention, it is described Threshold Analysis screen it is potential fall down data by way of it is as follows:
The Threshold Analysis that each frame data are carried out on smart phone calculates each frame data resultant acceleration, accelerates when closing
When degree is greater than 15g, be judged to having it is potential fall down possible behavioral data, and by totally two seconds data segments pass through before and after the time point
Internet is uploaded to server end, does and further falls down detection；Resultant acceleration calculation formula is as follows:
Wherein axt,ayt,aztThree axis accelerometer is respectively indicated in the output valve of the x, y, z axis of t moment.
In an embodiment of the present invention, before carrying out falling down detection, disaggregated model, specific build process need to first be built are as follows:
Firstly, carrying out low-pass filtering noise reduction and data normalization to initial data；Then, feature extraction is carried out to pretreated data
And the data after feature extraction are used for model training；Finally, the model after training to be used for the detection of new samples data.
In an embodiment of the present invention, as follows to the mode of data progress feature extraction:
The equation of feature extraction is used for using seven, first four equation is time series, take maximum value therein, latter three
Equation is single value；
1) three axis resultant accelerations are the amplitudes of the vector sum of 3-axis acceleration, and abbreviation resultant acceleration, this feature is in certain journey
The severity of human motion is symbolized on degree；The calculation formula of this feature is as follows:
Wherein axt,ayt,aztThree axis accelerometer is respectively indicated in the output valve of the x, y, z axis of t moment；
2) between cosine similarity characterization adjacent data, the angle change between the vector sum of 3-axis acceleration；This feature can
Reflect experimenter in activity, the instantaneous angular of body changes；The calculation formula of this feature is as follows:
Wherein A is 3-axis acceleration vector, i.e. (ax,ay,az)；| | A | | it is 2 norms of 3-axis acceleration, that is, closes and accelerate
Degree；
3) three axis resultant acceleration gradients are the absolute values of the difference of adjacent resultant acceleration；The calculation formula of this feature is as follows:
F3=| | At||-||At+1|| (3)
4) horizontal plane direction resultant acceleration, when person upright, sensor x-axis and z-axis direction are two in horizontal direction
Vertical direction, stationary state reading is approximately zero, and y-axis direction is gravity direction, and stationary state reads approximation 1g；The meter of this feature
It is as follows to calculate formula:
5) resultant acceleration standard deviation can react the fluctuation situation of resultant acceleration；The calculation formula of this feature is as follows:
Wherein AtIndicate the resultant acceleration value of each time point,Indicate the mean value of resultant acceleration；
6) horizontal plane direction resultant acceleration standard deviation；The calculation formula of this feature is as follows:
Wherein AtIndicate the horizontal plane direction resultant acceleration value of each time point,Indicate horizontal plane direction resultant acceleration
Mean value；
7) generation front and back, the cosine similarity of 3-axis acceleration mean value are fallen down；Take it is potential fall down in data, head and the tail each 0.5
The 3-axis acceleration data of second, average, and seek its cosine similarity；The calculation formula of this feature is as follows:
WhereinRespectively indicate the potential x, y, z axle acceleration mean value for falling down 0.5 second before data；
Respectively indicate the potential x, y, z axle acceleration mean value for falling down data trailer 0.5 second.
In an embodiment of the present invention, the potential data of falling down are in initial data using maximum resultant acceleration as midpoint
, when a length of two seconds data slots.
Compared to the prior art, the invention has the following advantages: the feature extraction equation that this motion proposes makes score
Class model is fitted not directed to specific action classification, but is not intended for daily behavior with the behavior two major classes of falling down
It closes, for " new category " sample being not present in training set, " new category " sample and daily behavior can be found out or fall down behavior
Common trait, to be classified；In view of the training set for model training can not exhaustive everything, therefore this motion design
Detection method of falling down be more in line with practical application scene.
Detailed description of the invention
Fig. 1 is that the present invention falls down detection system flow chart.
Fig. 2 is to fall down detection method flow chart the present invention is based on support vector machines.
Specific embodiment
With reference to the accompanying drawing, technical solution of the present invention is specifically described.
