Content uploaded by Huijuan Wu
Author content
All content in this area was uploaded by Huijuan Wu on Sep 16, 2020
Content may be subject to copyright.
Comparative study of lossy and
lossless data compression in
distributed optical fiber sensing
systems
David Atubga
Huijuan Wu
Lidong Lu
Xiaoyan Sun
David Atubga, Huijuan Wu, Lidong Lu, Xiaoyan Sun, “Comparative study of lossy and lossless data
compression in distributed optical fiber sensing systems,”Opt. Eng. 56(2), 024108 (2017),
doi: 10.1117/1.OE.56.2.024108.
Downloaded From: http://opticalengineering.spiedigitallibrary.org/ on 03/07/2017 Terms of Use: http://spiedigitallibrary.org/ss/termsofuse.aspx
Comparative study of lossy and lossless data
compression in distributed optical fiber sensing systems
David Atubga,aHuijuan Wu,a,*Lidong Lu,band Xiaoyan Sunb
aUniversity of Electronic Science and Technology of China, Key Laboratory of Optical Fiber Sensing and Communications (Ministry of Education),
School of Communication and Information Engineering, Chengdu, China
bState Grid Corporation of China, The Global Energy Interconnection Research Institute, Nanjing, Jiangsu, China
Abstract. Typical fully distributed optical fiber sensors (DOFS) with dozens of kilometers are equivalent to tens
of thousands of point sensors along the whole monitoring line, which means tens of thousands of data will be
generated for one pulse launching period. Therefore, in an all-day nonstop monitoring, large volumes of data are
created thereby triggering the demand for large storage space and high speed for data transmission. In addition,
when the monitoring length and channel numbers increase, the data also increase extensively. The task of mit-
igating large volumes of data accumulation, large storage capacity, and high-speed data transmission is, there-
fore, the aim of this paper. To demonstrate our idea, we carried out a comparative study of two lossless methods,
Huffman and Lempel Ziv Welch (LZW), with a lossy data compression algorithm, fast wavelet transform (FWT)
based on three distinctive DOFS sensing data, such as Φ-OTDR, P-OTDR, and B-OTDA. Our results demon-
strated that FWT yielded the best compression ratio with good consumption time, irrespective of errors in signal
construction of the three DOFS data. Our outcomes indicate the promising potentials of FWT which makes it
more suitable, reliable, and convenient for real-time compression of the DOFS data. Finally, it was observed that
differences in the DOFS data structure have some influence on both the compression ratio and computational
cost. ©2017 Society of Photo-Optical Instrumentation Engineers (SPIE) [DOI: 10.1117/1.OE.56.2.024108]
Keywords: data compression; distributed optical fiber sensor; fast wavelet transform; Huffman; Lempel Ziv Welch.
Paper 161216 received Aug. 14, 2016; accepted for publication Jan. 17, 2017; published online Feb. 22, 2017.
1 Introduction
In recent years, fully distributed optical fiber sensors (DOFS)
based on the optical time-domain reflectometry technolo-
gies, such as phase-sensitive optical time domain reflectom-
etry (Φ-OTDR),1polarization-sensitive optical time-domain
reflectometry (P-OTDR),2and Brillouin optical time-domain
analyzers (B-OTDA),3have been developing rapidly with the
exploration of wide applications in the field of infrastructure
security monitoring. These offer different kinds of conven-
ient and cost-effective environmental measuring methods for
many important applications in long or ultralong distance
monitoring with precise locations such as safety monitoring
of long borderlines, perimeters of airports, oil or gas pipe-
lines, power transmission lines, telecommunication cables,
large structures, among others. However, a typical DOFS
with dozens of kilometers is equivalent to tens of thousands
of point sensors along the whole monitoring line, which
means countless information will be generated for one pulse
launching period. With constant monitoring, large volumes
of data will be generated thereby triggering the demand
for large storage space and high-speed data transmission.
