You are on page 1of 30

ADAPTIVE COMPRESSION TECHNIQUE FOR WIRELESS SENSOR NETWORKS

Tarinder singh M.tech 2k13/cse/24

WIRELESS SENSER NETWORK MAIN CHARACTERISTI CS OF A WSN


Power consumption constrains for

nodes using batteries or energy harvesting

bi!ity to cope with node fai!ures

Mobi!ity of nodes "ommunication fai!ures #eterogeneity of nodes

$ca!abi!ity to !arge sca!e of

dep!oyment

bi!ity to withstand harsh environmenta! conditions

%ase of use

POWER CONSUMPTION CONSTRAINS


!most &'( of the energy is spent in the communication modu!e for data transmission in wire!ess sensor network which depends upon.

Energy Conservation Sche es


)uty cyc!e
Power management Topo!ogy contro!

)ata )riven
)ata reduction

*n n/w processing )ata compression )ata prediction

)ata ac+uistion

Mobi!ity based
Mobi!e sink Mobi!e re!ay

!ATA COMPRESSION
)istributed approach
Coding by Ordering data compression Pipelined In-Network Compression Distributed Compression

,oca! approach
,ossy ,oss!ess

CO!IN" #$ OR!ERIN" !ATA COMPRESSION

-irst. a data pass from sensor nodes in the interested region to a co!!ector node t node ). the aggregated data is transmitted to node / *n coding by ordering 44( data is !oss during compression

Pi%e&ine' In( Net)or* Co %ression !istri+,te' Co %ression RELATIONSHIP

!ATA COMPRESSION
-or !oca! !oss!ess compression we use

,01 and #uffman.


#uffman coding basica!!y divided in

to two categories

$tatic #uffman coding daptive/ dynamic #uffman coding

$tatic #uffman coding suffers from the fact


that the uncompressed need have some know!edge of the probabi!ities of the symbo! in the compressed fi!es. The fi!e re+uires two passes. -*2$T P $$ finds the fre+uency of each symbo! and constructs the #uffman tree. $%"34) P $$ is used to compress the fi!e.

A!APTI-E HUFFMAN CO!IN"


To get the code for every node. in

case of binary tree we cou!d 5ust traverse a!! the path from the root to the node. writing down 6for e7amp!e8 919 if we go to the right and 9'9 if we go to the !eft.
4ot yet transmitted9 64:T8. 1hen we transmit an 4:T symbo!.

we have to transmit code for the 4:T node. then for its generic code.

A!APTI-E HUFFMAN CO!IN"


-or every symbo! transmitted both

the transmitter and receiver e7ecute the update procedure;


*f current symbo! is 4:T. add two

chi!d nodes to 4:T node. 3ne< 4:T . other <!eaf node for symbo!. *ncrease weight for the new !eaf node and the o!d 4:T and go to step 4. *f not. go to symbo!=s !eaf node.
*f this node does not have the highest

number in a b!ock. swap it with the node having the highest number. e7cept if that node is its parent
*ncrease weight for current node

*f this is not the root node go to

parent node then go to step 2. *f this is the root. end.

A!APTI-E HUFFMAN CO!IN"

E.AMPLE /TENNESSEE
T
$tage 1 6-irst occurrence of t 8 r />

' 3rder; '.t618

t618

$end binary represntation of t ? r represents the root ? ' represents the nu!! node ? t618 denotes the occurrence of T with a fre+uency of 1

TE
$tage 2 6-irst occurrence of e8

r / > 1 t618

/ > ' e618

3rder; '.e618.1.t618 $end ' then binary representaion of e.

TEN
$tage 3 6-irst occurrence of n 8

r / > 2 / > 1 / > ' n618 e618 t618

3rder; '.n618.1.e618.2.t618

Misfit
$end 1' then binary of n

Reor'er/ TEN
r / > t618 2 / 1 / > ' n618 > e618

3rder; '.n618.1.e618.t618.2

TENN
$tage 4 6 2epetition of n 8

r / > t618 3 / > 2 / > ' n628 ; e618

3rder; '.n628.2.e618.t618.3

Misfit
$end 1'1

Reor'er/ TENN

r / > n628 2 / > 1 / > ' t618 e618

3rder; '.t618.1.e618.n628.2

t618.n628 are swapped

TENNE
$tage @ 62epetition of e 8

r / >

n628 3 / > 1 / > ' t618 e628

3rder; '.t618.1.e628.n628.3

TENNES
$tage A 6-irst occurrence of s8 r / > n628 2 / > 4 / > e628

1 / > '

t618 s618

3rder; '.s618.1.t618.2.e628.n628.4

TENNESS
$tage B 62epetition of s8 r / > n628 @ / > 3 / > 2 / > t618 e628

'

s628 ;