The present invention provides one kind to fall down detection method based on acceleration transducer, real-time by single acceleration transducer
Acquire initial data, by Threshold Analysis screen it is potential fall down data, feature extraction is carried out to data, and with support vector machines pair
Data carry out falling down detection.Before carrying out falling down detection, disaggregated model, specific build process need to be first built are as follows: firstly, to original
Data carry out low-pass filtering noise reduction and data normalization；Then, feature extraction is carried out to pretreated data and proposes feature
Data after taking are used for model training；Finally, the model after training to be used for the detection of new samples data.
It is as follows to initial data Threshold Analysis screening mode:
The Threshold Analysis that each frame data are carried out on smart phone calculates each frame data resultant acceleration, accelerates when closing
When degree is greater than 15g, be judged to having it is potential fall down possible behavioral data, and by totally two seconds data segments pass through before and after the time point
Internet is uploaded to server end, does and further falls down detection.Resultant acceleration calculation formula is as follows:
Wherein axt,ayt,aztThree axis accelerometer is respectively indicated in the output valve of the x, y, z axis of t moment；
The mode for carrying out feature extraction to data is as follows:
The equation of feature extraction is used for using seven, first four equation is time series, take maximum value therein, latter three
Equation is single value；
1) three axis resultant accelerations are the amplitudes of the vector sum of 3-axis acceleration, and abbreviation resultant acceleration, this feature is in certain journey
The severity of human motion is symbolized on degree；The calculation formula of this feature is as follows:
Wherein axt,ayt,aztThree axis accelerometer is respectively indicated in the output valve of the x, y, z axis of t moment；
2) between cosine similarity characterization adjacent data, the angle change between the vector sum of 3-axis acceleration；This feature can
Reflect experimenter in activity, the instantaneous angular of body changes；The calculation formula of this feature is as follows:
Wherein A is 3-axis acceleration vector, i.e. (ax,ay,az)；| | A | | it is 2 norms of 3-axis acceleration, that is, closes and accelerate
Degree；
3) three axis resultant acceleration gradients are the absolute values of the difference of adjacent resultant acceleration；The calculation formula of this feature is as follows:
F3=| | At||-||At+1|| (3)
4) horizontal plane direction resultant acceleration, when person upright, sensor x-axis and z-axis direction are two in horizontal direction
Vertical direction, stationary state reading is approximately zero, and y-axis direction is gravity direction, and stationary state reads approximation 1g；The meter of this feature
It is as follows to calculate formula:
5) resultant acceleration standard deviation can react the fluctuation situation of resultant acceleration；The calculation formula of this feature is as follows:
Wherein AtIndicate the resultant acceleration value of each time point,Indicate the mean value of resultant acceleration；
6) horizontal plane direction resultant acceleration standard deviation；The calculation formula of this feature is as follows:
Wherein AtIndicate the horizontal plane direction resultant acceleration value of each time point,Indicate horizontal plane direction resultant acceleration
Mean value；
7) generation front and back, the cosine similarity of 3-axis acceleration mean value are fallen down；Take it is potential fall down in data, head and the tail each 0.5
The 3-axis acceleration data of second, average, and seek its cosine similarity；The calculation formula of this feature is as follows:
WhereinRespectively indicate the potential x, y, z axle acceleration mean value for falling down 0.5 second before data；
Respectively indicate the potential x, y, z axle acceleration mean value for falling down data trailer 0.5 second.
It is described it is potential fall down data be using maximum resultant acceleration as midpoint in initial data, when a length of two seconds data slices
Section.
The following are specific implementation processes of the invention.
This motion design falls down detection system frame as shown in Figure 1, firstly, acquiring in real time by external inertial sensor
The activity data of the elderly, and smart phone is real-time transmitted to by bluetooth.Then, each frame data are carried out on smart phone
Threshold Analysis, when resultant acceleration is greater than 15g, be judged to having it is potential fall down possible behavioral data, and will be before the time point
Totally two seconds data segments are uploaded to server end by internet afterwards, do and further fall down detection.Finally, in server end,
Further behavioral value is done by the machine learning algorithm of more high-class precision, when differentiating result is to fall down behavior, service
Device sounds an alarm, and notifies preset contact person.
