CN108648417A - Raspberry Pi old man based on deep learning falls down detecting system - Google Patents
Raspberry Pi old man based on deep learning falls down detecting system Download PDFInfo
- Publication number
- CN108648417A CN108648417A CN201810598145.4A CN201810598145A CN108648417A CN 108648417 A CN108648417 A CN 108648417A CN 201810598145 A CN201810598145 A CN 201810598145A CN 108648417 A CN108648417 A CN 108648417A
- Authority
- CN
- China
- Prior art keywords
- data
- acceleration
- raspberry
- axis
- deep learning
- 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
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
- G06F18/2135—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on approximation criteria, e.g. principal component analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
- G06F3/01—Input arrangements or combined input and output arrangements for interaction between user and computer
- G06F3/03—Arrangements for converting the position or the displacement of a member into a coded form
- G06F3/033—Pointing devices displaced or positioned by the user, e.g. mice, trackballs, pens or joysticks; Accessories therefor
- G06F3/0346—Pointing devices displaced or positioned by the user, e.g. mice, trackballs, pens or joysticks; Accessories therefor with detection of the device orientation or free movement in a 3D space, e.g. 3D mice, 6-DOF [six degrees of freedom] pointers using gyroscopes, accelerometers or tilt-sensors
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/02—Preprocessing
- G06F2218/04—Denoising
- G06F2218/06—Denoising by applying a scale-space analysis, e.g. using wavelet analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Gerontology & Geriatric Medicine (AREA)
- Data Mining & Analysis (AREA)
- Emergency Management (AREA)
- Business, Economics & Management (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Computation (AREA)
- Human Computer Interaction (AREA)
- Psychiatry (AREA)
- Psychology (AREA)
- Social Psychology (AREA)
- Alarm Systems (AREA)
Abstract
The invention discloses the Raspberry Pi old men based on deep learning to fall down detecting system, includes two 3-axis acceleration sensors being placed in parallel in each collector, collector is individually positioned in the waistband, trouser pocket, coat pocket position of picker；It, will alert when acceleration, angular speed variation are more than a certain range；Include redundancy in data, data is reduced to from 150 dimensions within 100 dimensions, then the data of behavior state are trained using DBN deep neural networks, the network model of trained generation is for detecting instant behavior state.The beneficial effects of the invention are as follows fall down testing principle be based on DBN deep neural networks, can effectively prevent simulation, such as when equipment be freely falling body to ground when can't be mistaken for falling down；Raspberry Pi hardware device is cheap, can preferably be promoted.
Description
Technical field
The invention belongs to technical field of electronic devices, it is related to the Raspberry Pi old man based on deep learning and falls down detecting system.
Background technology
Increasingly huge old group has become focus of people's attention.The features such as inconvenient due to the elderly's body movement,
The 4th reason for having become China's casualties is fallen down, and accidental falls are the major health risks of over-65s crowd.I
The related scholar of state carried out research to the elderly's problem of falling down, but the product of research institute's production is mainly crutch, walk helper etc..This
Although a little products can reduce the probability that old man falls down, it effectively can not be rescued when old man falls down at the first time
Shield.Therefore, the application that this research is combined based on traditional product with the Internet of things era, inquire into out its to the elderly when falling down not
It will appear the result of delay treatment opportunity problem.This system is for statistical analysis to gathered data, can be to senior activity's situation
Society research provides effectively reference.Old man's tumbling alarm system that we are dedicated to solving safety problem is highly desired, in raspberry
On the basis of group's exploitation version, it is equipped with acceleration transducer ADXL345, good movement can be realized by putting in network by depth
State-detection, to triggering warning function when falling down；Accident fallen down also for the old man to take place frequently there is prevention to a certain extent
And adaptibility to response.
Currently, research and development falling over of human body detecting system in terms of technology there are many kinds of, most commonly image analysis and
Acceleration analysis.The former is the automatic checkout system of falling down based on video image analysis, and this technology accuracy is high, human action
It is high-visible, but multi-section video camera is needed to work at the same time, and the individual privacy of user is exposed, monitoring range is limited, by environment
Influence it is also very big.The latter is based primarily upon MEMS sensor.MEMS (MEMS) technology is obtained in recent years
Quickly development, is widely used in falling down detection, state-detection, motion detection etc..It is domestic at present some be based on MEMS skills
Though art falls down detection and can preferably realize and fall down detection, most calculation amount is larger, design complexity, expensive, it is difficult to obtain
It is extensive to promote.