3rder; '.s628.2.t618.3.e628.n628.@ Misfit

Reor'er/ TENNESS
r / > n628 3 / > 1 / > ' t618 s 628 @ / > e628

3rder ; '.t618.1.s628.3.e628.n628.@ s628 and t618 are swapped

TENNESSE
$tage & 6$econd repetition of e 8 r / > n628 3 / > 1 / > ' t618 s628 A / > e638

3rder ; '.t618.1.s628.3.e638.n628.A ; Misfit

Reor'er/ TENNESSE
r / > e638 3 / > 1 / > ' t618 s628 @ / > n628

3rder ; 1.t618.1.s628.3.n628.e638.@ 4628 and e638 are swapped

TENNESSEE
$tage C 6$econd repetition of e 8 r
'

/ >1 @
'

e648 3
'

/ >1 n628 s628

/ >1

1
'

/ >1 t618

'

3rder ; 1.t618.1.s628.3.n628.e648.@

ENCO!IN"
The !etters can be encoded as fo!!ows;
e;' n ; 11 s ; 1'1 t ; 1''1

MO!IFIE! A!APTI-E AL"ORITHM


-or $tatic #uffman a!gorithm we

need prior know!edge of the incoming source se+uence

-or daptive #uffman a!gorithm.

effective on!y for very fre+uent!y and first occurring data.


-urthermore. the binary tree

construction is based on the order of arriva! of incoming data.

MO!IFIE! A!APTI-E AL"ORITHM


Dses a tree with !eaves that represent

sets of symbo!s with the same fre+uency. rather than individua! symbo!s

MO!IFIE! A!APTI-E AL"ORITHM


/inary tree construction "onsidering the TEMPORAL

CORRELATION in sensor data on!y the difference of the sensor data d6i8E r6i8<r6i<18 is encoded.

FOR E.AMPLE/ 0123454260


)6i8E2.1.2.1.1.<1.<3.<2.1

set F and G difference.


$teps
$end binary of 3.

FOR E.AMPLE/ 0123454260


)6i8E2.1.2.1.1.<1.<3.<2.1 $end binay 2.

FOR E.AMPLE/ 0123454260


)6i8E2.1.2.1.1.<1.<3.<2.1 $end ' fo!!owed by binay of 1.

FOR E.AMPLE/ 0123454260

)6i8E2.1.2.1.1.<1.<3.<2.1 $end 1Eprefi7. 'Esuffi7. therefore

code E 1'

FOR E.AMPLE/ 0123454260


)6i8E2.1.2.1.1.<1.<3.<2.1 $end '1Eprefi7. 'Esuffi7. therefore

code E '1'

FOR E.AMPLE/ 0123454260


)6i8E2.1.2.1.1.<1.<3.<2.1

$end 1Eprefi7. 1'Esuffi7. therefore

code E 11'

FOR E.AMPLE/ 0123454260


)6i8E2.1.2.1.1.<1.<3.<2.1 $end 1Eprefi7. 'Esuffi7. therefore

code E 1'

FOR E.AMPLE/ 0123454260


)6i8E2.1.2.1.1.<1.<3.<2.1 $end '' then binay of 3.

FOR E.AMPLE/ 0123454260


)6i8E2.1.2.1.1.<1.<3.<2.1 $end '1'.

FOR E.AMPLE/ 0123454260


)6i8E2.1.2.1.1.<1.<3.<2.1 $end '1'.

RUN LEN"TH CO!IN"

2un !ength coding rep!aces se+uences

of the same data va!ues within a fi!e by a count number and a sing!e va!ue.
%7amp!e; /////////")%%%% The 1B byte of data string has to be

compressed as 1' bytes of ?4% -

?&/ " )

THANKIN" $OU REFERENCES


H1I Jason ,ester #i!!. K$ystem

rchitecture for 1ire!ess $ensor 4etworksL Dniversity of "a!ifornia. /erke!ey $pring 2''3. H4I *an -. kyi!diM. 1ei!ian $u. :ogesh $ankarasubramaniam. and %rda!

H2I "esare !ippi. 2omo!o "amp!ani.

"ristian Na!perti )ipartimento di %!ettronica e *nformaMione. Po!itecnico di Mi!ano. K,oss!ess "ompression Techni+ues in 1ire!ess $ensor 4etworks; Monitoring Micro acoustic %missionsL IEEE
H3I M.Parameswari K daptive

"ompression Techni+ue for


1ire!ess $ensor 4etworksL IJAIS H4I #ttp;//en.wikipedia.org/wiki/

daptiveO#uffmanOcoding

You might also like