Algorithms For Data Compression in Wireless Computing Systems

Algorithms for Data Compression in Wireless
Computing Systems

Upasana Mahajan, Dr. Prashanth C.S.R

CSE, VTU, Bangalore, Karnataka, India

Professor & HOD, Dept. of CSE, NHCE, VTU,
Bangalore, Karnataka, India

Abstract
Compression is a technique used for reducing data
size by manipulating data redundancy; so that the
packet transmission time and storage cost can be
reduced. This can be achieved with the use of
suitable data compression algorithms. Choosing the
right algorithm can be accomplished by analyzing the
performance of the algorithm. This paper presents the
survey of various lossless data compression
algorithms.
Keywords: Lossless data compression, irreducible
substitution tables, wireless sensor networks,
compression algorithms, compression ratio.

1. INTRODUCTION
Data Compression can be defined as encoding the
information using the small number of bits rather
than original representation. There are two types of
data compression, lossless and lossy compression.
The lossy compression is a method of data encoding,
in which compression is done by discarding/losing
some data. This is commonly used in multimedia
data, especially in applications like streaming media
and internet telephony. In this some loss of
information is acceptable. Dropping nonessential
detail from the data source can save storage
space. There are two basic lossy compression
schemes
1
lossy transform codecs and lossy predictive
codecs. The lossless data compression can be defined
as reducing the bits by identifying and
eliminating statistical redundancy. The lossless data
compression is reversible of lossy compression, such
that the exact original data to be reconstructed from
the compressed data. Lossless compression can be
used for images, audios etc. but mostly it is used for
text data like executable program, text documents
and source code. In this paper focus is only on the
lossless data compression. There are different types
of lossless data compression algorithms
2
like
Huffmans coding, Run Length encoding, Dictionary
coders (LZW) etc. Based on the algorithm
performance factors
10
like compression ratio, saving
percentage and compression time, we choose the
algorithm for compressing the data. The ultimate goal
is to study different algorithms and select the best for
compression.
2. LITERATURE SURVERY
2.1. Lossless Data Compression Algorithms Based
on Substitution Tables [3] [4]
This paper introduces a class of new lossless data
compression algorithm. Each algorithm first tries to
transforms the original data, which is to be
compressed into an irreducible table representation
and then uses an arithmetic code to compress the
irreducible table representation. These are generally
known as universal coding algorithms as they try to
achieve the compression rate. These new range of
lossless data compression algorithm has been
developed to improve overall compression rate and
performance with the help of different variants of
hierarchical transformations.

Firstly, the tables are formed with the help of parallel
substitution which ends up with a unique string using
the reduction rules. In this research 5 different
reduction rules have been implemented and with the
help of which, less complex tables are formed. For
example, Let x be a string from A which is to be
IJCSI International Journal of Computer Science Issues, Vol. 10, Issue 5, No 1, September 2013
ISSN (Print): 1694-0814 | ISSN (Online): 1694-0784
www.IJCSI.org 71
Copyright (c) 2013 International Journal of Computer Science Issues. All Rights Reserved.

compressed. Starting from the table T consisting of
only one row (s, x), a hierarchical transformation
applies repeatedly the reduction rules 1-5 in some
order to reduce T to an irreducible substitution table.
To compress x, the corresponding algorithm then
uses the zero order arithmetic code to compress the
irreducible table. After receiving the code word of T',
one can fully recover T' from which x can be
obtained via parallel substitution. Some examples of
hierarchical transformation are Greedy Sequential
Transformation, SEQUITUR Transformation,
Multilevel Pattern Matching Transformation (MPM),
The greedy sequential transformation parses the
sequence, into non-overlapping substring and build a
sequentially an irreducible table for each substring.
This algorithm helps in sequential compression. The
SEQUITUR algorithm has two main rules:
1. No pair of adjacent symbols appears more
than once in the grammar.
2. Every rule is used more than once.
This helps to build irreducible table for each prefix
and then append a substring to the end of the row at
last apply the reduction rules 1-5 to reduce the table.
It transforms the binary sequence.
The MPM transformation bisects each distinct
substring repeatedly, until the length of substring is 2.
Then, assign a unique token to each substring and
create a substitution table. The MPM code and
Lempel-Ziv code have similarities like both are pure
pattern matching codes, so they do not directly
compress the data. But there are differences like
MPM is a hierarchical transformation so it does
pattern matching at multiple levels and the LZE is
non-hierarchical. The MPM code was developed for,
strictly for data of length a power of two, and named
the bisection algorithm.
This research helps in trying to solve a problem of
performance of an algorithm. It can be evaluated
mainly by calculating and comparing the two facts:
Frequency of a block of a sequence and Empirical
Entropy of a sequence.