In terms of sorting algorithm, using the algorithm of support vector machines, disaggregated model builds process such as Fig. 2 institute for this motion
Show.Firstly, carrying out low-pass filtering noise reduction and data normalization to initial data；Then, feature is carried out to pretreated data
It extracts and the data after feature extraction is used for model training；Finally, the model after training to be used for the detection of new samples.Originally it mentions
The main innovation point of case is characteristic extraction part.Seven introduced below are used for the equation of feature extraction, when first four equation is
Between sequence, take maximum value therein, rear three equations are single value.
1) three axis resultant accelerations are the amplitudes of the vector sum of 3-axis acceleration, and abbreviation resultant acceleration, this feature is in certain journey
The severity of human motion is symbolized on degree；The calculation formula of this feature is as follows:
Wherein axt,ayt,aztThree axis accelerometer is respectively indicated in the output valve of the x, y, z axis of t moment；
2) between cosine similarity characterization adjacent data, the angle change between the vector sum of 3-axis acceleration；This feature can
Reflect experimenter in activity, the instantaneous angular of body changes；The calculation formula of this feature is as follows:
Wherein A is 3-axis acceleration vector, i.e. (ax,ay,az)；| | A | | it is 2 norms of 3-axis acceleration, that is, closes and accelerate
Degree；
3) three axis resultant acceleration gradients are the absolute values of the difference of adjacent resultant acceleration；The calculation formula of this feature is as follows:
F3=| | At||-||At+1|| (3)
4) horizontal plane direction resultant acceleration, when person upright, sensor x-axis and z-axis direction are two in horizontal direction
Vertical direction, stationary state reading is approximately zero, and y-axis direction is gravity direction, and stationary state reads approximation 1g；The meter of this feature
It is as follows to calculate formula:
5) resultant acceleration standard deviation can react the fluctuation situation of resultant acceleration；The calculation formula of this feature is as follows:
Wherein AtIndicate the resultant acceleration value of each time point,Indicate the mean value of resultant acceleration；
6) horizontal plane direction resultant acceleration standard deviation；The calculation formula of this feature is as follows:
Wherein AtIndicate the horizontal plane direction resultant acceleration value of each time point,Indicate horizontal plane direction resultant acceleration
Mean value；
7) generation front and back, the cosine similarity of 3-axis acceleration mean value are fallen down；As stated earlier, adding for feature extraction
Speed data is (i.e. potential to fall down data) in each sample, using maximum resultant acceleration as midpoint, when a length of two seconds data
Segment；It takes in two seconds data slots, each 0.5 second 3-axis acceleration data of head and the tail are averaged, and seek its cosine similarity；
The calculation formula of this feature is as follows:
WhereinRespectively indicate the potential x, y, z axle acceleration mean value for falling down 0.5 second before data；
Respectively indicate the potential x, y, z axle acceleration mean value for falling down data trailer 0.5 second.For sample rate be respectively 50Hz, 100Hz,
Three data sets of 200Hz, 0.5 second each number of axle strong point number are 25,50,100.
Usage mode using the product of the method for the present invention is as follows: sensor is fixed on waist by user, make sensor with
Smart phone is by bluetooth connection, and after the detection switch for opening mobile phone terminal, sensor will acquire user data in real time, and in mobile phone
End and server end are measured in real time data, to monitor whether user falls down.
The above are preferred embodiments of the present invention, all any changes made according to the technical solution of the present invention, and generated function is made
When with range without departing from technical solution of the present invention, all belong to the scope of protection of the present invention.
Claims (5)
1. one kind falls down detection method based on acceleration transducer, which is characterized in that adopted in real time by single acceleration transducer
Collect initial data, by Threshold Analysis screen it is potential fall down data, feature extraction is carried out to data, and with support vector machines logarithm
According to carrying out falling down detection.
