1 Of 26

Thesis Title
Impact of Signal Length in Cross-
Correlation Based Underwater Network
Size Estimation

Presented by Supervised by
Samir Ahmed Shah Ariful Hoque Chowdhury
Roll No: 104018 Assistant Professor
Dept. of ETE, RUET.

September 17, 2019

 2 of 26

Contents
• Introduction
• Importance of node estimation
• Importance of signal length
• Underwater environment
• Literature review
• Impact of signal length in node estimation
• Corresponding works
• Comparison
• Future work
• Conclusion
September 17, 2019
 3 of 26

Introduction
• Node - communication endpoint, terminal equipment.
• Sensor- receiving node, capable of performing some
processing, gathering sensory information and communicating
with other connected nodes.
• Cross-correlation- a measure of similarity between
two waveforms
• Underwater wireless acoustic sensor network (UWASN)
• Signal length – Energy related term
• TS case – triangular sensors case, sensors placed in triangular
shape

September 17, 2019

 4 of 26

Importance of node estimation

• To ensure proper network operation
• Successful data collection
• Network maintenance
• To maintain communication quality
• Background noise calculation

September 17, 2019

 5 0f 26

Importance of Signal Length

• Signal length possesses a very important role in size
estimation of underwater wireless sensor network (UWSN)
• The greater the signal length the greater energy is required
for estimation
• Ideally the signal length is infinity (we consider 106
samples)
• It plays a great role in estimating number of nodes
• Accurate node estimation is being observed and discussed
in this thesis
September 17, 2019
 6 of 26

Underwater environment
• Long propagation delay
• High path loss
• Strong background noise
• Non-negligible capture effect
• Multipath signal propagation

September 17, 2019

 7 0f 26

Cross-correlation based node

estimation using two sensors [1]
• Basic theory: cross-correlation of two Gaussian signals
results a delta.
• Estimation parameter: ratio of standard deviation to
the mean, R of the cross-correlation function (CCF).
• Low protocol complexity
• Delay insensitive
• Not affected by capture effect
• Less time required
• Applicable to any environment network

September 17, 2019

 8 of 26

Cross-correlation based node

estimation using two sensors [1]
Distribution of nodes and sensors
D
Sensors

z-axis
Nodes

0 D
D
x-axis
y-axis
0 0 Distance between sensors, dDBS

Figure. Distribution of underwater network nodes and sensors.

September 17, 2019
 9 of 26

Cross-correlation based node

estimation using two sensors [1]
100
Coefficient value of CCF

80 Bins, b

60

40

20

0
-1.0 -0.75 -0.5 -0.25 0 0.25 0.5 0.75 1.0
Distance (m)

Figure. Bins, b in the cross-correlation process.

September 17, 2019
 10 0f 26

Cross-correlation based node

estimation using two sensors [1]
• Estimation parameter, R :
1 1
𝜎 𝑁× × 1− 𝑏−1
𝑏 𝑏
𝑅= = 1 =
𝜇 𝑁× 𝑁
𝑏
𝑏−1
so, 𝑁 = 2
𝑅
where b is the number of bins, which is twice the number of samples
between the sensors (NSBS), m minus one and can be expressed as:
2×𝑑𝐷𝐵𝑆 ×𝑆𝑅
𝑏= ̶1
𝑆𝑝

September 17, 2019

 11 of 26

Cross-correlation based node

estimation using two sensors [1]
8

7 Theoretical
6 Simulated

5
R of CCF

0
0 20 40 60 80 100
Number of nodes, N

Figure. R versus N for b = 19 with dDBS = 0.5m and SR = 30kSa/s.

September 17, 2019
 12 of 26

Cross-correlation based node estimation

using three sensors (SL) case[2]
Distribution of nodes and sensors Sensors
D D N1 Node
Sensors
d11 d13
d12
z-axis Nodes z-axis
H1 dDBS12 H2 dDBS23 H3

0 D 0 D
D D
x-axis x-axis
y-axis y-axis
0 0 0 0

Figure. Distribution of underwater network nodes for sensors in line (SL) case
with N transmitting nodes (left) and only one node N1 (right).
September 17, 2019
 13 0f 26

Cross-correlation based node estimation

using three sensors (TS) case [3]
Distribution of nodes and sensors Sensors
D Sensors D Node
d13 H3
N1 dDBS31
z-axis Nodes z-axis d11 dDBS23
d12

H1 dDBS12 H2
0 D 0 D
D D
x-axis x-axis
y-axis y-axis
0 0 0 0

Figure. Distribution of underwater network nodes for triangular sensors (TS)

case with N transmitting nodes (left) and only one node N1 (right).
September 17, 2019
 14 of 26

Block diagram for TS case

C31(τ) μ31
Cross-correlation Mean Ratio
Nodes σ31 / μ31
N1 Sr c1(t) Standard deviation
σ31
Cross-correlation R31
Sensors C12(τ)
dDBS31 d11 d13
N2 H 1
Standard ddeviation
12
Mean
σ12
Sr c3(t) 3CCF
N3 H3 μ12 Ratio R12 Raverage
σ12 / μ12 Average
dDBS12
⁞ H2 dDBS23
⁞
⁞ σ23
Sr c2(t)
NN Standard deviation
C23(τ)
Cross-correlation Ratio R23
Gaussian σ23 / μ23
signals Composite μ23
Gaussian Mean
September 17, 2019 signals
 15 of 26

Cross-correlation based node estimation

using three sensors (TS) case [3]
3CCF can be expressed as:
For TS case, estimation parameter, 𝑅average
3CCF 𝑅12 + 𝑅23 + 𝑅31
𝑅average =
3
𝑏12 − 1 𝑏23 − 1 𝑏31 − 1
+ +
𝑁 𝑁 𝑁
=
3
For efficient estimation,b12=b23=b31=b,
𝑏−1
so, 𝑁𝑒𝑠𝑡 = 3CCF )2
(𝑅average

