All sensors send data I−frameSensors send updates when their value changes P_frame.
Trang 1Prediction-based Monitoring in Sensor Networks: Taking
Lessons from MPEG
Samir Goel
Department of Computer Science
Rutgers, The State University Of New Jersey
Piscataway, NJ 08854
gsamir@cs.rutgers.edu
Tomasz Imielinski
Department of Computer Science Rutgers, The State University Of New Jersey
Piscataway, NJ 08854
imielins@cs.rutgers.edu
ABSTRACT
! "#
$
! %&' "(
*
/
1
1
Æ
56667787789:);
( !
( &;= %8>
8)' 1
"
%9' #
(
(
1
! "
Trang 2
#$ ! % &'#$ (
#
! "# %8:' /
; ! "#
$"!-5 1
! "#
@
(
1.1 Background: Brief review of MPEG
en-coding process
! "#&
! "#&
! "#
8 ) %%*
1
(
(
& %%*
! "# (
1
! "#
! % 1
!% 1
/.&0
? &
1
! "#&
- @
+! ,& /? 80 @
Trang 3
!
!
1.2 Organization of the paper
$"!-5
;
6
9
1
%9'
14!
%9' 1 14!
%% %
/.;0
$"!-55 (
$"
!-5 ?
(
Trang 4
sensor base−station wired base−station
!
3.1 A naive approach to monitoring
! " 1
1
3.1.1 Prediction Based Monitoring (PREMON)
-%%. "
%8E' /
(
(
1
! %
G
3.1.2 Classes of prediction models
)* 1
"( 2 & %% / %
%,% % & %% 0 , %
%"
* 1
"( 2 & %% / % %
&,% & , %"
)! * "(
2 &%%/ %%,%
%, & .%%0, %"
1
&
F
"( 2 &% %% / %%1
1 2 %3"
1
(
@
3.1.3 Key characteristics of PREMON
1
%:'
$"!-5
Trang 5
( H
/ 0 (
3.1.4 Classes of Solutions
%! @ $"
!-5 (
(
.
. @
? (
@ (
3.1.5 A New Framework for Visualizing the Problem
3 < (
%!&
%!
&
+ @
C
(
@ #
! "#%8:'G
%!&
0 ! "#
1
1
! "#
! "#
(
0 1
! "#
0
$"!-5 (
3.1.6 Differences between MPEG and PREMON
1
1
! "#
1
$"!-5
! "#
Trang 6
All sensors send data (I−frame)
Sensors send updates when their value changes (P_frame) Dots indicate the sesors whose alue has changed since the last frame
frames update
Sensor Field sensor
1 ! "#
$"
!-5 #
( 877 0
(
/ 0
(
1
1 ! "#
!-51 @ $"!-5
1
( - (
@
3.1.7 Architecture
1
! "#
1 ! "#
2
(
1
(14!
# ! "#
@
@
(
Trang 7
! "#
5
( 1
0
@ %6'
1
%" 1
(
+ (
1 (
(
(
(&
! "# / (80
( H
1
8 & 3 <
!
! /
0 ? !
& ; ; E
1
8 4
%&' 3- IH#H3-
7 8
0
1
1
(
5
#"-1"! -$3 1
D
(
( (; E
%
3* 2
&
& +
& 1
& +
* 1
0
4%* 1
/
0
5*1
Trang 8T = t0
T = t0+1
T = t0+2
T =t0+2
Frame#1
Frame#2
Frame#3
macro−block
dy
dy
dy dx
dx
Motion vector = <dx,dy>
dx
' ( (
3.1.8 Effect of losses
1
(
0
4 EXPERIMENTS: SETUP AND RESULTS
$"!-5
(
4.1 Experimental Setup
(
$ ! %&' /. 60
/1!"3J$:73+9);) E!H 0
1-+%:' %;'
! (
*
- !
!
! 87
%:'
7) L 8 ?
( 8:&K/IB 0
1 !
$ ! /.60 1
7 > 7 >
* + )
4.1.1 Creating a motion-sensor
!
Trang 9,
! /
1
(
/?>0
?9
1
( ! 1 (
!
4.1.2 Setup
"
? $"!-5
4.1.3 Generating predictions
?
%
"
- (
! )/ )$
% (
1
113.
(
)
;
(
78 38 78! 38
2 .!!&,
. &,!!. ?
4 &, & ,
&0
Trang 10
/ 90
!%& ( .9
+
( /
0
1 (
Æ
1 %8&'
%8M &NM'
( '1 (9
:'1 (9 !Æ
'1 (9 1Æ;
1 (
(
4.2 Results and Analysis
(
E8&/.880 1 (
$"!-5 ;
0 1 &
#
( @
@
$"!-5
8 * O82 %%! % *
#
( /. 880
& * O&2 <%!% &% 3 %!
%*
$"!-5
; * O;2 <%!% &% , %!
%*
0
$"!-5 @
8
$
/ 0 /LI0 /L0
*
*
!
!
)
2 3 4
4.2.1 Case#1: Default mode
8977 /;7(670 /
2
$
1 8
...
1.1 Background: Brief review of MPEG
en-coding process
! "#&
! "#&... ! "
Trang 2
#$ ! % &''#$ (