Moreover, the data increase several times over when the
monitoring length and channel numbers increase.4As a
result, excessive pressure for massive data storage and
speed for data transmission are required. Compressing the
amount of data to mitigate these predicaments is equivalent
to increasing the speed capacity of the communication
channels and saving the limited storage resource, which is
a necessary part for many practical applications of DOFS.
However, until recently the corresponding big data compres-
sion methods in the DOFS system have not been mentioned
in related research and applications. Thus, lossless and lossy
data compression algorithms in three typical DOFS sensing
data, such as Φ-OTDR, P-OTDR, and B-OTDA, are compa-
ratively studied in this paper. The organization of this paper
is arranged as follows: Sec. 2presents the compressibility
of DOFS sets of data. Section 3deals with typical data com-
pression methodology of both lossless and lossy ones.
Section 4compares the compression results for the above
two classes of compression methodologies in three typical
DOFS systems. Finally, Sec. 5concludes this paper with
remarks and future works.
2 Data Compression in DOFS System
2.1 Compression Requirement in DOFS System and
the Data Compressibility Analysis
For a typical OTDR-based DOFS, whether it is Φ-OTDR or
P-OTDR, its measurement data are generated and accumulated
as a two-dimensional (2-D) matrix as shown in Fig. 1(a).At
each pulse launching moment, an OTDR curve along the
fiber length is acquired by an ultrahigh sampling frequency
of about tens of MHz, and the data size depends on the fiber
length and the sampling rate. With repetitive pulse launch-
ing, more and more OTDR curves are accumulated along the
time axis and contribute to the matrix, which is in fact,
a spatial–temporal event evolving graph obtained from the
*Address all correspondence to: Huijuan Wu, E-mail: hjwu@uestc.edu.cn 0091-3286/2017/$25.00 © 2017 SPIE
Optical Engineering 024108-1 February 2017 •Vol. 56(2)
Optical Engineering 56(2), 024108 (February 2017)
Downloaded From: http://opticalengineering.spiedigitallibrary.org/ on 03/07/2017 Terms of Use: http://spiedigitallibrary.org/ss/termsofuse.aspx
differentiation of each two adjacent OTDR curves, which
demonstrates when and where the change happens. For
the BOTDA system, at one moment, the trail along the spa-
tial axis is obtained by Lorenz fitting from the scanning lines
of different frequencies at all fiber locations. Thus, the
BOTDA data are composed of layers of spatial frequency
scanning graphs at different moments as shown in Fig. 1(b),
which is a three-dimensional (3-D) matrix with frequency,
spatial and temporal axes. Whenever a curve or a 2-D graph
is obtained at one moment, a huge amount of data will
definitely be generated with the time elapsing. In general,
a monitoring range of dozens of kilometers means tens of
thousands of point sensors along the whole line, which
also means tens of thousands of data will be generated at
one pulse launching or one frequency scanning period,
e.g., for sampling a sensing fiber of 50 km with an acquis-
ition rate of 50 MHz and an optical pulse launching fre-
quency of 1 kHz, the amount of data could reach up to
1 GB per hour. In an all-day round nonstop monitoring,
the data will fill up the disk space quickly. Thus, it is urgent
and necessary to compress the data reasonably at real time
for most DOFS systems. In the DOFS safety monitoring sys-
tem, the abnormal events to be detected or measured are
always occasional and happening at local areas, and for
most times, the environments remain the same or just change
slowly. Then, the curve or graph at one moment will be sim-
ilar to that of the next moment. Thus, the original data of
different DOFS systems could be redundant and the data
similarity from moment to moment makes the compressible
space quite large.
2.2 Data Compression Index and Other Issues
When data compression is applied in information transmis-
sion, the primary goal is to achieve high speed and less stor-
age space. Speed of data transmission over the bandwidth
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 ensem-
ble. In a data storage application, although the degree of
compression is the key concern, it is nonetheless necessary
that the algorithm be efficient in order for the scheme to be
practical enough to ascertain the desired results. The design
for data compression schemes involves trade-offs among
various factors such as the computational resources required
to compress and decompress the data, as well as the amount
or degree of distortion introduced involving lossy data
compression.5,6It is very possible to compress varied
types of digital data to reduce the size of the original file
needed to be transmitted over telecommunication lines by
the bandwidth and store it, with no loss of the full informa-
tion contained in the original file. In signal processing, data
compression, bit-rate reduction, or source coding involves
encoding information using less bits than the original
representation.7Compression can either be lossless or lossy.