2. a kind of acceleration transducer that is based on according to claim 1 falls down detection method, which is characterized in that described to pass through
It is as follows that Threshold Analysis screens the potential mode for falling down data:
The Threshold Analysis that each frame data are carried out on smart phone calculates each frame data resultant acceleration, when resultant acceleration is big
When 15g, be judged to having it is potential fall down possible behavioral data, and by totally two seconds data segments pass through interconnection before and after the time point
Server end is reached on the net, does and further falls down detection；Resultant acceleration calculation formula is as follows:
Wherein axt,ayt,aztThree axis accelerometer is respectively indicated in the output valve of the x, y, z axis of t moment.
3. a kind of acceleration transducer that is based on according to claim 1 falls down detection method, which is characterized in that fallen
Before detecting, disaggregated model, specific build process need to be first built are as follows: firstly, carrying out low-pass filtering noise reduction sum number to initial data
According to normalization；Then, feature extraction is carried out to pretreated data and the data after feature extraction is used for model training；Most
Afterwards, the model after training is used for the detection of new samples data.
4. a kind of acceleration transducer that is based on according to any one of claims 1 to 3 falls down detection method, which is characterized in that
The mode for carrying out feature extraction to data is as follows:
The equation of feature extraction is used for using seven, first four equation to be time series, takes maximum value therein, rear three equations
For single value；
1) three axis resultant accelerations are the amplitudes of the vector sum of 3-axis acceleration, and abbreviation resultant acceleration, this feature is to a certain extent
Symbolize the severity of human motion；The calculation formula of this feature is as follows:
Wherein axt,ayt,aztThree axis accelerometer is respectively indicated in the output valve of the x, y, z axis of t moment；
2) between cosine similarity characterization adjacent data, the angle change between the vector sum of 3-axis acceleration；This feature is able to reflect
In activity, the instantaneous angular of body changes experimenter；The calculation formula of this feature is as follows:
Wherein A is 3-axis acceleration vector, i.e. (ax,ay,az)；| | A | | it is 2 norms of 3-axis acceleration, i.e. resultant acceleration；
3) three axis resultant acceleration gradients are the absolute values of the difference of adjacent resultant acceleration；The calculation formula of this feature is as follows:
F3=| | At||-||At+1|| (3)
4) horizontal plane direction resultant acceleration, when person upright, sensor x-axis and z-axis direction are that two in horizontal direction are vertical
Direction, stationary state reading is approximately zero, and y-axis direction is gravity direction, and stationary state reads approximation 1g；The calculating of this feature is public
Formula is as follows:
5) resultant acceleration standard deviation can react the fluctuation situation of resultant acceleration；The calculation formula of this feature is as follows:
Wherein AtIndicate the resultant acceleration value of each time point,Indicate the mean value of resultant acceleration；
6) horizontal plane direction resultant acceleration standard deviation；The calculation formula of this feature is as follows:
Wherein AtIndicate the horizontal plane direction resultant acceleration value of each time point,Indicate the mean value of horizontal plane direction resultant acceleration；
7) generation front and back, the cosine similarity of 3-axis acceleration mean value are fallen down；Take it is potential fall down in data, head and the tail each 0.5 second
3-axis acceleration data, average, and seek its cosine similarity；The calculation formula of this feature is as follows:
WhereinRespectively indicate the potential x, y, z axle acceleration mean value for falling down 0.5 second before data；Respectively
Indicate the potential x, y, z axle acceleration mean value for falling down data trailer 0.5 second.