Invention content
The purpose of the present invention is to provide the Raspberry Pi old men based on deep learning to fall down detecting system, and of the invention is beneficial
Effect is to fall down testing principle to be based on DBN deep neural networks, can effectively prevent simulation, and it is freely falling body to ground such as to work as equipment
When can't be mistaken for falling down；Raspberry Pi hardware device is cheap, can be received by most of families, opposite micro-electromechanical technology
The shortcomings of calculation amount is larger, design is complicated, expensive, our invention has more advantage, can preferably be promoted.
The technical solution adopted in the present invention is the Internet of Things operating system Android Things based on Google, acquires number
According to using the acceleration signal collector based on 3-axis acceleration sensor, in each collector comprising two be placed in parallel three
Axle acceleration sensor, in this way, picker can collect two 3-axis acceleration data simultaneously when doing each action；
In gatherer process, collector is individually positioned in the waistband, trouser pocket, coat pocket position of picker；Accelerate
The data that degree sensor generates are the data using the time as independent variable, and the primary data sample of difference action includes the data of X-axis,
The data of Y-axis and the data of Z axis；The noise for solving to mix in original sampled signal using wavelet de-noising；
The wherein data of X-axis：Between the variation of normal walking brief acceleration is 10 to 20,0~70 represents acceleration when falling down
Degree variation, 80~120 represent acceleration change when running, and 130~170 represent acceleration change when walking；Angular speed side
To variation：Hour angle velocity variations are fallen down in 0~80 representative, and 80~120 represent running hour angle velocity variations, and 120~160 representatives are walked
Hour angle velocity variations；It, will alert when acceleration, angular speed variation are more than a certain range；
Include redundancy in data, select PCA modes as dimensionality reduction mode, data are reduced to 100 dimensions from 150 dimensions
Within, then using DBN deep neural networks the data of behavior state are trained, the network model of trained generation is for examining
Survey instant behavior state.
Further, the noise method for solving to mix in original sampled signal using wavelet de-noising is to be adopted first to original
Sample signal carries out wavelet decomposition, then noise section is generally comprised in high frequency coefficient；Then to the high frequency coefficient of wavelet decomposition with
The forms such as threshold value carry out quantification treatment；It is again finally the purpose that can reach noise reduction to signal reconstruction.
Further, instant behavior state is detected to include the following steps：
Step 1, input data establish the data that length is 50 groups；
Step 2：The instant data acquired in data acquisition are sequentially placed into last position of array；
Step 3：Carry out Noise reducing of data processing；
Step 4：Judge whether to fall；
Step 5：It is no, continue since step 2；
Step 6：It is to send warning message.
Description of the drawings
Fig. 1 is the primary data sample of different actions；
Fig. 2 be fall down, run, walk acceleration change figure；
Fig. 3 be fall down, run, walk angular speed direction change figure.
Specific implementation mode
The present invention is described in detail With reference to embodiment.
Internet of Things operating system Android Things of the inventive algorithm based on Google.Gathered data first is based on three
The acceleration signal collector of axle acceleration sensor.Include two 3-axis acceleration being placed in parallel sensings in each collector
Device, in this way, picker can collect two 3-axis acceleration data simultaneously when doing each action.Sensor there are one preventing
Loss of data.
Include the 10 class action datas of 44 different acquisition persons in CUT-NAA databases used.In gatherer process, adopt
Storage is individually positioned in the waistband, trouser pocket, coat pocket position of picker.10 classes act and description is listed in the table below 1：
Table 1
The data that acceleration transducer generates are the data using the time as independent variable, and the primary data sample of difference action is such as
Shown in Fig. 1, include the data (the first row) of X-axis, the data (the second row) of Y-axis and the data (the third line) of Z axis.
To solve the noise problem that mixes in original sampled signal our technologies for the use of wavelet de-noising.
Wavelet decomposition is carried out to original sampled signal first, then noise section is generally comprised in high frequency coefficient；Then to wavelet decomposition
High frequency coefficient quantification treatment is carried out in the form of threshold value etc.；It is again finally the purpose that can reach noise reduction to signal reconstruction.It is right
Signal de-noising be substantially inhibit signal in nonuseable part, restore signal in useful part process, i.e., to external environment into
Row processing, to reduce noise.The external environmental interference that Raspberry Pi is subject to can be reduced, can promptly and accurately be sounded the alarm
It rings.
Acceleration transducer data processing
Fig. 2 be it is measured fall down, run, the oscillogram for three kinds of states of walking.Horizontal axis 0~70 represents acceleration when falling down
Degree variation, horizontal axis 80~120 represent acceleration change when running, and horizontal axis 130~170 represents acceleration change when walking.
Normal walking brief acceleration variation range is smaller, remains between 10 to 20.In falling down generating process, acceleration
When having increased process after first reducing with size, and remaining static after falling down, acceleration can also be sent out compared with original state
Raw significant change.
Measured angular speed direction change figure is as shown in Figure 3：Hour angle velocity variations, horizontal axis are fallen down in the representative of horizontal axis 0~80
80~120 represent running hour angle velocity variations, and horizontal axis 120~160 represents hour angle velocity variations of walking.The above analysis finds, nothing
Which fallen down by from direction, the physical condition of people can all undergo the states such as weightless, overweight, static and acceleration change.Cause
These factors are carried out rational comprehensive analysis, you can as the condition for judging whether human body is fallen down by this.
Algorithm detecting state is carried out after the completion of system initialization.State as above and acceleration are detected within a certain period of time
It, will alert when degree, angular speed variation are more than a certain range.
Data are trained
In gesture recognition task, input each time has 3*50*1 row, and (3-axis acceleration sensor ADXL345 is with 50Hz
1 second data of frequency collection), export as 5 class categories.
For exercise data, the feature possessed by information in time domain contained by each row is not obvious.So we
Fast Fourier Transform (FFT) is carried out to the time series data, and retains the real part of preceding 50 row of the data after transformation.Because
Include redundancy in data, it, can be under the premise of accuracy rate influences less in classification speed after the dimensionality reduction of data
On have and more preferably show, to reduce the load of headend equipment CPU.We are to compared PCA, LDA and AutoEncoder etc. a variety of
After influence of the dimensionality reduction mode to our model accuracy rate, final choice PCA as our dimensionality reduction mode, by data from
150 dimensions are reduced within 100 dimensions, then carry out corresponding data training using deep neural network.Using DBN deep neural networks
The data of behavior state are trained, the network model of trained generation is for detecting instant behavior state.Detection is instant
Behavior state include the following steps：
Step 1, input data establish the data that length is 50 groups；
Step 2：The instant data acquired in data acquisition are sequentially placed into last position of array；
Step 3：Carry out Noise reducing of data processing；
Step 4：Judge whether to fall；
Step 5：It is no, continue since step 2；
Step 6：It is to send warning message.
Present system can be monitored the motion state of old man.Detect fall down when, can be in back-stage management
Website display alarm information, while warning reminding user.This system is equipped with acceleration transducer ADXL345, can identify user
Current motion state.If it find that the abnormality fallen down currently occurs in user, alarm signal will be just sent out.It can realize
Good user of service's safety monitoring and processing function.
The hardware that the present invention uses：Raspberry Pi 3B mono-, one, 16GBTF cards, ADXL acceleration transducers one,
TDA2030 power amplifier modules one, toy trumpet one, earphone one are secondary.
1. the ADXL acceleration transducers of collection exercise data are connected to by Du Pont's line on the pin of Raspberry Pi, connection
Mode is：The 5V in 5V connection Raspberry Pis, the GND of GND connection Raspberry Pis on ADXL, the SCL.1, SDA of SCL connection Raspberry Pis
Connect another SDA.1 pin of Raspberry Pi.
2. wired power amplifier module.Earphone is cut first, 3.5mm is connected to the audio port of Raspberry Pi.TDA2030 power amplifier moulds
The OUT of block connects the anode of loudspeaker, and GND connects the cathode of loudspeaker, and VCC meets the 5V of Raspberry Pi, and GND meets the GND of Raspberry Pi, and IN connects raspberry
The anode for the earphone cable of audio interface connect, GND is sent to connect the cathode of earphone cable.
3. after connecting all the things, Raspberry Pi even being connected to the Net, makes he and computer under the same network.Next
On computers open Android Studio Integrated Development Environment, the code finished writing before is burnt to Raspberry Pi, so that it may with into
Row, which is fallen down, to be tested.
The present invention devises the acceleration signal collector based on 3-axis acceleration sensor.Include two in each collector
A 3-axis acceleration sensor being placed in parallel can collect two 3-axis acceleration numbers simultaneously when picker does each action
According to.The purpose for the arrangement is that preventing the loss of data there are one sensor.Quasi- research carries out noise reduction process to original signal, reduces
Interference of the external environment to Raspberry Pi, can promptly and accurately send out alarm.
Study falling over of human body testing principle：Relevant acceleration transducer is connected using the development board based on Raspberry Pi to carry out
Behavioral data acquisition and noise reduction.Research is learnt, handled and is sorted out to human body behavioral data, and excavates its interior change rule
Rule, classifies to data according to the changing rule of different motion behavior.Study to changing rule by coding establish mould
Type is integrated into client,, can be clear by model analysis when receiving the behavioral data of acceleration collection again next time
The state recognized at this time be walking, fall down or run, realized with this and fall down detection.
The above is only the better embodiment to the present invention, not makees limit in any form to the present invention
System, every any simple modification that embodiment of above is made according to the technical essence of the invention, equivalent variations and modification,
Belong in the range of technical solution of the present invention.
Claims (3)
1. the Raspberry Pi old man based on deep learning falls down detecting system, it is characterised in that：Internet of Things net operation system based on Google
Unite Android Things, and gathered data uses the acceleration signal collector based on 3-axis acceleration sensor, each to acquire
Include two 3-axis acceleration sensors being placed in parallel in device, in this way, picker can collect simultaneously when doing each action
Two 3-axis acceleration data；
In gatherer process, collector is individually positioned in the waistband, trouser pocket, coat pocket position of picker；Acceleration passes
The data that sensor generates are the data using the time as independent variable, and the primary data sample of difference action includes the data of X-axis, Y-axis
Data and Z axis data；The noise for solving to mix in original sampled signal using wavelet de-noising；
The wherein data of X-axis：Between the variation of normal walking brief acceleration is 10 to 20,0~70 acceleration for representing when falling down becomes
Change, 80~120 represent acceleration change when running, and 130~170 represent acceleration change when walking；Angular speed direction becomes
Change：Hour angle velocity variations are fallen down in 0~80 representative, and 80~120 represent running hour angle velocity variations, and 120~160 represent hour angle of walking
Velocity variations；It, will alert when acceleration, angular speed variation are more than a certain range；
Include redundancy in data, select PCA modes as dimensionality reduction mode, data are tieed up from 150 be reduced to 100 dimensions with
It is interior, then the data of behavior state are trained using DBN deep neural networks, the network model of trained generation is for detecting
Instant behavior state.
2. falling down detecting system according to the Raspberry Pi old man based on deep learning described in claim 1, it is characterised in that：It is described to make
The noise method for solving to mix in original sampled signal with wavelet de-noising is to carry out small wavelength-division to original sampled signal first
Solution, then noise section is generally comprised in high frequency coefficient；Then to the high frequency coefficient of wavelet decomposition in the form of threshold value etc. into
Row quantification treatment；It is again finally the purpose that can reach noise reduction to signal reconstruction.
3. falling down detecting system according to the Raspberry Pi old man based on deep learning described in claim 1, it is characterised in that：The inspection
Instant behavior state is surveyed to include the following steps：
Step 1, input data establish the data that length is 50 groups；
Step 2：The instant data acquired in data acquisition are sequentially placed into last position of array；
Step 3：Carry out Noise reducing of data processing；
Step 4：Judge whether to fall；
Step 5：It is no, continue since step 2；
Step 6：It is to send warning message.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810598145.4A CN108648417A (en) | 2018-06-12 | 2018-06-12 | Raspberry Pi old man based on deep learning falls down detecting system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810598145.4A CN108648417A (en) | 2018-06-12 | 2018-06-12 | Raspberry Pi old man based on deep learning falls down detecting system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108648417A true CN108648417A (en) | 2018-10-12 |
Family
ID=63752362
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810598145.4A Pending CN108648417A (en) | 2018-06-12 | 2018-06-12 | Raspberry Pi old man based on deep learning falls down detecting system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108648417A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2020106212A1 (en) * | 2018-11-22 | 2020-05-28 | Optimax Management Services Pte. Ltd. | Intelligent impact sensor and uses |
| CN111912433A (en) * | 2020-07-14 | 2020-11-10 | 威步智能科技(苏州)有限公司 | High-precision falling real-time judgment system |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090161915A1 (en) * | 2007-12-24 | 2009-06-25 | National Chiao Tung University Of Taiwan | Multi-person pose recognition system using a zigbee wireless sensor network |
| CN102542724A (en) * | 2011-12-15 | 2012-07-04 | 上海移为通信技术有限公司 | Falling detecting and alarming method |
| CN106203512A (en) * | 2016-07-12 | 2016-12-07 | 北京安易康科技有限公司 | The detection method of falling down based on multi-sensor information fusion |
| CN106473265A (en) * | 2016-12-03 | 2017-03-08 | 石家庄学院 | A kind of intelligence is fallen and is guarded T-shirt and its fall detection algorithm |
| CN106971059A (en) * | 2017-03-01 | 2017-07-21 | 福州云开智能科技有限公司 | A kind of wearable device based on the adaptive health monitoring of neutral net |
-
2018
- 2018-06-12 CN CN201810598145.4A patent/CN108648417A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090161915A1 (en) * | 2007-12-24 | 2009-06-25 | National Chiao Tung University Of Taiwan | Multi-person pose recognition system using a zigbee wireless sensor network |
| CN102542724A (en) * | 2011-12-15 | 2012-07-04 | 上海移为通信技术有限公司 | Falling detecting and alarming method |
| CN106203512A (en) * | 2016-07-12 | 2016-12-07 | 北京安易康科技有限公司 | The detection method of falling down based on multi-sensor information fusion |
| CN106473265A (en) * | 2016-12-03 | 2017-03-08 | 石家庄学院 | A kind of intelligence is fallen and is guarded T-shirt and its fall detection algorithm |
| CN106971059A (en) * | 2017-03-01 | 2017-07-21 | 福州云开智能科技有限公司 | A kind of wearable device based on the adaptive health monitoring of neutral net |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2020106212A1 (en) * | 2018-11-22 | 2020-05-28 | Optimax Management Services Pte. Ltd. | Intelligent impact sensor and uses |
| CN111912433A (en) * | 2020-07-14 | 2020-11-10 | 威步智能科技(苏州)有限公司 | High-precision falling real-time judgment system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Saleh et al. | FallAllD: An open dataset of human falls and activities of daily living for classical and deep learning applications | |
| Gao et al. | Adaptive weighted imbalance learning with application to abnormal activity recognition | |
| Lim et al. | Fall-detection algorithm using 3-axis acceleration: combination with simple threshold and hidden Markov model | |
| CN108109336B (en) | Human body falling identification method based on acceleration sensor | |
| CN104269025B (en) | Wearable single node feature and the position choosing method of monitoring is fallen down towards open air | |
| CN109886068B (en) | Motion data-based action behavior identification method | |
| CN106580282A (en) | Human body health monitoring device, system and method | |
| CN103955699A (en) | Method for detecting tumble event in real time based on surveillance videos | |
| Cao et al. | A fall detection method based on acceleration data and hidden Markov model | |
| CN113963192A (en) | Fall detection method and device and electronic equipment | |
| CN105139869B (en) | A kind of baby crying detection method based on section Differential Characteristics | |
| CN108648417A (en) | Raspberry Pi old man based on deep learning falls down detecting system | |
| Kambhampati et al. | Unified framework for triaxial accelerometer‐based fall event detection and classification using cumulants and hierarchical decision tree classifier | |
| Nurwulan et al. | Window selection impact in human activity recognition | |
| Cheng et al. | A Fall detection system based on SensorTag and Windows 10 IoT core | |
| Kao et al. | GA-SVM applied to the fall detection system | |
| Qu et al. | Convolutional neural network for human behavior recognition based on smart bracelet | |
| CN112699744A (en) | Fall posture classification identification method and device and wearable device | |
| CN113367686A (en) | Method, device, computer equipment and storage medium for detecting human body falling | |
| Xu et al. | Research of HMM-based fall detection system for elderly | |
| Pipanmaekaporn et al. | Mining Acceleration Data for Smartphone-based Fall Detection | |
| Castro et al. | Fall Detection with LSTM and Attention Mechanism | |
| CN115886794A (en) | Multi-classifier selection method, frozen gait detection system, device and storage medium | |
| Abd Aziz et al. | Wearable Device-based Fall Detection System for Elderly Care Using Support Vector Machine (SVM) classifier | |
| Xie et al. | A multistage collaborative filtering method for fall detection |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181012 |