CN103815912B - Old solitary people based on thermal infrared sensor array falls down behavior method of real-time - Google Patents

Abstract

Old solitary people based on thermal infrared sensor array is fallen down behavior method of real-time and is belonged to areas of information technology, and the method comprises the steps: step one: adopt thermal infrared sensor data collecting system to carry out real-time data acquisition to the crawler behavior of indoor old solitary people; Step 2: mathematical modeling is carried out to the real time position of old solitary people indoor residing for it that step one monitors; Step 3: utilize the locus p falling down monitoring algorithm and step 2 is located _i,jt () judges, judge monitor that whether personage's state is as falling down.The invention has the beneficial effects as follows: the method, while carrying out real-time high-efficiency monitoring to the behavior of falling down, effectively protects individual privacy, and to daily life without bothering, the round-the-clock monitoring without blind spot can be carried out to target area.

Description

Old solitary people based on thermal infrared sensor array falls down behavior method of real-time

Technical field

The invention belongs to areas of information technology, propose a kind of old solitary people based on the observation of thermal infrared sensor array and fall down behavior method of real-time.The method is specially adapted to behavior of falling down Real-Time Monitoring that is solitary to empty nest and geriatric nursing home old man, and system can send warning immediately after old man falls down.

Background technology

At present, at home and abroad old people falls down monitoring field, and most of old people falls down monitoring equipment and have employed wearable portable sensor, such as acceleration and angular-rate sensor, or more simply helps call button.But most old people is unwilling body-worn sensor device miscellaneous often.After old people falls down, the situation that health can not move or brain loses consciousness happens occasionally, and in this case, helps call button also can lose efficacy.The method that another monitoring old people falls down adopts traditional camera head monitor, identifies fast the behavior of falling down.But the often invasion of privacy of traditional camera head monitor, especially among the home environment of old solitary people, is caught on camera the discomfort that the long-term supervision of head often brings psychology.

Thermal infrared sensor is also called infrared motion sensor, higher to personage's body temperature sensitivity, and the sensor array be made up of this sensor can gather the action/behavioural information of mobile personage, provides data basis to behavior analysis and Deviant Behavior monitoring etc.But the defect of thermal infrared sensor array is: the data precision of collection is low, be only 0,1 two-value data of reaction action severe degree.

Summary of the invention

To fall down deficiency in monitoring old people in order to overcome portable sensor and traditional camera device, the thermal infrared sensor technology that the inventive method adopts and old people fall down monitoring algorithm can by monitoring equipment and monitoring method to user's daily life bother and the infringement degree of privacy drops to minimum; Further, method of the present invention does not affect by light intensity, by day or night light change can not bring behavioral data error; Meanwhile, comprehensive each corner being installed on indoor roof of sensor array, can realize without blind monitoring.

The technical solution used in the present invention is as follows:

Old solitary people based on thermal infrared sensor array is fallen down behavior method of real-time and is comprised the steps:

Step one: adopt thermal infrared sensor data collecting system to carry out real-time data acquisition to the crawler behavior of indoor old solitary people;

Described thermal infrared sensor data collecting system comprises controller module, thermal infrared sensor array and PC industrial computer; Multiple thermal infrared sensor is arranged at indoor roof and is formed top view video camera array, for taking the crawler behavior real time data in monitor area; Controller module comprises MCU singlechip microprocessor, USB transducer and multiple RS-422 transceiver; Multiple thermal infrared sensor forms a thermal infrared sensor bus by serial or parallel connection, and every bar thermal infrared sensor bus is connected with MCU singlechip microprocessor by a RS-422 transceiver; Containing falling down monitoring algorithm in PC industrial computer, it is connected with MCU singlechip microprocessor by USB transducer, PC industrial computer sends to MCU singlechip microprocessor and starts or stops instruction, and receives and store the crawler behavior real time data transmitted by MCU singlechip microprocessor;