September 17, 2019

 16 of 26

Cross-correlation based node estimation

using three sensors[]
8 8

7 Theoretical 7 Theoretical
6 Simulated 6 Simulated
Raverage of CCFs

Raverage of CCFs
5 5

4 4
2CCF

3CCF
3 3

2 2

1 1

0 0
0 20 40 60 80 100 0 20 40 60 80 100
Number of nodes, N Number of nodes, N

2CCF 3CCF
Figure: Raverage versus N for SL case (left) and Raverage versus N for TS case
(right) with b = 19. [dDBS = 0.5m and SR = 30kSa/s]
September 17, 2019
 17 of 26
Effect of signal length in estimation
process, (SL case) [4]
35

30
Nest vs Ns plot for SL case
25 Nest vs Ns plot for Two sensor case
Number of nodes, N

Exact 32 Nodes
20

15

10

0
2 3 4 5 6
10 10 10 10 10
Signal length in Number of samples (N s)

Figure. Estimated N versus Ns plot (x-log, y-normal scale) for two & three
sensor (SL case) method with fixed value of b = 119 using exact 32 nodes
September 17, 2019
 18 of 26

Impact of signal length in cross-

correlation based estimation (TS case)

Some initial assumption

• The channel is considered as ideal
• Receivers are assumed to be ideal
• No multipath effect is considered
• No Doppler shift is considered
• Network dimension — 3D spherical
• Transmitted Signal — White Gaussian
• Signal power — Equal received powers from all nodes

September 17, 2019

 19 of 26

Nominal simulation parameters

• Dimension of the sphere, D = 2000m

• Speed of acoustic wave propagation, SP = 1500m/s
• Signal length, Ns = 106 samples (varied for comparison)
• Absorption coefficient, a = 1
• Dispersion factor, k = 1.5
• Distance between equidistance sensors = 1m (can be varied)
Estimation parameter
𝑏−1
𝑁𝑒𝑠𝑡 = 3CCF )2
(𝑅average

September 17, 2019

 20 of 26

Corresponding Result
2
10
number of nodes, N

1
10

N vs Ns plot for exact 32 nodes

Estimated N vs Ns plot for b=119
Estimated N vs Ns plot for b=39
0
Estimated N vs Ns plot for b=19
10
2 3 4 5 6
10 10 10 10 10
signal Length, Ns

Figure. Estimated N versus Ns plot (x-log, y-log scale) three sensor (TS case)
method with value of b = 19, b=39 and b=119 using exact 32 nodes
September 17, 2019
 21 of 26

Corresponding Result
2
10
number of nodes, N

1
10

Estimated N vs Ns plot for 64 nodes,b=39

N vs Ns plot for exact 64 nodes
Estimated N vs Ns plot for 32 nodes,b=39
0
N vs Ns plot for exact 32 nodes
10
2 3 4 5 6
10 10 10 10 10
signalLength, Ns

Figure. Estimated N versus Ns plot (x-log, y-log scale) three sensor (TS case)
method with fixed value of b = 39 using 32 nodes and 64 nodes.

September 17, 2019

 22 of 26

Comparison between TS case and SL case

2
10
number of nodes, N

1
10

N vs Ns plot for exact 64 nodes

Estimated N vs Ns plot for TS case,b=19
0
Estimated N vs Ns plot for SL case,b=19
10
2 3 4 5 6
10 10 10 10 10
signalLength, Ns

Figure. Estimated N versus Ns plot (x-log, y-log scale) three sensor (TS and SL
case) method with fixed value of b = 19 using exact 64 nodes

September 17, 2019

 23 of 26

Future work
• Estimation with unequal distances between the sensors
• Estimation with non-uniform distribution of nodes
• Estimation with different shape of network
• Estimation with random placement of the sensors
• Estimation with variable propagation delay
• Use of Non-Gaussian signals for estimation
• Estimation with M number of sensors
• This thesis consider only ERP case, so ETP, RTRP cases
requires further work

September 17, 2019

 24 of 26

Conclusion
• Using three sensors method, TS case, we can estimate the
number of nodes easily with reduced signal length for which
the required energy will be less than the SL case, three sensors
method
• In this thesis we use smaller signal length than two sensor
technique and provide better performance in estimation
process
• TS case, three sensors techniques provide better performance
than any other techniques in small area

September 17, 2019

 25 of 26

References
[1] M. S. Anower, M. R. Frater, and M. J. Ryan, “Estimation by cross-correlation of the number of
nodes in underwater networks,” In Proceedings of Australasian Telecommunication Networks
and Applications Conference (ATNAC), 10–12 November, 2009, pp. 1–6. doi:
10.1109/ATNAC.2009.5464716.
[2] S. A. H. Chowdhury, M. S. Anower, and J. E. Giti (2014), “A signal processing approach of
underwater network node estimation,” In Proc. International Conference on Electrical
Engineering and Information Communication Technology (ICEEICT) 2014, Dhaka, 10−12
April, 2014.
[3] S. A. H. Chowdhury, M. S. Anower, and J. E. Giti (2014), “Effect of sensor number and location
in cross-correlation based node estimation technique for underwater communications network,” in
Proceedings of 3rd International Conference on Informatics, Electronics & Vision (ICIEV 2014),
23–24 May, 2014, Dhaka, Bangladesh
[4] M.A. Hossen, S.A.H. Chowdhury, M. S. Anower (2015), “Effect of signal length in cross-
correlation based underwater network size estimation” Paper id 528_ICEEICT 2015

September 17, 2019

 26 of 26

Thank you

September 17, 2019