Lossless compression reduces the bits by first of all identi-
fying and eliminating statistical data redundancy with no loss
of information. On the other hand, lossy compression as the
preferred method in this paper reduces the bits by The
differences in the data structure of DOFS data types have
some amount of influence on the compression ratio and
both the compression and decompression.8According to
Al-Dubaee and Ahmad,9the lossless compression method
works by removing the redundant information present in
data signals. It, however, suffers from two major disadvan-
tages; it has small (poor) compression ratios and does not suf-
ficiently meet economic needs and as well does not guarantee
a constant output data rate. On the other hand, the lossy com-
pression method reduces data redundancy and provides
much higher compression than lossless compression. Also,
Palaniappan and Latifi10 proposed three character-based lossy
text compressions in 2007. They included the dropped vowels,
letter mapping, and replacement of characters which were
applied in their compression processes and obtained desired
results. By adopting lossy text compression, using the wavelet
transform or Fourier transform will deal effectively with text
files as signals. Thus, in this paper, a comparative study of
lossless and lossy data compression is conducted in the
three typical DOFS systems mentioned above.
3 Data Compression Methodologies
3.1 Lossless Compression by Huffman
Huffman coding,5,6which dates to 1952, was proposed
mostly for lossless compression in one of his papers “A
method for the construction of minimum-redundancy
codes.”It was among the first to appear and a widely
used method for lossless compression even to date. Its
usage has yielded an improved performance that considers
Fig. 1 Original data styles of different DOFS systems. (a) X-OTDR data structure for a time period.
(b) B-OTDA data structure at one moment.
Optical Engineering 024108-2 February 2017 •Vol. 56(2)
Atubga et al.: Comparative study of lossy and lossless data compression in distributed optical fiber sensing systems
Downloaded From: http://opticalengineering.spiedigitallibrary.org/ on 03/07/2017 Terms of Use: http://spiedigitallibrary.org/ss/termsofuse.aspx
the core binary-tree structure with an efficient method of
coding the characters to successfully attain desired results
based on the algorithm. In the Huffman compression algo-
rithm, probability of all the characters in the data is evaluated
first, and then the characters appearing with high frequency
are represented as shorter digits, whereas characters that
rarely appear are denoted as longer digits. Thus, the algo-
rithm performance mainly depends on the probability distri-
bution of the characters.
3.2 Lossless Compression by Lempel Ziv Welch
LZW compression coding is a variation on the basis of LZ78,
by Terry Welch’s“A technique for high-performance data
compression”11 and also improves the compression speed
in addition to inheriting LZ78 algorithm compression’s
good advantages and further simplify the code output.
The basic idea is to create a code table or string table dic-
tionary, input signal as a bitstream, then enter into a long
string or frequent characters with simple numbers of symbols
based on the dictionary to effectively run the algorithm. The
algorithm is then run by first initializing the dictionary to
include all possible characters. In this dictionary, long strings
or some combinations of letters with high frequency can be
represented as one entry, which is indexed by a short symbol.
Hence, this algorithm’s performance mainly depends on the
appearing frequency of the string and the size of the diction-
ary. It is also one of the most currently used algorithms for
lossless compression.
3.3 Lossy Compression by Fast Wavelet Transform
In this paper, we also use fast wavelet transform (FWT) as a
lossy compression method to solve the problem and compare
it with the above two lossless compression methods,
Huffman and LZW. The purpose is to overcome the weak-
nesses of the two lossless compression methods for DOFS
data compression to obtain good compression ratio and con-
sumption time. An illustration and the implementation of the
application of FWT involved seven main steps and is dis-
played in Fig. 2.