Step 2: mathematical modeling is carried out to the real time position of old solitary people indoor residing for it that step one monitors;

With expression formula s _ij(t), (i=1,, m, j=1, n) represent and be positioned at the sensor active state 0 of (i, j) position or 1(0 represents sensor an inactive state in t, 1 represents sensor active state), when sample frequency is set to H, definition t(t>H/2) pixel value of moment institute's position location pixel is:

$p_{i,j}\left(t\right)=\frac{1}{H+1}\underset{u=t-H/2}{\overset{t+H/2}{\mathrm{\Σ}}}{s}_{i,j}\left(u\right)---\left(1\right)$

P _i,j(t) ∈ [0,1]; Suppose that t has N(N≤m*n) individually enliven sensor, position is represented by (i, j), makes its pixel value be p _ij, utilize calculated with weighted average method to go out t to the position that old solitary people is located to be:

$(x_t,y_t)=\frac{{\mathrm{\Σ}}_{i=1}^m{\mathrm{\Σ}}_{j=1}^n(i,j)p_{i,j}\left(t\right)}{{\mathrm{\Σ}}_{i=1}^m{\mathrm{\Σ}}_{j=1}^n{p}_{i,j}\left(t\right)}---\left(2\right)$

The time series P of pixel value vector _t={ p ₍₁₎, p ₍₂₎..., p _(t), p _(t)=(p _1,1(t) ..., p _m,n(t)), the pixel value vector p of current time _(t)about pixel value vector time series P _tsingular value be defined as:

s _t＝||p _(t)-c|| （3），

C is cluster centre, and its expression formula is || || be Euclidean distance;

Step 3: utilize the locus p falling down monitoring algorithm and step 2 is located _i,jt () judges, judge monitor that whether personage's state is as falling down, it specifically comprises following sub-step:

Step 3.1: adopt with acc power halter strap tolerance pixel value vector p _(t)about pixel value vector time series P _thalter strap value, it is defined as

$M_n^{\left(\mathrm{\ϵ}\right)}=\underset{t=1}{\overset{n}{\mathrm{\Π}}}\left(\mathrm{\ϵ}{\hat{q}}_t^{\mathrm{\ϵ}-1}\right)---\left(4\right),$

ε ∈ [0,1], be worth by value function calculates:

${\hat{q}}_t\left(\right\{p\left(1\right),p\left(2\right),...,p\left(t\right)\},{\mathrm{\θ}}_t)=\frac{\#\{r:s_r>s_t\}+{\mathrm{\θ}}_t\#\{r:s_r=s_t\}}{t}---\left(5\right),$

In formula (5), #{r:s _r> s _trepresent and work as s _r> s _ttime r counting; s _rfor in the singular value of moment r measured by formula (3), r=1,2 ... t; θ _tevenly randomly drawed by interval [0,1] at moment t; Initial halter strap value is set as value in ε is set as [0.9,1];

Step 3.2: establish H ₀represent to fall down and do not occur, and establish H ₁represent and fall down generation, λ is default detection threshold threshold value, then, when time, expression is fallen down and is not occurred, and falling down monitoring algorithm program Output rusults is H ₀, then re-execute step 3.2;

When time, then fall down monitoring algorithm programmed decision for falling down generation, program Output rusults is H ₁, fall down monitoring algorithm and make PC industrial computer record and store this Output rusults data, then establish re-execute step 3.2, and then realize falling down behavior method of real-time based on the old solitary people of thermal infrared sensor array.

The invention has the beneficial effects as follows: the method, while carrying out real-time high-efficiency monitoring to the behavior of falling down, effectively protects individual privacy, and to daily life without bothering, the round-the-clock monitoring without blind spot can be carried out to target area.