2.2. A Simple Algorithm for Data Compression in
Wireless Sensor Networks [5] [6]
Sensor Nodes have small batteries which cannot be
changed or recharged frequently, so the WSN have
an issue of Energy. Power saving can be done by
either duty cycling (coordinated sleep/wakeup
schedules between nodes) or by in-network
processing (compression/aggregation techniques).
Data compression is the best option and appreciated
only if the execution of compression algorithms
requires lesser amount of energy than the one saved
in reducing transmission. This paper introduces the
algorithm known as Lossless Entropy Compression
(LEC), which shows the correlation between the data
collected by sensor nodes and the entropy
compression. This algorithm follows same scheme
used in baseline JPEG algorithm for compressing the
DC-coefficients of a digital image. The Huffman
table proposed in JPEG to entropy encoding the
groups has been adopted.

Fig.1 Pseudo-code of the encode algorithm
The difference d
i
computed by algorithm for the input
to an entropy encoder. The d
i
= bs
i
(bit sequence) =
s
i
|a
i,
s
i
codifies the number n
i
and a
i
represents d
i
. If
i.
d
i
> 0, a
i
= n
i
low order bits of the 2s
complement representation of d
i.
ii.
d
i
< 0, a
i
= n
i
low order bits of the 2s
complement representation of d
i
1.

iii.
d
i
= 0, s
i
is coded as 00 and a
i
is not
represented.

Encode (d
i
, Table)

IF d
i
=0 THEN
SET n
i
= 0
ELSE
SET n
i
= log
2
(|d
i
|)
ENDIF
SET s
i
TO Table [n
i
]
IF n
i
= 0 THEN
SET bs
i
= s
i

ELSE
IF d
i
> 0 THEN
SET a
i
= (d
i
)|
ni
ELSE
SET a
i
= (d
i
1)|
ni

ENDIF
SET bs
i
TO <<s
i
, a
i
>>
ENDIF
RETURN bs
i

www.IJCSI.org 72

Table1: Huffman variable length codes used in the experiment.

EXPERIMENTAL RESULTS:
The performance of a compressed algorithm can be
defined by compression ratio as shown below:

With the help of datasheets given in SHT11 for
temperature & relative humidity and using the above
formula, following ratios are obtained. Thus the
comparison between other compression algorithm
ratios as per following results shows that the LEC
algorithm performs better.

2.3. Online Adaptive Compression in Delay
Sensitive Wireless Sensor Networks [7] [8]
In wireless sensor networks (WSN), compression
reduces the data size by exploiting the redundancy
residing in sensing data. This reduction of the data
can be measured as compression ratio which is
calculated as original data size divided by the
compressed data size. The higher the compression
ratio means more data reduction is done and results in
shorter communication delays. To understand the
effect of compression, firstly obtain the processing
time of compression, which depends on several
factors like compression algorithm, CPU frequency,
processor architecture and compression data. There
are so many compression algorithm have been
developed, but one of the best is Lempel-Ziv-Welch
(LZW). LZW is a dictionary based lossless
compression algorithm suitable for sensor nodes
which replaces the strings of characters with single
codes in the dictionary. To calculate the compression
delay, the software estimation approach is adopted.
The source code of this algorithm is written in C and
then converted into the assembly codes, which have
fixed number of execution cycles.

Fig. 2 the LZW Compression Algorithm
The total count of cycles can be obtained at the
completion of LZW algorithm. The processing time
of the algorithm can be calculated by dividing the
total execution cycles by the working (i.e. CPU)
frequency. There are different experiments conducted
in the NS-2 simulator to check out the effect of
n
i
s
i
d
i
0 00 0
1 010 -1,+1
2 011 -3,-2,+2,+3
3 100 -7..-4,+4..+7
4 101 -15..-8,+8..+15
5 110 -31..-16,+16..+31
6 1110 -63..-32,+32..+63
7 11110 -127..-64,+64..+127
8 111110 -255..-128,+128..+255
9 1111110 -511..-256,+256..+511
10 11111110 -1023..-512,+512..+1023
11 111111110 -2047..-1024,+1024..+2047
12 1111111110 -4095..-2048,+2048..+4095
13 11111111110 -8191..-4096,+4096..+8191
14 111111111110 -16383..-8192,+8192..+16383
Temperature Relative Humidity
Comp_size Comp_ratio Comp_size Comp_r
atio
LEC 7605bits 66.99% 7527bits 67.33%
S-
LZW
16760bits 27.25% 13232bits 42.57%
Gzip 15960bits 30.73% 13320bits 42.19%
Bzip2 15992bits 30.59% 13120bits 43.05%
STRING = get the first character
while there are still input character
C = get next character
look up STRING + C in the dictionary
if STRING+C is in the dictionary
STRING = STRING + C
else
output the code for STRING
add STRING+C to the dictionary
STRING = C
end if
end while
output the code for STRING
Orig _Size
Comp_Ratio = 100 X 1 -
Comp_Size
www.IJCSI.org 73