5. a kind of acceleration transducer that is based on according to claim 4 falls down detection method, which is characterized in that described potential
Fall down data be using maximum resultant acceleration as midpoint in initial data, when a length of two seconds data slots.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810643391.7A CN109035696A (en) | 2018-06-21 | 2018-06-21 | One kind falling down detection method based on acceleration transducer |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810643391.7A CN109035696A (en) | 2018-06-21 | 2018-06-21 | One kind falling down detection method based on acceleration transducer |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109035696A true CN109035696A (en) | 2018-12-18 |
Family
ID=64610537
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810643391.7A Pending CN109035696A (en) | 2018-06-21 | 2018-06-21 | One kind falling down detection method based on acceleration transducer |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109035696A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112438721A (en) * | 2019-08-30 | 2021-03-05 | 奇酷互联网络科技（深圳）有限公司 | State determination method, electronic device, and computer storage medium |
| CN112949552A (en) * | 2021-03-22 | 2021-06-11 | 浙江大华技术股份有限公司 | Fall detection processing method and device |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102610056A (en) * | 2012-03-16 | 2012-07-25 | 清华大学 | Mobile phone wearing mode-oriented falling event detection system and method |
| CN104125337A (en) * | 2014-07-22 | 2014-10-29 | 厦门美图移动科技有限公司 | Smart phone falling detection and alarming method |
| CN104361361A (en) * | 2014-11-14 | 2015-02-18 | 北京天地弘毅科技有限公司 | Method and system for judging fall through cloud computing and machine learning algorithm |
| CN105632101A (en) * | 2015-12-31 | 2016-06-01 | 深圳先进技术研究院 | Human body anti-tumbling early warning method and system |
-
2018
- 2018-06-21 CN CN201810643391.7A patent/CN109035696A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102610056A (en) * | 2012-03-16 | 2012-07-25 | 清华大学 | Mobile phone wearing mode-oriented falling event detection system and method |
| CN104125337A (en) * | 2014-07-22 | 2014-10-29 | 厦门美图移动科技有限公司 | Smart phone falling detection and alarming method |
| CN104361361A (en) * | 2014-11-14 | 2015-02-18 | 北京天地弘毅科技有限公司 | Method and system for judging fall through cloud computing and machine learning algorithm |
| CN105632101A (en) * | 2015-12-31 | 2016-06-01 | 深圳先进技术研究院 | Human body anti-tumbling early warning method and system |
Non-Patent Citations (1)
| Title |
|---|
| 陈翔，杨明静: "基于SVM 与多数据集的摔倒检测方法研究", 《信息通信》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112438721A (en) * | 2019-08-30 | 2021-03-05 | 奇酷互联网络科技（深圳）有限公司 | State determination method, electronic device, and computer storage medium |
| CN112949552A (en) * | 2021-03-22 | 2021-06-11 | 浙江大华技术股份有限公司 | Fall detection processing method and device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105678222B (en) | A kind of mobile device-based Human bodys' response method | |
| CN103997572B (en) | A kind of step-recording method based on mobile phone acceleration sensor data and device | |
| CN107220617A (en) | Human body attitude identifying system and method | |
| CN105310696B (en) | A kind of fall detection model building method and corresponding fall detection method and device | |
| CN106228200B (en) | Action identification method independent of action information acquisition equipment | |
| Ahmed et al. | An approach to classify human activities in real-time from smartphone sensor data | |
| CN106981174A (en) | A kind of Falls Among Old People detection method based on smart mobile phone | |
| WO2014089238A1 (en) | Gait analysis system and method | |
| CN108549900A (en) | Tumble detection method for human body based on mobile device wearing position | |
| CN103083025A (en) | Gait Analysis Device | |
| CN105877757A (en) | Multi-sensor integrated human motion posture capturing and recognizing device | |
| CN112464738B (en) | Improved naive Bayes algorithm user behavior identification method based on mobile phone sensor | |
| WO2021115064A1 (en) | Fitness exercise recognition method based on wearable sensor | |
| CN107277222A (en) | User behavior state judging method based on mobile phone built-in sensors | |
| CN109035696A (en) | One kind falling down detection method based on acceleration transducer | |
| CN111582361A (en) | Human behavior recognition method based on inertial sensor | |
| Chen et al. | Detection of falls with smartphone using machine learning technique | |
| CN107016411B (en) | Data processing method and device | |
| CN111603750A (en) | Motion capture recognition evaluation system and method based on edge calculation | |
| CN107239147A (en) | A kind of human body context aware method based on wearable device, apparatus and system | |
| CN105551191A (en) | Falling detection method | |
| Dwiyantoro et al. | A simple hierarchical activity recognition system using a gravity sensor and accelerometer on a smartphone | |
| Cao et al. | ActiRecognizer: Design and implementation of a real-time human activity recognition system | |
| Miyamoto et al. | Human activity recognition system including smartphone position | |
| CN115393956A (en) | CNN-BilSTM fall detection method for improving attention mechanism |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181218 |
|
| RJ01 | Rejection of invention patent application after publication |