Accompanying drawing explanation

Fig. 1 is thermal infrared sensor behavioral data acquisition system topological structure of the present invention;

Fig. 2 is the circuit theory diagrams of the thermal infrared sensor module in thermal infrared sensor behavioral data acquisition system of the present invention;

Fig. 3 is the circuit theory diagrams of the controller module in thermal infrared sensor behavioral data acquisition system of the present invention;

Fig. 4 is the design sketch of personage location under the top view camera application of thermal infrared sensor array of the present invention (4*5 divide plant).

Detailed description of the invention

The old solitary people that the present invention is based on thermal infrared sensor array is fallen down behavior method of real-time and is divided into two steps: the design of thermal infrared sensor behavioral data acquisition system, the design basis ground motion of behavior of the falling down real time monitoring algorithm of Behavior-based control data.

Step 1: the design of thermal infrared sensor behavioral data acquisition system

1.1) maximum 128 thermal infrared sensors of behavioral data acquisition system easily extensible, and be installed on indoor roof.Fig. 1 shows the topological structure of thermal infrared sensor array, has 8 interface area (being drawn by controller) and constitutes bus network.Wherein, the sensor of band T is the end sensor in network.Bus also can branch, as shown in the part of Fig. 1 lower left.Each interface area easily extensible to 16 sensor node, therefore maximum easily extensible 128 sensor nodes in sensor topology network, fully can meet the mounting arrangement in single room or apartment.

Controller (shown in Fig. 3) has been come by data packets the request of sensor assembly (shown in Fig. 4) and the response of sensor assembly to controller.Packet comprises only 8, and minimum 4 (the 0th to the 3rd) represent the address of each sensor assembly, can be arranged by thumb-acting switch.When human action being detected, the 4th will be set to 1.5th is the LED instruction on sensor node, is set to 1 when lamp is bright.7th is usually set to 0, is set to 1 when system errors.

1.2) Fig. 2 is the sensor assembly electronic-circuit diagram in the present invention, and passive heat infrared sensor is also called infrared motion sensor, the variations in temperature that the action of the human body different from ambient temperature or other object causes can be detected.NaPiOn(AMN11111 is adopted, SUNX in sensor assembly) inductive probe.It comprises 16 camera lenses, is used for gathering the infrared ray exposing to four quadrants in thermal infrared inductive probe surface.When on the roof that this sensor is arranged on high 2.5 meters, infrared detection scope can reach 7.42m*5.66m.

The sensor AMN11111 adopts Matsushita Electric Industries's infrared ray sensor, be Matsushita Electric Industries in the title that on January 1st, 2012 is issued be the infrared ray sensor in the data of " the burnt electric type MP motion sensor NaPiOn of motion sensor ", this file name is: bltn_cn_mp.pdf; Referral web site is: http://device.panasonic.cn/ac/c/dl/catalog/index.jsp series_cd=1391 & part_no=AMN11111.

Infrared sensor AMN11111 controls outfan (Output or Out) according to body temperature, and when detecting that human body enters surveyed area, outfan exports as numeral 1, otherwise exports as numeral 0.The digital output end of described infrared sensor AMN11111 connects the A2 pin of microprocessor PIC16F84A, and as the 4th of 8 bit data transmission package.The address information that low 4 (the 0th to the 3rd) of 8 bit data transmission package is each sensor assembly, connects B0, B1, B2, B3 pin of microprocessor PIC16F84A, is set by thumb-acting switch DIP-SW.The 5th of 8 bit data transmission package is LED light video data, connects the B4 pin of microprocessor PIC16F84A.7th is generally 0, is set to 1 during system errors, connects other pin of microprocessor PIC16F84A.6th for retaining position.Low-power consumption transceiver MAX489 in accompanying drawing, is applicable to RS422 standard traffic, can carry out full duplex communication, have short-circuit current-limiting function, realize data-transformation facility in the present system, be connected between microprocessor PIC16F84A and bus RS422.Last 8 bit data transmission package be conveyed through 6P4C(RJ14) standard interface completes.

1.3) Fig. 3 is the controller module electronic-circuit diagram in the present invention, and controller module comprises singlechip microprocessor MCU(PIC16F873), USB transducer (ELECOMUC-SGT) and adopt eight transceivers of RS-422 communication standard.

(address is indicated by low 4 of 8 bit data transmission package for the data of each sensor assembly, behavioral data is indicated by the 4th) send to controller module after, microprocessor PIC16F873 extracts and arranges behavior two-value data, its implementation is: the RB0 to RB7 on microprocessor PIC16F873 amounts to 8 pins and connects a 6P4C(RJ14 respectively) standard interface, each standard interface can connect maximum 16 sensor nodes (4 thumb-acting switch: can set 2 ⁴individual sensor node address), amount to and maximumly connect 128 sensor nodes.Finally, two-value behavioral data and sensor node address information are delivered to computer by USB transducer (ELECOMUC-SGT) by serial communication and are shown.Finally, according to actual needs, the factors such as such as room-size, design and develop out some sensor assemblies and a controller module, the topological structure composition behavioral data acquisition system according to Fig. 1.

Step 2: behavior of the falling down real time monitoring algorithm design basis ground motion of Behavior-based control data

Because the data precision of thermal infrared sensor array acquisition is low, be only 0,1 two-value data of reaction action severe degree.Therefore, invention applies the top view camcorder method of roof sensor array, by two-value data 0,1 is converted into the continuous data on interval [0,1], improves the gray level of pixel (sensing station).Meanwhile, of the present inventionly the resolution that behavior monitoring algorithm improves monitored area is fallen down.

Core algorithm of the present invention-fall down behavior monitoring algorithm is as follows:

With expression formula s _ij(t), (i=1 ..., m, j=1 ..., n) represent the sensor active state (0 or 1) being positioned at (i, j) position in t.When sample frequency is set to H (Hz), the pixel value of definition t (t>H/2) this pixel of moment:

$p_{i,j}\left(t\right)=\frac{1}{H+1}\underset{u=t-H/2}{\overset{t+H/2}{\mathrm{\Σ}}}{s}_{i,j}\left(u\right)---\left(1\right)$

p _i,j(t)∈[0,1]。Suppose that t has N(N≤m*n) individually enliven sensor, position is represented by (i, j), makes its pixel value be p _ij, utilize weighted mean method to locate personage at moment t:

$(x_t,y_t)=\frac{{\mathrm{\Σ}}_{i=1}^m{\mathrm{\Σ}}_{j=1}^n(i,j)p_{i,j}\left(t\right)}{{\mathrm{\Σ}}_{i=1}^m{\mathrm{\Σ}}_{j=1}^n{p}_{i,j}\left(t\right)}---\left(2\right)$

Effect after being positioned personage by formula (1) and (2) as shown in Figure 4, in the indoor environment of room layout shown in Fig. 4 upper left corner, behavior sequence: walk, clear up desk, be sitting on sofa, see TV, withdraw from a room etc. and carry out successively, amount within 20 seconds every 2 seconds and select a view data (top view picture: t=0 ... t=9).The bright dark degree of each pixel (namely often opening the square-shaped frame in image) represents its pixel value size (in [0,1] interval, calculated and obtain by formula (1)).The white circular cast often opened in image represents the position t experimenter (calculate according to the pixel value of each pixel and formula (2) and obtain) estimated.

The time series P of pixel value vector _t={ p ₍₁₎, p ₍₂₎..., p _(t), p _(t)=(p _1,1(t) ..., p _m,n(t)), the pixel value vector p of current time _(t)about pixel value vector time series P _tsingular value be defined as

s _t＝||p _(t)-c|| （3）

C is cluster centre, obtains it be expressed as by averaging method || || be Euclidean distance, Euclidean distance embodies current time pixel value vector p _(t)and the diversity between the cluster centre c of pixel value vector time series.

The present invention adopts and measures pixel value vector p with acc power halter strap (randomized power martingale) _(t)about pixel value vector time series P _thalter strap value, it is defined as

$M_n^{\left(\mathrm{\ϵ}\right)}=\underset{t=1}{\overset{n}{\mathrm{\Π}}}\left(\mathrm{\ϵ}{\hat{q}}_t^{\mathrm{\ϵ}-1}\right)---\left(4\right),$

ε ∈ [0,1], be worth by value function calculates:

${\hat{q}}_t\left(\right\{p\left(1\right),p\left(2\right),...,p\left(t\right)\},{\mathrm{\θ}}_t)=\frac{\#\{r:s_r>s_t\}+{\mathrm{\θ}}_t\#\{r:s_r=s_t\}}{t}---\left(5\right)$

In formula (5), #{r:s _r> s _trepresent and work as s _r> s _ttime r counting.S _rfor in the singular value of moment r measured by formula (3), r=1,2 ... t.θ _tevenly randomly drawed by interval [0,1] at moment t.Initial halter strap value is set as value in ε is set as [0.9,1].

Carrying out falling down in the process of monitoring with halter strap framework, after a new data sample (the pixel value vector of moment t) is acquired, is adopting testing of hypothesis to differentiate to fall down behavior and whether occur: H ₀do not occur for falling down, H ₁for falling down generation.Meet

$0<M_n^{\left(\mathrm{\&epsiv;}\right)}<\mathrm{\&lambda;}---\left(6\right)$

Time, halter strap detects and is undertaken continuing.Wherein, λ is the detection threshold threshold value that user sets.When time, suppose H ₀be rejected, abnormal (namely falling down behavior) is monitored to, and halter strap detects and restarts:

Due to { M _n: 0 < n < ∞ } be non-negative halter strap value, and E (M _n)=E (M ₀)=1, according to Du's cloth martingale spaces maximal inequality (Doob ' s maximal inequality), has

$P(\underset{k\&le;n}{\max }{M}_k\&GreaterEqual;\mathrm{\&lambda;})\&le;\frac{1}{\mathrm{\&lambda;}},\mathrm{\&lambda;}>0,n\&Element;N---\left(7\right)$

Its meaning is, for any halter strap value M _k, the probability with higher value is very little, when halter strap value is greater than λ, and H ₀be rejected.Falling down in monitoring, when behavior of not falling down occurs, inequality (7) provides the upper bound for false alarm rate (False Alarm Rate, FAR).

Of the present inventionly fall down monitoring algorithm, namely halter strap detects (Martingale Test, MT), and false code is as follows:

Initialize: M (0)=1; T=1; P _t={ }.

Setting: λ.

1: enter circulation loop

2: obtain pixel value vector P (t) at moment t.

3: if P _t={ }:

4: singular value P (t) :=0. of setting P (t)

5: otherwise:

6: calculate P _tin the singular value of each sample and P (t).

7: calculate value (formula 5).

8: calculate M (t) (formula 4).

9: if M (t) > is λ:

10: monitor the behavior of falling down, system alarm indicates.

11: make M (t)=1.

12: reinitialize P _t={ }.

13: otherwise:

14: add P (t) to time series P _t.

15: make time loyal t:=t+1.

18: terminate epicycle circulation loop.

Claims (1)

1. the old solitary people based on thermal infrared sensor array falls down behavior method of real-time, and it is characterized in that, the method comprises the steps:

Step one: adopt thermal infrared sensor data collecting system to carry out real-time data acquisition to the crawler behavior of indoor old solitary people;

Described thermal infrared sensor data collecting system comprises controller module, thermal infrared sensor array and PC industrial computer; Multiple thermal infrared sensor is arranged at indoor roof and is formed top view video camera array, for taking the crawler behavior real time data in monitor area; Controller module comprises MCU singlechip microprocessor, USB transducer and multiple RS-422 transceiver; Multiple thermal infrared sensor forms a thermal infrared sensor bus by serial or parallel connection, and every bar thermal infrared sensor bus is connected with MCU singlechip microprocessor by a RS-422 transceiver; Containing falling down monitoring algorithm in PC industrial computer, it is connected with MCU singlechip microprocessor by USB transducer, PC industrial computer sends to MCU singlechip microprocessor and starts or stops instruction, and receives and store the crawler behavior real time data transmitted by MCU singlechip microprocessor;

Step 2: mathematical modeling is carried out to the real time position of old solitary people indoor residing for it that step one monitors;

With expression formula s _ijt () expression is positioned at (i in t, j) the sensor active state 0 or 1,0 of position represents sensor an inactive state, and 1 represents sensor active state, when sample frequency is set to H, the pixel value of definition t institute position location pixel is:

$p_{i,j}\left(t\right)=\frac{1}{H+1}\underset{u=t-H/2}{\overset{t+H/2}{\mathrm{\&Sigma;}}}{s}_{i,j}\left(u\right)---\left(1\right)$

I=1 ..., m, j=1 ..., n, t>H/2, p _i,j(t) ∈ [0,1]; Suppose that t has and N number ofly enliven sensor, N≤m*n, position is represented by (i, j), makes its pixel value be p _ij, utilize calculated with weighted average method to go out t to the position that old solitary people is located to be:

$(x_t,y_t)=\frac{{\mathrm{\&Sigma;}}_{i=1}^m{\mathrm{\&Sigma;}}_{j=1}^n(i,j)p_{i,j}\left(t\right)}{{\mathrm{\&Sigma;}}_{i=1}^m{\mathrm{\&Sigma;}}_{j=1}^n{p}_{i,j}\left(t\right)}---\left(2\right)$

The time series P of pixel value vector _t={ p ₍₁₎, p ₍₂₎..., p _(t), p _(t)=(p _1,1(t) ..., p _m,n(t)), the pixel value vector p of current time _(t)about pixel value vector time series P _tsingular value be defined as:

s _t＝||p _(t)-c|| (3)，

C is cluster centre, and its expression formula is || || be Euclidean distance;

Step 3: utilize the locus p falling down monitoring algorithm and step 2 is located _i,jt () judges, judge monitor that whether personage's state is as falling down, it specifically comprises following sub-step:

Step 3.1: adopt with acc power halter strap tolerance pixel value vector p _(t)about pixel value vector time series P _thalter strap value, it is defined as

$M_n^{\left(\mathrm{\&epsiv;}\right)}=\underset{t=1}{\overset{n}{\mathrm{\&Pi;}}}\left(\mathrm{\&epsiv;}{\hat{q}}_t^{\mathrm{\&epsiv;}-1}\right)---\left(4\right),$

ε ∈ [0,1], be worth by -value function calculates:

${\hat{q}}_t\left(\right\{p\left(1\right),p\left(2\right),...,p\left(t\right)\},{\mathrm{\&theta;}}_t)=\frac{\#\{r:s_r>s_t\}+{\mathrm{\&theta;}}_t\#\{r:s_r=s_t\}}{t}---\left(5\right),$

In formula (5), #{r:s _r> s _trepresent and work as s _r> s _ttime r counting; s _rfor in the singular value of moment r measured by formula (3), r=1,2 ... t; θ _tevenly randomly drawed by interval [0,1] at moment t; Initial halter strap value is set as value in ε is set as [0.9,1];

Step 3.2: establish H ₀represent to fall down and do not occur, and establish H ₁represent and fall down generation, λ is default detection threshold threshold value, then, when time, expression is fallen down and is not occurred, and falling down monitoring algorithm program Output rusults is H ₀, then re-execute step 3.2;

When time, then fall down monitoring algorithm programmed decision for falling down generation, program Output rusults is H ₁, fall down monitoring algorithm and make PC industrial computer record and store this Output rusults data, then establish re-execute step 3.2, and then realize falling down behavior method of real-time based on the old solitary people of thermal infrared sensor array.