ACS

compression on the packet delays. The results of the
experiments shows that delay can cause severe
performance degradation under light traffic load and
if traffic loads is heavy than compression reduces the
delay of packet, increase the maximum throughput.
So the compression is favored only when the packet
generation rate is higher than the threshold rate.
Therefore to determine whether the compression of
data is required or not the online adaptive algorithm
has been developed.
The adaptive compression algorithm is distributively
implemented on each sensor node as ACS (Adaptive
Compression Service) in an individual layer created
in a network stack. The main goal of this algorithm is
to take a right decision, that whether packet
transmission is required or not at a particular node.
Before moving to algorithm, lets have a look of the
architecture of ACS. There are 4 functional units: 1)
Controller manages the traffic flow and makes
compression decisions on each incoming packet in
this layer. 2) The LZW compressor performs actual
compression of packet with the help of LZW
algorithm. 3) The information collector helps in
collecting local statics information about network and
hardware conditions. 4) The packet buffer helps in
temporarily storing the packets to be compressed.

Statistics data Packet
Fig.3 Architecture of ACS
As compression is managed by the node state, so the
adaptive algorithm helps to determine the node state.
In this algorithm the utilization of the queuing model
is done for estimation of the node state conditions.
The queuing model includes the network model and
the MAC model. The network model defines the
network topology and traffic (i.e. estimates the arrival
rates of each node). The MAC model defines the
packet service time with the help of DCF (Distributed
Coordination Function), which can be calculated as
the time when packet enters the MAC layer to the
time when packet is successfully transmitted or
discarded.
The Adaptive Compression Algorithm is divided into
two stages: Information collection and State
determination. Firstly, in ACS the information
collector collects the statistics information like
compression statistics (compression ratio r
c
, average
compression processing time T
p
, the coefficient of
variance of processing time c
p
), MAC layer service
time and packet arrival rates. Once the collector
finishes its job, the controller in the ACS defines the
state of the node i.e. whether compression is required
or not. For making the decision the following State
Determination Procedure has been adopted which is
performed at the end of each time slot for a node in a
No-Compression state.

Fig. 4 State Determination Procedure

Upper Layer
C
o
n
t
r
o
l
l
e
r

L
Z
W

C
o
m
p
r
e
s
s
o
r

C
o
m
p
r
e
s
s
i
n

Information
collector
Buffer
Mac Layer
For each node at level i:
if state = No-Compression then
read statistics from the information collector
compute T
com
, T
min

if T
com
T
min
then
set state to Compression
else
set i to the nodes level number
T
mac
= 0
while i > 0
calculate
i
and
compute reduction T
mac
(i)
add
i
T
mac
(i) to T
mac

decrease i by one
end while
if
c
T
com
T
mac
then
set state to Compression
end if
end if
www.IJCSI.org 74

T
com
the average packet waiting time at the
compression queue.
T
min
lower bound of total delay reduction

T
mac
MAC layer service time
T
mac
(i) Delay reduction in level i
c
Arrival rate compression
i
Mean arrival rate for nodes in level i
With help of the queuing model, it is possible
to calculate the terms/equation used in algorithm. So
the outcome this paper is that using the online
adaptive compression algorithm, each node can
decide whether the packet is compressed or not,
adapting to the current network and hardware
environment.
2.4. A Statistical Lempel-Ziv Compression
Algorithm for Personal Digital Assistant (PDA)
[9]
This paper introduces a compression algorithm
named as Statistical Lempel-Ziv Compression
algorithm (SLZ), which is suitable for the
applications of hand held PDAs and can be viewed as
a variant of LZ77.
The first step of the algorithm is to build a dictionary
which may include up to 2
|c|
entries (supposing each
fixed length codeword c contains |c| bits). To build a
good dictionary, a two pass approach is adopted. The
first pass is to collect most useful phrases from the
file for building a dictionary. The second pass is to do
compression by creating codewords that refer to the
phrases in the dictionary. While building a dictionary,
there must be a balance between the dictionary size
and codeword length to avoid large number of
phrases. Therefore for a file of T symbols long, the
total number of phrases will be:
T-1
i=0
T-i =T
2
+ T O(T
2
) (1)
2
The sliding window approach which has been used in
LZ77 can be adopted to reduce the number of
phrases. Lets imagine the sliding window of size W
symbols, such that W<<T then

W-1
i=0
T-i = W(2T W + 1) O(WT)
2 (2)
Once done with the number of phrases, time is to
decide which phrases have to collect. If the phrase
collected from the file is found in the dictionary, then
there is no need to add that phrase in the dictionary
but the number of counts/ frequency of that phrase is
incremented. On the other hand, two identical phrases
having overlap in the input file must be counted as
single occurrence instead of two. This overlap
detection can be done by adding a time stamp (last
time at which the phrase occurred in the file) to each
entry in the dictionary. When a phrase is fetched
from the input file and an identical phrase is found in
the dictionary, compare the timestamp of that phrase
in the dictionary with the current time stamp. If time
stamp difference is less than the phrase length,
overlap is detected.
After collecting the phrases, time to put all phrases in
the dictionary with respect of dictionary size. The
size of dictionary should not be too large and too
small; it must contain all useful phrases. The number
of entries in the dictionary can be reduced by pruning
the phrases having unit frequency. Which means
prune the phrases which occur once in a file not the
phrases that are one symbol long, and it can be done
at end of the first pass. With this method most of the
time the newly appears phrases are purged. So to
avoid this problem another method of pruning the
phrases known as Move-To-Front approach is used.
In this approach, when a new phrase inserted it is
move to the front of the dictionary. The time
dictionary is full, discard the phrase at the end of the
dictionary. With this method the phrases which have
high frequency will be at the front and phrases with
least frequency located at the end of the dictionary.
Once the dictionary has been build, its time for
compression which can be done with the help of
entropy coding method. The symbols of the file are
shifted into the sliding window and once its full, the
symbol sequence in window is compared with the
phrases one by one in the dictionary in the order of
entropy. When matched phrase is found, the matched
symbols in the window are coded by the index of that
phrase. The symbols that matched the phrases are
removed from the sliding window and new symbols
are moved to sliding window. As soon as window is
full repeat the process again until all the symbols get
coded.
www.IJCSI.org 75

We cant say that this is the best compression
algorithm but a simple entropy coding scheme
designed using the prefix codes to eliminate look-up
table for decoding. Using the combination of
dictionary based algorithm and sliding window
approach, the overall compression ratio decreases.
2.5. Comparison of Lossless Data Compression
Algorithms for Text Data [10]
Data compression helps in reducing the size of the
file, in other words compression represents the
information in a compact form rather than its original
form without any data loss. When data compression
is done while transmitting the data, the main concern
is speed. Speed of the transmission depends upon the
number of bits sent, the time required for the encoder
to generate the coded message and the time required
for the decoder to recover the original message.
Sometimes the lossless compression algorithms are
also known as reversible algorithms, as the original
message can achieve by the decompression process.
Some of the famous lossless compression algorithms
are Run-Length Encoding (RLE), Huffman
Encoding, Adaptive Huffman Encoding, Shannon
Fano algorithm, Arithmetic Encoding and Lempel
Zev Welch algorithms.
This paper introduces the comparison of
performances of above algorithms, based on different
factors. There are many different ways to measure the
performance of a compression algorithm. The main
concern is space and time efficiency, while
measuring the performance. Following are some
factors used to evaluate the performances of the
lossless algorithms.
Compression Ratio =size after compression
size before compression (3)

Compression Factor = size before compression
size after compression (4)

Saving Percentage =
size before compression size after compression %
size before compression
(5)
Compression Time can be defines as time taken to
compress particular file. Time taken for the
compression and decompression should be
considered separately. For a particular file, if the
compression and decompression time is less and in
an acceptable level, it means that algorithm is
acceptable with respect to time.
Entropy can be used as a performance factor, if the
compression algorithm is based on statistical
information of the source file. Let set of event be S=
{s
1
, s
2,
s
3
s
n
} for an alphabet and each s
j
is a symbol
used in this alphabet. Let the occurrence probability
of each event be p
j
for event s
j.
Then the self-
information I(s) is defined as follows:
I(s) = log
b
1/ p
j
or I(s) = - log
b
1/ p
j
(6)
The first order Entropy value H(P) can be calculated
as follows:
H(P) =
n
j=1
p
j
I(s
j
) or H(P) = -
n
j=1
p
j
I(s
j
) (7)
Code Efficiency is the ratio between the entropy of
the source and the average code length.
E (P, L) = H (P) 100%, (8)
l (P, L)
E (P, L) is the code efficiency, H (P) is entropy and
l (P, L) is the average code length.

Average code length defined as the average number
of bits required to represent a single code word. It can
be calculated as: l =
n
j=1
p
j*
l
j
, where p
j
is the
occurrence probability of j
th
symbol of the source
message, l
j
is the length of the particular code word
for that symbol and L = {l
1
, l
2
, ,l
n
}.
In order to test the performance of above mentioned
lossless compression algorithms, first step is to
implement them and then test them with some set of
files. Performances evaluated by computing above
mentioned factors. After the implementation and
testing the results shows that the Adaptive Huffman
algorithm needs larger time period for processing,
because the tree should be updated or recreated. LZW
works better as the file size grows up to certain limit,
because there are more chances of replacing the
words by using the small index number. But it cannot
be used for all cases, so cant say it is one of the
efficient algorithms.
www.IJCSI.org 76

Arithmetic Encoding algorithm has an Underflow
problem, which gives an erroneous result after few
numbers of iterations. Therefore it is not suitable for
comparison. Huffman Encoding and Shannon Fano
algorithm shows similar results except in
compression times. Shannon Fano algorithm has
faster compression time than Huffman Encoding, so
this factor can be used to determine the more efficient
algorithm from these two.
While considering the major performance factors like
compression time, decompression time and saving
percentages of the all the selected algorithms. The
Shannon Fano algorithm is considered as the most
efficient algorithm, as the values of this algorithm
lies acceptable range and it also shows better results
for the large files.
3. CONCLUSION
This study introduces data compression and simple
algorithms for compression. Each algorithm has its
own advantages and disadvantages. With the help of
various performance factors, it is easy to choose
algorithms that are more efficient. This paper
demonstrates that if we use the right data
compression techniques, it will certainly be helpful in
reducing the storage space and the computational
resources. This is definitely more critical in the case
of wireless systems where network bandwidth is
always a cause for concern.
REFERENCES
[1] https://en.wikipedia.org/wiki/Lossy_compression.
[2] https://en.wikipedia.org/wiki/Lossless_
Compression.
[3] John.C.Kieffer and En-hui Yang, Lossless Data
Compression Algorithms Based on Substitution
Tables, IEEE, 1998.
[4] J.C.Kieffer, E-H.Yang, G.Nelson, and P.Cosman,
Lossless data compression via multi-level pattern
matching, IEEE, 1996.
[5] Massimo Vecciho, A Simple Algorithm for Data
Compression in Wireless Sensor Networks,
IEEE, June 2008.
[6] https://en.wikipedia.org/wiki/Huffman_coding.
[7] Xi Deng and Yuanyuan Yang, Online Adaptive
Compression in Delay Sensitive Wireless
Sensor Networks, IEEE, October 2012.
[8] https://en.wikipedia.org/wiki/LempelZivWelch.
[9] S.Kwong and Y.F.Ho, A Statistical Lempel-Ziv
Compression Algorithm for Personal Digital
Assistant (PDA), IEEE, February 2001.
[10] S.R.Kodituwakku and U.S.Amarasinghe,
Comparison of Lossless Data Compression
Algorithms for Text Data, Indian Journal of
Computer Science and Engineering, Vol 1 No. 4
416- 425.

Upasana Mahajan is M.Tech in Computer Science
Engineering (2006) from VTU and passed B.Tech
(CSE) from Guru Nanak Dev University in year of
2006. She is currently pursuing her Ph.D and has
active research interest in area of Data Compression
and there applications in wireless networks.

Dr. Prashanth C.S.R holds B.E, M.S (USA), Ph.D
(USA). At present he is a Professor and Head of the
Department for Computer Science in New Horizon
College of Engineering Bangalore, Karnataka.
www.IJCSI.org 77

Algorithms For Data Compression in Wireless Computing Systems

Uploaded by

Copyright:

Available Formats

Algorithms For Data Compression in Wireless Computing Systems

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Algorithms For Data Compression in Wireless Computing Systems

Uploaded by

Copyright:

Available Formats

Algorithms for Data Compression in Wireless

You might also like