4 Comparison of the Compression Results of
Huffman, LZW, and FWT for Φ-OTDR, POTDR,
and BOTDA data
4.1 Compression Ratio and Computational Cost
Analysis
EQ-TARGET;temp:intralink-;e001;326;620compression ratio ¼the size of compressed signal
The size of uncompressed signal :(1)
The experiment was applied on Huffman, LZW, and FWT
for compression of Φ-OTDR, POTDR, and BOTDA data
via MATLAB to evaluate their practical performances. A
clear interpretation of the results generated by FWT,
Huffman, and LZW during the compression of the DOFS
sets of data ranging from compression ratio, compression
time, and decompression time in milliseconds, are given,
respectively. The compression ratio is computed as shown
in Eq. (1), and the compression and decompression time
for the different types of methods are computed in millisec-
onds. In this study, to select a suitable filter and level for the
lossy compression algorithm, FWT, Daubechies (db1 to 10)
wavelets from level 1 to level 10 according to Misiti et al.12 in
2008, were, respectively, applied in compressing the DOFS
sets of data. The results showed that the optimum compres-
sion results were achieved at Db6 level 4 as indicated in
Table 1, Figs. 3and 4. All the results are obtained when the
proposed data procession method is applied in dynamic sens-
ing networks. Table 1shows a detailed presentation of the
compression time of the proposed method in terms of com-
pression and decompression time as compared to the lossless
compression methods for the three sets of DOFS data.
Figures 3and 4show an intuitive and vivid illustration of
the compression ratio and computational cost, respectively.
In Fig. 3, the first bars from each group of DOFS data are
for FWT, second bars for Huffman and third bars for LZW.
START
Read the data
Wavelet decomposition
Apply threshold and choose some key coefficients
Reconstruct the signal with the chosen coefficients and
their locations
FWT evaluation with the compression ratio, time and
reconstruction error rate
More files?
END
YES
NO
Store the chosen coefficients and their locations
Fig. 2 Lossy compression by FWT algorithm.
Table 1 Comparison results of FWT, Huffman, and LZW.
DOFS
data type
Compression
method Ratio
Compression
time in (ms)
Decompression
time in (ms)
Φ-OTDR FWT 0.19 7.226 6.4804
Huffman 0.45 13,712 24,282
LZW 0.40 4555 1716
P-OTDR FWT 0.20 5.6623 4.4365
Huffman 0.42 354 491
LZW 0.40 999 500
B-OTDA FWT 0.23 3.3482 4.2577
Huffman 0.42 384 673
LZW 0.40 126 47
Optical Engineering 024108-3 February 2017 •Vol. 56(2)
Atubga et al.: Comparative study of lossy and lossless data compression in distributed optical fiber sensing systems
Downloaded From: http://opticalengineering.spiedigitallibrary.org/ on 03/07/2017 Terms of Use: http://spiedigitallibrary.org/ss/termsofuse.aspx
It is very lucid from the values shown below that FWT
performance in compression Φ-OTDR data produced a ratio
of 0.19 as against Huffman and LZW with 0.45 and 0.40,
respectively. In compressing P-OTDR data, FWT had a
ratio of 0.20 as against 0.42 and 0.40 for Huffman and
LZW, respectively. FWT again showed good compression
ratio results during the compressing of B-OTDA data with
0.23, whereas 0.42 and 0.40 were obtained by Huffman
and LZW accordingly. It showed that FWT as a lossy method
of compression generated better compression ratio results
of about one-half the two lossless methods of Huffman
and LZW during the compression of the DOFS sets of
data, Φ-OTDR, P-OTDR, and BOTDA data. Furthermore,
Fig. 4(a) shows the analysis of the computational cost: the
first bars indicate compression time and the second bars re-
present decompression time for each set of data. It showed
that the FWT method had in both the compression and
decompression time with respect to the three sets of DOFS
data less consumption time (7.226, 5.6623, 3.3482 ms and
6.4804, 4.4365, 4.2577 ms), respectively, for Φ-OTDR,
P-OTDR, and B-OTDA data sets followed by LZW with
4555, 999, 384 ms and 1716, 500, 47 ms, respectively,
whereas Huffman revealed the worst consumption time of
13,712, 354, 384 ms and 13,712, 354, 384 ms. In general,
FWT showed a marked difference and yielded the best com-
pression ratio and least consumption time as observed among
the three methods with respect to the three sets of DOFS
data. Huffman and LZW methods of compression did not
prove many vital margins in the compression ratio demon-
strated in Fig. 3. On the other hand, the LZW algorithm
exhibited less compression/decompression time as compared
to the Huffman algorithm especially for the Φ-OTDR and
B-OTDA data as illustrated in Fig. 4, whereas LZW turned
out to be higher than Huffman for the compression of
P-OTDR data with 999 ms as against Huffman of 354 ms.
Notwithstanding, the same compression algorithm again
showed different compression ratios in Fig. 3and different
compression/decompression time for the three different
types of DOFS data. Thus, the differences in the data struc-
ture of the DOFS sets of data have some significant impact
on both the compression ratio and the computational cost.
The above computation of the algorithm is done in a personal
computer, while the processing procedure could be sped up
when implemented in an embedded unit, which could be
future work.
4.2 Signals Reconstruction and Error Analysis
for the Three Groups of DOFS Data
The mean square error (MSE) is the cumulative squared error
between the compressed signal and the uncompressed signal
as illustrated as follows:
EQ-TARGET;temp:intralink-;e002;326;571MSE ¼1
MX
M
m¼1
½IðmÞ−I0ðmÞ2;(2)
where IðmÞis the uncompressed signal, I0ðmÞis the approxi-
mated version (which is actually the decompressed signal),
and Mis the dimension of the signal. A lower value for MSE
means lesser error and vice versa.
Even though the FWT behaves the best for the compres-
sion ratio and consumption time, it is a lossy compression
algorithm and the error rate must be considered to confirm
that the reconstructed information is acceptable. Figures 5–7
show the sample results of compression signals of FWT
at Db6 level 4 from reconstructed signals to error signals
(original signal minus reconstructed signals) for, POTDR,
Φ-OTDR, and B-OTDA sensing data. The purpose of the
error rate is to determine the efficiency of the FWT compres-
sion method. The error rate by the FWT compression also
yielded good results especially in P-OTDR and B-OTDA.
The differences in the data structure of DOFS data types
have some amount of influence on the compression ratio
and both the compression and decompression time.
Figure 5(a) shows the illustration of the reconstructed sig-
nal compared with the uncompressed signal of Φ-OTDR
data by the FWT compression method. Also, Fig. 5(b) shows
the error rate of Φ-OTDR by FWT compression method. It
was measured by differentiating the reconstruction data from
the uncompressed signal. The error rate intensity was
observed at 0.05. It is shown that the signal trends are con-
sistant with each other, however, there were some slight fluc-
tuations due to the differences between the uncompressed
data and the reconstructed data for the Φ-OTDR data.
Figure 6(a) shows the sample signal of the reconstruction
results compared with the uncompressed data of POTDR
by the FWT method. Figure 6(b) shows the error rate of
FWT in compressing the P-OTDR. The error rate intensity
was observed less than 0.001 with an insignificant reduction
on the sample axis. It is also shown that the reconstructed and
the uncompressed data trends are consistant with each other
quite well for P-OTDR data. This means there was an
impressive performance by FWT for the compression of
POTDR data. Figure 7(a) shows the sample signal of the
reconstruction results compared with the uncompressed
data of B-OTDA by the FWT method. Figure 7(b) shows
the error rate of B-OTDA data. These figures show that
the reconstructed and uncompressed signal trends are nearly
Fig. 3 Comparison of compression ratio of FWT, Huffman, and LZW.
Fig. 4 Computational cost analysis of FWT, Huffman, and LZW.
Optical Engineering 024108-4 February 2017 •Vol. 56(2)
Atubga et al.: Comparative study of lossy and lossless data compression in distributed optical fiber sensing systems
Downloaded From: http://opticalengineering.spiedigitallibrary.org/ on 03/07/2017 Terms of Use: http://spiedigitallibrary.org/ss/termsofuse.aspx
Fig. 6 Error analysis for P-OTDR data. (a) Uncompressed and reconstructed signals of P-OTDR data.
(b) Error rate signal: uncompressed and reconstructed signals of P-OTDR data.
Fig. 7 Error analysis for B-OTDA data. (a) Uncompressed and reconstructed signals of B-OTDA data.
(b) Error rate signal: uncompressed and reconstructed signals of B-OTDA data.
Fig. 5 Error analysis for Φ-OTDR data. (a) Uncompressed and reconstructed signals of Φ-OTDR data.
(b) Error rate signal: uncompressed and reconstructed signals of Φ-OTDR data.
Optical Engineering 024108-5 February 2017 •Vol. 56(2)
Atubga et al.: Comparative study of lossy and lossless data compression in distributed optical fiber sensing systems
Downloaded From: http://opticalengineering.spiedigitallibrary.org/ on 03/07/2017 Terms of Use: http://spiedigitallibrary.org/ss/termsofuse.aspx
the same for B-OTDA data and the error rate is also very
small. Based on this, it can be concluded that the perfor-
mance by FWT for the compression of B-OTDA data on
the part of generating error rate was also impressive and effi-
cient. Conclusively, it is apparent from Figs. 5–7that
P-OTDR has the best error rate followed by B-OTDA,
and Φ-OTDR giving the worst error rate due to its less deter-
mined data structure as compared to the other two groups of
DOFS data. These differences revealed the compression
results and the reconstructed performances depend greatly
on the data structure; P-OTDR data have the most stable
temporal–spatial structure; B-OTDA has less stable data
structure and with stochastic noises in it; Φ-OTDR with non-
linear demodulation of the phase change has the chaotic data
structure thus the reconstruction error is the largest.
However, this error level can meet the application require-
ments for the abnormal event detection and location. In con-
clusion for the experimental study, we can see FWT as a
lossy compression method is a good choice and an efficient
way for compressing the typical DOFS data at real time in
many practical applications.
5 Conclusions
After dealing with the three sets of DOFS data to mitigate the
problems of large volumes of data generation, large storage
space and high speed for data transmission, two lossless
methods, Huffman and LZW, and a lossy data compression
algorithm of FWT were comparatively studied and tested
based on three distinctive DOFS sensing data. Although it
revealed that there were some slight errors in the recon-
structed signals with respect to the DOFS data, FWT proved
a better choice on the compression ratio, and time as
observed among the three methods in achieving the stated
problems above. However, the differences in the structure
of the DOFS sets of data have some amount of influence
on both the compression ratio and the computational cost.
We can, therefore, conclude that FWT is cost effective, con-
venient, and more suitable for real-time compression of the
DOFS sets of data per the results exhibited. For future works,
it can withstand the test of time and we suggest FWT could
be applied in embedded systems for real-time application.
Acknowledgments
First, the authors appreciatively acknowledge previous sup-
port provided by the National High China (863 Program,
Grant No. 2007AA01Z245). Second, we are grateful for
the aid provided for this research by the Major Instrument
Special (Grant No. 41527805), Major Program (Grant
No. 61290312) and Youth Foundation (Grant No. 61301275)
of the National Science Foundation of China (NSFC).
Gratitude also goes for the fund of State Grid Corporation
of China: research on distributed multiparameter sensing
and measurement control technology for electric power optical
fiber communication networks (Grant No. 5455HT160014)
and the Fundamental Research Funds for the Central
Universities (Grant No. ZYGX2011J010). Finally, we are
indeed glad for being supported by the Program for
Changjiang Scholars and Innovative Research Team in
University (Nos. PCSIRT, IRT1218), and the 111 Project
(No. B14039).
References
1. J. C. Juarez et al., “Distributed fiber-optic intrusion sensor system,”
J. Lightwave Technol. 23, 2081–2087 (2005).
2. Z. Zhang and X. Bao, “Distributed optical fiber vibration sensor based
on spectrum analysis of polarization-OTDR system,”Opt. Express 16,
10240–10247 (2008).
3. X. Zhang et al., “Development of fully distributed fiber sensors based on
Brillouin scattering,”Photonic Sens. 1(1), 54–61 (2011).
4. J. Xu, H. Wu, and S. Xiao, “Distributed intrusion monitoring system
with fiber link backup and on-line fault diagnosis functions,”
Photonic Sens. 4(4), 354–358 (2014).
5. S. Mittal and J. S. Vetter, “A survey of architectural approaches for data
compression in cache and main memory systems,”IEEE Trans. Parallel
Distrib. Syst. 27, 1524–1536 (2016).
6. M. K. Tank, Implementation of Lempel-ZIV Algorithm for Lossless
Compression using VHDL, pp. 275–278, Springer India, New Delhi
(2011).
7. O. A. Mahdi, M. A. Mohammed, and A. J. Mohamed, “Implementing a
novel approach an convert audio compression to text coding via hybrid
technique,”Int. J. Comput. Sci. Issues 9(6), 53–59 (2012).
8. J. H. Pujar and L. M. Kadlaskar, “A new lossless method of image
compression and decompression using Huffman coding techniques,”
J. Theor. Appl. Inf. Technol.,15,18–23 (2010).
9. S. A. Al-Dubaee and N. Ahmad, “New strategy of lossy text compres-
sion,”in First Int. Conf. on Integrated Intelligent Computing (ICIIC
‘10), pp. 22–26 (2010).
10. V. Palaniappan and S. Latifi, Lossy Text Compression Techniques,
pp. 205–210, Springer, London (2007).
11. J. Ziv and A. Lempel, “A universal algorithm for sequential data
compression,”IEEE Trans. Inf. Theory 23, 337–343 (1977).
12. M. Misiti et al., “Wavelet Toolbox™User’s Guide © COPYRIGHT
2016,”MathWorks, Inc., Massachusetts, United States, http://www.
mathworks.com/help/pdf_doc/wavelet/wavelet_ug.pdf (15–07–2016).
David Atubga received his BA degree from the Department of
Information and Communication Studies at the University of Ghana
in 2012 and his master’s degree in communication and information
engineering from the University of Electronic Science and
Technology of China 2016. He is currently pursuing his PhD in infor-
mation and communication engineering at the University of Electronic
Science and Technology of China. His research interest is on distrib-
uted optical fiber sensing technology. He joined as a new member of
OSA on October 15, 2016.
Huijuan Wu received her BEng and PhD degrees in opto-electronic
engineering from Chongqing University, Chongqing, China, in 2003
and 2009, respectively. During 2007 to 2008, she studied at the
Department of Engineering Science in Oxford University as a joint
training doctoral student. Since 2013, she has been an associate pro-
fessor at the University of Electronic Science and Technology of
China. Her research interests include sensing signal extraction and
processing and optical fiber sensing systems. She is a member of
IEEE and OSA.
Lidong Lu received his BE, ME, and PhD degrees in optical engineer-
ing from Xi’an Technological University, Chinese Academy of
Science, and Nanjing University in 2005, 2009, and 2012 respec-
tively. He joined the Global Energy Interconnection Research
Institute of the State Grid Corporation of China in 2012 and has
been working on optical fiber sensing technology and its application
in power industry. He is currently a senior engineer of a research
group developing optical fiber communication networks and
advanced sensor technologies.
Xiaoyan Sun received her master’s degree from Southeast
University, Nanjing, China, in 2010. Currently, she is a research
and development engineer in the Global Energy Interconnection
Research Institute of the State Grid Corporation of China. Her
research interests include communication system and image
processing.
Optical Engineering 024108-6 February 2017 •Vol. 56(2)
Atubga et al.: Comparative study of lossy and lossless data compression in distributed optical fiber sensing systems
Downloaded From: http://opticalengineering.spiedigitallibrary.org/ on 03/07/2017 Terms of Use: http://spiedigitallibrary.org/ss/termsofuse.aspx
