Sparse Signal Processing for Undersea Acoustic Links

CHAPTER 1

INTRODUCTION

Underwater acoustic channels are generally recognized as one of the most

difficult and challenging communication media in use today. This chapter
addresses the various factors affecting the propagation of sound in water.
Underwater acoustic communication is characterised by certain
undesirable effects like multipath, ambient noise, attenuation, spreading,
etc. Evaluation of new communication schemes and devices under
realistic conditions being very expensive and time-consuming, simulators
for modeling the acoustic underwater communication channel accurately
like Bellhop and Rayleigh models have been studied.

1.1 Background

Undersea links are complex due to the effects of multipath,

scattering, absorption, ambient noise, etc. Even though complex, a reliable
communication is achievable using acoustic waves. During the last three
decades, the realm of undersea communication scenario has witnessed
remarkable progress and there was an exponential migration from systems
with low bit rate to high bit rate and high power to low power consumption,
which facilitated the development and deployment of undersea acoustic
networks. Current research in this area focuses on judicially formulating
an integrated solution for meeting the emerging demands for applications
of undersea networks in environmental data collection, pollution
monitoring, offshore surveillance, coastal surveillance, etc.

1
Chapter 1 Introduction

1.2 Undersea Acoustic Communication

Acoustic signals are widely used in ocean exploration,

oceanographic data collection, underwater communication, etc. Optical
waves and electromagnetic waves do not propagate over long distances in
underwater environment. The underwater channel is too complex due to
various phenomena like multipath propagation, spreading, attenuation, etc.
The available bandwidth of an underwater acoustic channel depends on the
transmission loss, which increases with range and frequency. Moreover,
the ambient noise also affects the performance of underwater acoustic
communication systems. The sound speed is typically 1500m/s; but varies
with depth, climatic conditions, etc. Sea trials being very expensive and
time consuming, various toolboxes available for simulating the underwater
acoustic communication have been used in this thesis, for investigating the
behavior of the underwater channels.

1.3 Underwater Acoustic Sensor Networks

Underwater sensor network consists of a number of sensor nodes

deployed to perform collaborative monitoring tasks over a given area for
applications like oceanographic data collection, pollution monitoring,
offshore exploration, mapping of the ocean floor for detection of objects,
offshore oil/gas field monitoring, etc. A typical underwater acoustic sensor
network is shown in Figure 1.1. Sensor nodes are mounted on the seabed
by using anchors for observations of environmental conditions such as
temperature, pollution data and other parameters of strategic and civilian
applications.

2
 Sparse Signal Processing for Undersea Acoustic Links

Fig. 1.1 A typical Underwater Acoustic Network {Courtesy: [1]}

1.3.1 Static Topology

1.3.1.1 Two dimensional

In two-dimensional underwater networks, a group of sensor nodes

are anchored to the bottom of the ocean with deep ocean anchors for
applications like environmental monitoring. The topology of a two-
dimensional underwater network is shown in Figure 1.2. Underwater
sensor nodes are interconnected to one or more cluster heads by means of
wireless acoustic links. Cluster heads are network devices in charge of
relaying data from the ocean bottom network to a surface station. To
achieve this objective, cluster heads are equipped with two acoustic
transceivers, namely a vertical and a horizontal transceiver. The horizontal
transceiver is used by the cluster head to communicate with the sensor
nodes in order to send commands and configuration data to the sensor
nodes and collect monitored data. The vertical link is used by the cluster
heads to relay data to a surface station. In deep water applications, vertical

3
Chapter 1 Introduction

transceivers must be long range transceivers as the ocean can be as deep as

10km. The surface station is equipped with an acoustic transceiver that is
able to handle multiple parallel communications with the cluster heads. It
is also endowed with a long range RF and/or satellite transmitter to
communicate with the onshore sink and/or to a surface sink.

Fig. 1.2 Two-dimensional underwater network {Courtesy: [2]}

1.3.1.2 Three-dimensional

Three-dimensional underwater networks are used to detect and

observe phenomena that cannot be adequately observed by means of ocean
bottom sensor nodes, i.e., to perform cooperative sampling of the 3D ocean
environment. In three-dimensional underwater networks, sensor nodes are
anchored to the bottom of the ocean and they float at different depths in
order to observe a given phenomenon.

4
 Sparse Signal Processing for Undersea Acoustic Links

In three-dimensional topology, as depicted in Figure 1.3, each node

is equipped with a floating buoy that can be inflated by a pump. The buoy
pushes the node towards the ocean surface. The depth of the node can then
be regulated by adjusting the length of the wire that connects the node to
the anchor, by means of an electronically controlled engine that resides on
the node. A challenge to be addressed in such a topology is the effect of
ocean currents on the described mechanism to regulate the depth of the
sensor nodes.

Fig. 1.3 Three-dimensional underwater network {Courtesy: [2]}

1.3.2 Hybrid Topology - Networks with mobile sensor nodes

Autonomous Underwater Vehicles (AUVs) can function without

tethers, cables, or remote control, and therefore they have a multitude of
applications in oceanography, environmental monitoring, and underwater
resource study. Hence, they can be used to enhance the capabilities of
underwater sensor networks in many ways. The integration and

5
Chapter 1 Introduction

enhancement of fixed sensor networks with AUVs, Remotely Operated

Vehicles (ROVs) or any other sea gliders, is a promising area of research in
undersea networks. This is called a Hybrid topology where an underwater
acoustic sensor network consists of lots of static nodes together with a
number of mobile nodes, as shown in Figure 1.4. In hybrid topology,
mobile nodes play a key role for additional support in accomplishing the
task, perhaps for data harvesting or enhancing the network capacity.
Mobile nodes could be considered as super nodes which has more energy
and can move independently, it could be a router between fixed nodes, a
manager for network reconfiguration, or even a normal node for data
sensing.

Fig. 1.4 Hybrid Underwater Network {Courtesy: [3]}

1.4 Factors Affecting Propagation of Sound in Water

Ocean forms a complex medium for the propagation of sound.

Underwater acoustic communications include undesirable effects like
ambient noise, attenuation, spreading, etc. Transmission loss (TL) can be

6
 Sparse Signal Processing for Undersea Acoustic Links

defined as ten times the log (base 10) of the ratio of the reference intensity
(Iref) measured at a point 1m from the source, to the intensity (I), measured
at a distant point, and is expressed in units of decibels (dB):

= 10 log . (1.1)

Transmission loss may be considered to be the sum of loss due to

spreading and attenuation. Spreading effect is the regular weakening of
sound as it spreads outwards from the source. Attenuation loss includes the
effects of absorption, scattering and leakage out of sound channels.

Fig. 1.5 Geometry for (a) Spherical Spreading (b) Cylindrical Spreading {Courtesy: [4,5]}

1.4.1 Spreading

Spreading can be cylindrical, in which, the intensity decreases

proportional to the range or spherical, in which the intensity decreases as
7
Chapter 1 Introduction

the square of the range. Figure 1.5 shows the geometry for spherical
spreading and cylindrical spreading. Cylindrical spreading occur when the
medium has plane parallel upper and lower bounds and is given by

= 10 , (1.2)

where r is the range from the transmitter. For spherical spreading, the
transmission loss varies as,

= 20 log . (1.3)

1.4.2 Attenuation

Absorption involves conversion of acoustic energy into heat. The

Thorpe attenuation model for absorption is given by

= 0.1 !
+ 40 $ %%!
+ 0.000275 (
, (1.4)

where attenuation is in decibels per kiloyard and frequency f is in

kHz. