CN108230618A - A kind of low-power consumption fall detection system of Community-oriented based on ZigBee - Google Patents
A kind of low-power consumption fall detection system of Community-oriented based on ZigBee Download PDFInfo
- Publication number
- CN108230618A CN108230618A CN201711397818.1A CN201711397818A CN108230618A CN 108230618 A CN108230618 A CN 108230618A CN 201711397818 A CN201711397818 A CN 201711397818A CN 108230618 A CN108230618 A CN 108230618A
- Authority
- CN
- China
- Prior art keywords
- data
- fifo
- zigbee
- value
- moment
- 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
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/103—Detecting, measuring or recording devices for testing the shape, pattern, colour, size or movement of the body or parts thereof, for diagnostic purposes
- A61B5/11—Measuring movement of the entire body or parts thereof, e.g. head or hand tremor, mobility of a limb
- A61B5/1116—Determining posture transitions
- A61B5/1117—Fall detection
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/74—Details of notification to user or communication with user or patient ; user input means
- A61B5/746—Alarms related to a physiological condition, e.g. details of setting alarm thresholds or avoiding false alarms
-
- G—PHYSICS
- G08—SIGNALLING
- G08B—SIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
- G08B21/00—Alarms responsive to a single specified undesired or abnormal condition and not otherwise provided for
- G08B21/02—Alarms for ensuring the safety of persons
- G08B21/04—Alarms for ensuring the safety of persons responsive to non-activity, e.g. of elderly persons
- G08B21/0438—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
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W84/00—Network topologies
- H04W84/18—Self-organising networks, e.g. ad-hoc networks or sensor networks
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D30/00—Reducing energy consumption in communication networks
- Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks
Landscapes
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Engineering & Computer Science (AREA)
- Animal Behavior & Ethology (AREA)
- Public Health (AREA)
- Biophysics (AREA)
- Pathology (AREA)
- Biomedical Technology (AREA)
- Heart & Thoracic Surgery (AREA)
- Medical Informatics (AREA)
- Molecular Biology (AREA)
- Surgery (AREA)
- General Physics & Mathematics (AREA)
- Emergency Management (AREA)
- Physiology (AREA)
- Veterinary Medicine (AREA)
- Business, Economics & Management (AREA)
- Gerontology & Geriatric Medicine (AREA)
- Social Psychology (AREA)
- Psychology (AREA)
- Psychiatry (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Dentistry (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Navigation (AREA)
Abstract
The present invention discloses a kind of low-power consumption fall detection system of Community-oriented based on ZigBee, including:Multiple motion perception modules and server;Wherein, motion perception module realizes the acquisition of low-power consumption human body movement data using Zigbee and interrupt function, and the exercise data of acquisition is reached to the server for being integrated with ZigBee aggregation nodes by ZigBee-network, server carries out Kalman filtering and passes through kNN Algorithm Analysis and detect whether to fall, once detect tumble system automatic alarm.The present invention is on transmission range, the open scope of activities of user well;It in acquisition and transmits above user's tumble data, employ the equipment of low-power consumption and interrupts the data transfer algorithms of driving, solve the problems, such as wearable device high energy consumption very well;In tumble data processing and detection, the precision that detection identification is fallen is improved;Alarm falling above, timely can alarmed and be notified.
Description
Technical field
The invention belongs to electronic information fields, are related to a kind of low-power consumption fall detection system of Community-oriented based on ZigBee
System.
Background technology
Traditional wearable tumble technology based on 3 axis accelerometers, gyroscope has ignored the calculating energy of wearable device
The objective constraints of power and limited power.In wearable device 3 axis accelerometer of Main Processor Unit on-line real-time measuremen and
3 axle accelerations of reading and angular velocity data are sent to the computing device of receiving terminal by gyroscope by wireless radio frequency modules, then
Fall detection is carried out by it.Since the Main Processor Unit in wearable device needs Real-time Monitoring Data, radio frequency unit to need reality
When transmission data, the power consumption for leading to system is big.
In addition, traditional fall detection is all based on greatly Bluetooth technology, and Bluetooth transmission distance transmits model usually at 10 meters or so
It is with limit;In addition, bluetooth communication is easily blocked interference, and the point-to-point communication mode of generally use, may be only available for a
People or domestic consumer, it is difficult to meet the needs that old community multi-user detects in real time.
Invention content
In view of the above-mentioned problems, the present invention provides a kind of low-power consumption fall detection system of the Community-oriented based on Zigbee.Needle
Low-power consumption and accuracy requirement to community fall detection devise the motion perception of integrated MPU6050 and ZigBee
Module, and design and realize the low-power consumption 3-axis acceleration for interrupting driving, angular velocity data acquisition and transmission algorithm, and by human body
Exercise data is transferred to server by ZigBee-network;Secondly, on the server design realize based on Kalman filtering and
The fall detection alarm module of k-NN algorithms, once detect that tumble can be by making a phone call or the modes such as short message are alarmed.
To achieve the above object, the present invention uses following technical method:
A kind of low-power consumption fall detection system of Community-oriented based on ZigBee, including:Multiple motion perception module kimonos
Business device;Wherein, motion perception module realizes the acquisition of low-power consumption human body movement data using Zigbee and interrupt function, and will acquisition
Exercise data reach the server for being integrated with ZigBee aggregation nodes by ZigBee-network, server carries out Kalman filtering
And pass through kNN Algorithm Analysis and detect whether to fall, once detect tumble system automatic alarm.
The beneficial effects of the invention are as follows:On transmission range, the open scope of activities of user well;Obtain and
It transmits above user's tumble data, employ the equipment of low-power consumption and interrupts the data transfer algorithms of driving, solve wear very well
The problem of wearing equipment high energy consumption;In tumble data processing and detection, the precision that detection identification is fallen is improved;It alarms falling
Above, timely it can be alarmed and be notified.
Description of the drawings
Fig. 1:Human body movement data acquires transfer process figure;
Fig. 2:Lay figure based on cartesian coordinate system;
Fig. 3 is the structure diagram of low-power consumption fall detection system of the present invention.
Specific embodiment
As shown in figure 3, the present invention provides a kind of low-power consumption fall detection system of the Community-oriented based on ZigBee, including:
Multiple motion perception modules and server, motion perception module realize low-power consumption human motion number using Zigbee and interrupt function
According to acquisition, and the exercise data of acquisition by ZigBee-network is reached to the server for being integrated with ZigBee aggregation nodes, serviced
Device carries out Kalman filtering and passes through kNN Algorithm Analysis and detect whether to fall, once detect tumble system automatic alarm.
Detailed process is:
Step 1. motion perception module is obtained as Zigbee terminal nodes, i.e. data acquisition node by system break
Human body movement data, the data are the data of 2 seconds and latter 2 seconds before generation interruption;This ensure that the data acquisition of low-power consumption;
Step 2. motion perception module obtain data after, by Zigbee network forward the data to server set into
Zigbee aggregation nodes.Zigbee network has 4 layers of structure.The 4th layer of Zigbee network is by many Zigbee terminal sections
Point (motion perception module) forms, these underlying devices are used for obtaining user's tumble data;The 3rd layer (Zigbee routing layer),
The 4th layer network data are received, if the 2nd layer of Zigbee coordinator nodes can pass data to distance not in its transmission range
The nearer Zigbee routing nodes of Zigbee coordinator nodes.If the 2nd layer of Zigbee coordinator nodes, will in its transmission range
Data are sent to the 1st layer of Zigbee aggregation nodes;The most data handover service at last after the 1st layer of aggregation node receives data
Device is handled;
After step 3. server receives the data of Zigbee aggregation nodes, data analysis is carried out, is designed on the server
Realize the fall detection alarm system based on Kalman filtering and k-NN algorithms;
Step 4. once detects that tumble can be by making a phone call or the modes such as short message are alarmed.
Motion perception module acquires user movement data
The size of motion perception module be 28mm × 39mm × 9mm, by CC2530 microcontrollers, MPU6050 sensors and
ZigBee radio frequencies and power management module composition, are sized for being positioned over wearable vest close to the position of neck, Ke Yishi
When perceive user exercise data.Wherein, the transmission rate of ZigBee module is 115200baud, and transmission maximum distance is
100m.MPU-6050 is integrated with MEMS gyroscope, 3 axis accelerometers and expansible digital moving processor (Digital
Motion Processor, DMP).Gyroscope measurement range maximum can reach ± 2000 °/sec, and accelerometer measures range is most
Big reachable ± 16g;The sample frequency of motion perception module is 50Hz.
MPU-6050 includes the fifo register of a 1KB for data cached.Meanwhile the DMP that MPU-6050 is integrated can
To read data from gyroscope, accelerometer in dormant state and deposit in FIFO cachings, since micro-control need not be accessed at this time
Device and ZigBee radio frequency processed, therefore module is in low-power consumption mode.In addition, there are one programmable interruption systems by MPU6050, carry
Freely falling body (Free Fall), static (Zero Motion), FIFO has been supplied to overflow (FIFO Overflow) to interrupt.
Freely falling body interrupts, and MPU6050 is by detecting whether the acceleration measurement on 3 axis judges in defined threshold
The movement of falling object.To sampled value each time, if not reaching threshold value will be ignored;Once reaching threshold value, just trigger certainly
It is interrupted by falling bodies, and generates flag bit.The FF_THR registers of MPU6050 are used to set the threshold value of freely falling body, and value is accurate
To 1mg;FF_DUR registers are accurate to 1ms for setting the freely falling body duration.
Static interruption:For ZRMOT_THR registers for setting static threshold value, value is accurate to 1mg.ZRMOT_DUR is posted
For storage to set static duration of interruption, value is accurate to 1ms.It is different from freely falling body interruption, it is detected when for the first time
Static and when no longer detecting, static interruption can all be triggered.
FIFO, which overflows, to interrupt:The data capacity value of FIFO cachings can be set by the FIFO_CNT registers of MPU6050,
FIFO, which is generated, when number data cached in FIFO is more than the value overflows interruption.
By the way that freely falling body, static and FIFO is set to overflow the reasonable threshold value interrupted, motion perception module is detecting certainly
During by falling bodies or active state, the activity data for using and sending user is acquired in real time by CC3250;When detecting static interruption,
CC3250 and MPU6050 is responsible for acquiring 3-axis acceleration and angular velocity data and depositing to FIFO delaying all in dormant state by DMP
In depositing, MPU and ZigBee radio frequencies need not be accessed at this time and then reduce the power consumption of motion perception module.
When human body is fallen, centre of body weight can rapidly drop to lower from eminence, and acceleration transducer can be undergone short at this time
Temporary zero-g period, the acceleration value exported at this time can be fast approached in 0, and this state can be interrupted by freely falling body and be captured
It arrives.Since the time that human body is fallen is usually less than 2 seconds, setting FF_THR is 0.5625g, FF_DUR 40ms.Work as system detectio
It is interrupted to freely falling body, and the duration is more than 40ms, then system can be by ZigBee immediately by the data cached hair in FIFO
Give server.
When acceleration value is less than 0.5g and time delay was more than 1 second, system generates static interruption, is set in static interrupt processing
Module is in suspend mode.MPU6050 with low-power consumption mode acquisition acceleration, gyro data and is stored in FIFO cachings at this time
In.
Since human body fell the time less than 2 seconds, add in data sampling frequency the position 50Hz, MPU6050 of motion detection block
Several times, magnitude of angular velocity uses 2 byte representations, and therefore, present invention setting FIFO_CNT is 600 bytes, i.e., when FIFO is data cached
Byte number be more than 600 when, MPU6050 can generate FIFO overflow interrupt.
It is set based on above-mentioned threshold value, the present invention devises the human body movement data perception transmission algorithm for interrupting driving.Fig. 1 is
Human body movement data acquisition transfer process figure corresponding with algorithm, using finite-state machine description.Each state description is such as in figure
Under:
F0:Init state.Initialization FIFO is cached and is set data sampling rate as 50Hz, the interruption of setting freely falling body,
Static interruption and FIFO, which overflow, to interrupt as enabled state;System enters low-power consumption mode, i.e., into F1.
F1:Stationary state.3 axle accelerations, angular velocity data are acquired by DMP and cached into FIFO;If in static
Update FIFO was cached and was removed static interrupt flag bit, continues low-power consumption mode more than 1 second the disconnected duration.In stationary state,
It is interrupted if FIFO occurs and overflows, DMP is according to the data in the principle update FIFO of first in first out;When the acceleration value of reading is big
It when 0.5g, is interrupted if interrupting to overflow with FIFO without generation freely falling body at this time, DMP continues reading data and caches extremely
FIFO;Otherwise the data during FIFO is cached by ZigBee are sent to server.
F2:Active state.If generating freely falling body to interrupt, and duration of interruption is more than 40ms, then module passes through
ZigBee is data cached to server transmission FIFO.If freely falling body duration of interruption is less than 40ms and produces FIFO spillings
It interrupts, then module is data cached to server transmission FIFO by ZigBee and returns to F1;Otherwise, addition number is cached to FIFO
According to, and return to F1.
The lay figure based on three-dimension altitude angle is established, and analyzes and compares daily routines with falling in attitude angle and adding
Difference on speed signal vector mould (Signal Vector Magnitude, SVM) refines classification daily routines and tumble
Characteristic parameter.
During the motion, acceleration and angular speed can real-time change for human body.Erdogan etc. is research shows that the upper body of human body
Dry (i.e. more than waist, neck with lower part) is to acquire acceleration information and identify tumble and the best portion of other everyday actions
Position.Motion perception module is placed on the neck of customization vest by the present invention from wearable device comfort and system reliability
Back, and the placement direction according to sensor establishes the lay figure based on cartesian coordinate system, as shown in Figure 2.
In Fig. 2, coordinate system of the left figure for human body acceleration, ax、ay、azIt represents respectively in physical activity along x-axis, y-axis and z
The acceleration of axis.Right figure be human body angular speed and three-dimension altitude angle coordinate system, ωx、ωy、ωzTo represent trunk respectively around x
The angular speed of axis, y-axis and z-axis.
In the right figure of Fig. 2, pitch represents the pitch angle of the rotation angle, i.e. human body forward/backward around y-axis;Roll represent around
The rotation angle of z-axis, i.e. the human body inclined roll angle in both sides to the left and right;Yaw represents the rotation angle around x-axis, i.e. human body towards left/right rotation
To deflection angle.According to the relationship of acceleration in three dimensions and gravity, 3 attitude angles are defined as follows:
In above-mentioned 3 formula, ax、ay、azThe acceleration measured along x-axis, y-axis, z-axis represented respectively.
In addition, the present invention uses SVMaAs a characteristic quantity for judging physical activity type.SVMaSize and human motion
Severe degree it is related, it is unrelated with the direction of motion, be defined as follows:
Select 3 Wei Rentizitaijiao and SVMaIt is characterized, and can effectively classify the daily of human body with reference to suitable algorithm
Activity and tumble.
Fall detection of the server end based on Kalman filtering and k-NN algorithms
(1) filter data is crossed using Kalman filtering
With reference to the model of figure 2, the attitude angle of human body can be calculated according to acceleration information.But accelerometer in a practical situation
Noises, these factors such as voltage fluctuation, gravitational acceleration component superposition can be generated under motion state and add the angle that can make to calculate
There are notable differences with real data for degree.In addition, the although product of human motion angular speed acquired in real time by computing gyroscope
Divide the corner that can be derived that human body.But after long-time integration operation, the tiny signal static drift of gyroscope inherently can be by
Step accumulation, causes the output error of system gradually to be promoted, and the angle calculated is caused deviation occur.Therefore accelerometer is used alone
Or the corner that gyro sensor calculates human body deviation can all occur, and then influence the accuracy of fall detection.It is in this regard, of the invention
Accelerometer and gyro data are merged using simple, efficient Kalman filtering.Accelerometer and top is respectively adopted
Measured value and predicted value of the attitude angle that spiral shell instrument meter calculates as system calculate accurate posture by two kinds of data complement fusions
Angle, and then improve the accuracy rate of fall detection.
If θ is human body attitude angle, ω and β represent angular speed and its static drift that gyroscope acquires respectively, and dt representatives are adopted
The size of sample timeslice, k and k-1 represent residing for system the moment respectively, and the people shown in formula 5 can be established based on gyro data
The linear model of body attitude angle.
θk=θk-1+ωk-1-βk-1dt (5)
If the static drift of gyroscope is constant, have:
βk=βk-1 (6)
To (5) (6) two formula simultaneous, state matrix equation can be obtained:
Using the attitude angle calculated according to acceleration information and the attitude angle that angular velocity data calculates is used as system
Measured value and predicted value, can Kalman filtering be realized by following 5 steps, and then the accurate attitude angle for calculating human body.
1) predicted value at k moment human body attitudes angle
X (k | k-l)=Ax (k-l | k-l)+BU (k) (8)
Wherein, X (k | k-1) is the predicted value at current time, i.e. the k moment is to the predicted value of attitude angle;X (k-1 | k-1) be
The optimal value of last moment, i.e. system are in the attitude angle filter result at k-1 moment;U (k) is control of the current k moment to system
Amount, the present invention take controlled quentity controlled variable of the integration for attitude angle predicted value of angular speed over the sampling time interval;A and B is systematic parameter.
According to formula (7), the value that can obtain A, B is respectively:
2) it calculates the covariance of the predicted value at k moment and measures its precision
P (k | k-1)=AP (k-1 | k-1) AT+Q (10)
Wherein, P (k | k-1) is the corresponding covariances of X (k | k-1);P (k-1 | k-1) is the corresponding association sides of X (k-1 | k-1)
Difference;ATTransposed matrix for systematic parameter A.Q is vectorCovariance matrix.Since gyroscopic drift noise is made an uproar with attitude angle
Sound is independent from each other, then the calculating of Q can abbreviation be:
3) the kalman gain COEFFICIENT K g (k) at the k moment is calculated
According to formula (7), θ and β is both needed to be estimated, so it is bivector to need to define Kg (k)Correspond to respectively θ with
The kalman gain coefficient of β.R is the covariance of systematic survey noise.H is system output matrix, is counted according to formula (1)-(3)
Jacobin matrix of the attitude angle value of calculating about X (k | k-1).Since attitude angle value and attitude angle predicted value θ are straight
Correlation is connect, and it is unrelated with β, so H=10.HTIt is the transposed matrix of H.
4) the attitude angle filter value at k moment is calculated
X (k | k)=X (k | k-1)+Kg (k) [Z (k)-HX (k | k-1)] (12)
Wherein, optimization estimated values of the X (k | k) for the k moment in sliding window, i.e., system is in filter of the k moment to attitude angle
Wave is as a result, Z (k) is the attitude angle value calculated respectively according to formula (1) (2) (3).
5) covariance of the attitude angle filter value at k moment is calculated
P (k | k)=(I-Kg (k) H) P (k | k-1) (13)
Wherein, I is unit matrix.When system enters the k+1 moment, P (k | k) is exactly the P (k-1 | k-1) in formula (10),
(11) (12) (13) are the state renewal equation of Kalman filter.It performs renewal equation repeatedly in this way, noise can be reduced to passing
The influence of sensor data, so as to improve the accuracy of fall detection.
Since system sampling frequency is 50Hz, so setting the sampling interval as dt=0.02s.Therefore Kalman filters in the present invention
Wave initial parameter is as follows:
Kalman filtering can eliminate a large amount of shakes in posture angular curve, this feature for being conducive to human body movement data carries
Take the design with sorting algorithm.
(2) the fall detection algorithm based on sliding window and k-NN
Since the continuity and infinite property, the human body movement data that sensor acquires in real time of physical activity constitute continuously
Flow data, the sorting algorithm of conventional process static data cannot directly apply such data.In this regard, the present invention is suitable by choosing
Sliding window, by after Kalman filtering 3 dimension attitude angles and SVMa data be stored in the daily work of structural classification in sliding window
Characteristic that is dynamic and falling.
Sliding window is a section in data flow, wherein arrival time there are one each data, and according to advanced
First go out the in store flow data received recently of principle.It is shown according to a large amount of statistics, the tumble process of human body and other daily routines
It is usually happened in 2s, therefore the time span of sliding window is set as 2s by the present invention.Due to the data of motion perception module
Sample frequency is 50Hz, therefore is calculated in each sliding window comprising 100 groups according to Kalman filter algorithm and formula (4)
Human body three-dimensional attitude angle and SVMa, they constitute a sample space.
The present invention realizes fall detection using k-NN algorithms, and estimates the similitude between sample using manhatton distance
Degree.
K-NN algorithms are defined as follows:For given test sample x, calculate and select the k sample closest with it
y1..., yk, and vote according to formula (14) it, select generic of most neighbor nodes as x of voting.
Wherein, c (yi) it is sample yiAffiliated classification, δ vote by proxy functions:As u=v, δ (u, v)=1.
In sliding window each characteristic value by three-dimension altitude angle (be abbreviated as θ,ω) formed with SVMa.Each sample record
100 data in quick variation 2s occur for SVMa, and therefore, the manhatton distance calculation formula between sample is as follows:
Wherein, D (x, t) is manhatton distance, and x is test sample, and t is training sample, and i is the data of each community-internal
Number.It is of the invention when seeking manhatton distance due to accelerometer and the range of gyroscope instrument registration evidence and precision difference
Normalized has been carried out to 3 dimension attitude angles in window and SVMa.
Claims (4)
1. a kind of low-power consumption fall detection system of Community-oriented based on ZigBee, which is characterized in that including:Multiple motion perceptions
Module and server;Wherein,
Motion perception module realizes the acquisition of low-power consumption human body movement data using Zigbee and interrupt function, and by the movement of acquisition
Data reach the server for being integrated with ZigBee aggregation nodes by ZigBee-network, and server carries out Kalman filtering and passes through
KNN Algorithm Analysis detects whether to fall, once detect tumble system automatic alarm.
2. low-power consumption fall detection system of the Community-oriented as described in claim 1 based on ZigBee, which is characterized in that movement
Sensing module is made of CC2530 microcontrollers, MPU6050 sensors and ZigBee radio frequencies and power management module, MPU-6050
Comprising:MEMS gyroscope, 3 axis accelerometers, fifo register and expansible digital moving processor DMP, DMP are in suspend mode shape
State reads data from gyroscope, accelerometer and deposits in FIFO cachings,
MPU6050 is also included:Programmable interruption system, provide freely falling body (Free Fall), static (Zero Motion),
FIFO, which overflows (FIFO Overflow), to interrupt;Wherein,
Freely falling body interrupts, and MPU6050 is by detecting whether the acceleration measurement on 3 axis in defined threshold judges freedom
Falling;To sampled value each time, if reaching threshold value, just trigger freely falling body and interrupt, and generate flag bit;
Static interruption:By ZRMOT_THR registers to set static duration of interruption, when detect for the first time it is static with
And when no longer detecting, static interruption can all be triggered;
FIFO, which overflows, to interrupt:The data capacity value of FIFO cachings is set by FIFO_CNT registers, when data cached in FIFO
Number generates FIFO when being more than the value and overflows interruption.
3. low-power consumption fall detection system of the Community-oriented as claimed in claim 2 based on ZigBee, which is characterized in that
MPU6050 sensors are as follows by the human body movement data perception for interrupting driving:
F0:Init state, initialization FIFO are cached and are set data sampling rate as 50Hz, and setting freely falling body interrupts, is static
It interrupts and FIFO is overflowed and interrupted as enabled state;System enters low-power consumption mode, i.e., into F1;
F1:Stationary state acquires 3 axle accelerations, angular velocity data by DMP and caches into FIFO;If static interruption is held
Update FIFO was cached and was removed static interrupt flag bit, continues low-power consumption mode more than 1 second the continuous time.In stationary state, if sending out
Raw FIFO, which overflows, to interrupt, then DMP is according to the data in the principle update FIFO of first in first out;When the acceleration value of reading is more than
It during 0.5g, is interrupted if interrupting to overflow with FIFO without generation freely falling body at this time, DMP continues reading data and caches extremely
FIFO;Otherwise the data during FIFO is cached by ZigBee are sent to server;
F2:Active state is interrupted if generating freely falling body, and duration of interruption is more than 40ms, then module passes through ZigBee
It is data cached that FIFO is sent to server.If freely falling body duration of interruption, which is less than 40ms and produces FIFO, overflows interruption,
Then module is data cached to server transmission FIFO by ZigBee and returns to F1;Otherwise, interpolation data is cached to FIFO, and returned
Return F1.
4. low-power consumption fall detection system of the Community-oriented as claimed in claim 2 based on ZigBee, which is characterized in that service
Device end group includes in the process that Kalman filtering and the fall detection of k-NN algorithms are alarmed:
Step (1) crosses filter data using Kalman filtering
If θ is human body attitude angle, ω and β represent angular speed and its static drift that gyroscope acquires respectively, when dt represents sampling
Between piece size, k and k-1 are represented residing for system the moment, the human body appearance shown in formula 5 can be established based on gyro data respectively
The linear model at state angle,
θk=θk-1+ωk-1-βk-1dt (5)
If the static drift of gyroscope is constant, have:
βk=βk-1 (6)
To (5) (6) two formula simultaneous, state matrix equation can be obtained:
It will be according to the attitude angle that acceleration information calculates and the survey using the attitude angle that angular velocity data calculates as system
Magnitude and predicted value can realize Kalman filtering by following 5 steps, calculate the attitude angle of human body,
1) predicted value at k moment human body attitudes angle
X (k | k-1)=AX (k-1 | k-1)+BU (k) (8)
Wherein, X (k | k-1) is the predicted value at current time, i.e. the k moment is to the predicted value of attitude angle;X (k-1 | k-1) it is upper one
The optimal value at moment, i.e. system are in the attitude angle filter result at k-1 moment;U (k) for the current k moment to the controlled quentity controlled variable of system, take
Controlled quentity controlled variable of the integration for attitude angle predicted value of angular speed over the sampling time interval;A and B is systematic parameter,
According to formula (7), the value that can obtain A, B is respectively:
2) it calculates the covariance of the predicted value at k moment and measures its precision
P (k | k-1)=AP (k-1 | k-1) AT+Q (10)
Wherein, P (k | k-1) is the corresponding covariances of X (k | k-1);P (k-1 | k-1) is the corresponding covariances of X (k-1 | k-1);AT
For the transposed matrix of systematic parameter A, Q is vectorCovariance matrix, since gyroscopic drift noise and posture angle noise are
It is mutually independent, then the calculating of Q can abbreviation be:
3) the kalman gain COEFFICIENT K g (k) at the k moment is calculated
According to formula (7), θ and β is both needed to be estimated, so it is bivector to need to define Kg (k)Correspond to θ's and β respectively
Kalman gain coefficient, R be systematic survey noise covariance, H be system output matrix, be according to attitude angle value about
The Jacobin matrix of X (k | k-1), HTIt is the transposed matrix of H.
4) the attitude angle filter value at k moment is calculated
X (k | k)=X (k | k-1)+Kg (k) [Z (k)-HX (k | k-1)] (12)
Wherein, optimization estimated values of the X (k | k) for the k moment in sliding window, i.e. system are at the k moment to the filtering knot of attitude angle
Fruit, Z (k) are the attitude angle value calculated respectively according to formula (1) (2) (3),
5) covariance of the attitude angle filter value at k moment is calculated
P (k | k)=(I-Kg (k) H) P (k | k-1) (13)
Wherein, I is unit matrix, and when system enters the k+1 moment, P (k | k) is exactly the P (k-1 | k-1) in formula (10);
Fall detection of the step (2) based on sliding window and k-NN
By choosing suitable sliding window, 3 dimension attitude angles after Kalman filtering and SVMa data are stored in sliding window
Middle structural classification daily routines and the characteristic fallen.
The human body three-dimensional posture being calculated in each sliding window comprising 100 groups according to Kalman filter algorithm and formula (4)
Angle and SVMa, they constitute a sample space.
Fall detection is realized, and estimate the similarity measurements between sample using manhatton distance using k-NN algorithms.
K-NN algorithms are defined as follows:For given test sample x, calculate and select the k sample y closest with it1...,
yk, and vote according to formula (14) it, select generic of most neighbor nodes as x of voting.
Wherein, c (yi) it is sample yiAffiliated classification, δ vote by proxy functions:As u=v, δ (u, v)=1.
In sliding window each characteristic value by three-dimension altitude angle (be abbreviated as θ,It ω) is formed with SVMa, each sample record
100 data in quick variation 2s occur for SVMa, and therefore, the manhatton distance calculation formula between sample is as follows:
Wherein, D (x, t) is manhatton distance, and x is test sample, and t is training sample, and i is that the data of each community-internal are compiled
Number.Due to accelerometer and the range of gyroscope instrument registration evidence and precision difference, when seeking manhatton distance, the present invention is right
3 dimension attitude angles and SVMa in window have carried out normalized.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711397818.1A CN108230618A (en) | 2017-12-21 | 2017-12-21 | A kind of low-power consumption fall detection system of Community-oriented based on ZigBee |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711397818.1A CN108230618A (en) | 2017-12-21 | 2017-12-21 | A kind of low-power consumption fall detection system of Community-oriented based on ZigBee |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108230618A true CN108230618A (en) | 2018-06-29 |
Family
ID=62648392
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201711397818.1A Pending CN108230618A (en) | 2017-12-21 | 2017-12-21 | A kind of low-power consumption fall detection system of Community-oriented based on ZigBee |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108230618A (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110604546A (en) * | 2019-07-29 | 2019-12-24 | 桂林医学院附属医院 | Parkinson disease monitoring method and device and storage medium |
CN111189443A (en) * | 2020-01-14 | 2020-05-22 | 电子科技大学 | Pedestrian navigation method for online step length calibration, motion deviation angle correction and adaptive energy management |
CN111311877A (en) * | 2020-02-20 | 2020-06-19 | 吉林农业大学 | Fall detection and positioning method and device based on attitude angle |
CN111401507A (en) * | 2020-03-12 | 2020-07-10 | 大同公元三九八智慧养老服务有限公司 | Adaptive decision tree fall detection method and system |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104080053A (en) * | 2014-07-16 | 2014-10-01 | 上海诺诚电气有限公司 | Human falling-down detecting and positioning system and method based on Zigbee technology |
CN104821061A (en) * | 2015-03-26 | 2015-08-05 | 苏州大学 | Wearable fall early warning system and application method thereof |
CN106037749A (en) * | 2016-05-18 | 2016-10-26 | 武汉大学 | Old people falling monitoring method based on smart mobile phone and wearable device |
CN106056848A (en) * | 2016-06-30 | 2016-10-26 | 深圳先进技术研究院 | Low-power-consumption realization method for early warning device for fall prevention |
CN106960542A (en) * | 2017-05-21 | 2017-07-18 | 北京工业大学 | A kind of low-power consumption falling detection device based on interruption driving |
CN206726416U (en) * | 2017-03-08 | 2017-12-08 | 广东工业大学 | A kind of falling detection device and system |
-
2017
- 2017-12-21 CN CN201711397818.1A patent/CN108230618A/en active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104080053A (en) * | 2014-07-16 | 2014-10-01 | 上海诺诚电气有限公司 | Human falling-down detecting and positioning system and method based on Zigbee technology |
CN104821061A (en) * | 2015-03-26 | 2015-08-05 | 苏州大学 | Wearable fall early warning system and application method thereof |
CN106037749A (en) * | 2016-05-18 | 2016-10-26 | 武汉大学 | Old people falling monitoring method based on smart mobile phone and wearable device |
CN106056848A (en) * | 2016-06-30 | 2016-10-26 | 深圳先进技术研究院 | Low-power-consumption realization method for early warning device for fall prevention |
CN206726416U (en) * | 2017-03-08 | 2017-12-08 | 广东工业大学 | A kind of falling detection device and system |
CN106960542A (en) * | 2017-05-21 | 2017-07-18 | 北京工业大学 | A kind of low-power consumption falling detection device based on interruption driving |
Non-Patent Citations (2)
Title |
---|
何坚等: "基于卡尔曼滤波与k-NN算法的可穿戴跌倒检测技术研究", 《电子与信息学报》 * |
郑淑孟: "基于MEMS惯性传感器的人体姿态和位移检测", 《中国硕士学位论文全文数据库信息科技辑》 * |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110604546A (en) * | 2019-07-29 | 2019-12-24 | 桂林医学院附属医院 | Parkinson disease monitoring method and device and storage medium |
CN111189443A (en) * | 2020-01-14 | 2020-05-22 | 电子科技大学 | Pedestrian navigation method for online step length calibration, motion deviation angle correction and adaptive energy management |
CN111311877A (en) * | 2020-02-20 | 2020-06-19 | 吉林农业大学 | Fall detection and positioning method and device based on attitude angle |
CN111401507A (en) * | 2020-03-12 | 2020-07-10 | 大同公元三九八智慧养老服务有限公司 | Adaptive decision tree fall detection method and system |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108230618A (en) | A kind of low-power consumption fall detection system of Community-oriented based on ZigBee | |
CN106705968B (en) | Indoor inertial navigation algorithm based on attitude identification and step size model | |
CN106095099B (en) | A kind of user behavior motion detection recognition methods | |
Klein et al. | StepNet—Deep learning approaches for step length estimation | |
CN109579853B (en) | Inertial navigation indoor positioning method based on BP neural network | |
AU2015316575B2 (en) | Inertial tracking based determination of the position of a mobile device carried by a user in a geographical area | |
CN103027687A (en) | Old people fall detection and alarm system based on 3D (3-dimensional) accelerometer and gyroscope | |
CN103884337A (en) | Multi-attitude step correction positioning system and method | |
CN110113116A (en) | Human bodys' response method based on WIFI channel information | |
CN103340634A (en) | Method for detecting human motion state based on accelerated speed changes | |
CN107766930B (en) | Equivalent ROM distance calculation method based on DTW cluster fuzzy clustering SOM neurons | |
CN107432739A (en) | A kind of smart garment system for health monitoring | |
Wu et al. | Indoor positioning system based on inertial MEMS sensors: Design and realization | |
Florentino-Liaño et al. | Human activity recognition using inertial sensors with invariance to sensor orientation | |
WO2017185222A1 (en) | System and method for motion trajectory collection and analysis based on ball games | |
Minh et al. | Evaluation of smartphone and smartwatch accelerometer data in activity classification | |
CN111603750A (en) | Motion capture recognition evaluation system and method based on edge calculation | |
Wang et al. | Robust and accurate step counting based on motion mode recognition for pedestrian indoor positioning using a smartphone | |
Zeng et al. | A novel step counting algorithm based on acceleration and gravity sensors of a smart-phone | |
Li et al. | Collaborative recognition of queuing behavior on mobile phones | |
Tang et al. | Deep CNN-BiLSTM model for transportation mode detection using smartphone accelerometer and magnetometer | |
CN110455284A (en) | A kind of pedestrian movement patterns' recognition methods and device based on MEMS-IMU | |
Bai et al. | Application and research of MEMS sensor in gait recognition algorithm | |
CN110929766B (en) | Self-adaptive pedestrian mobile phone attitude identification method based on Gaussian mixture clustering algorithm | |
Wu et al. | Fall detection monitoring system based on mems sensor |
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 | ||
WD01 | Invention patent application deemed withdrawn after publication | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20180629 |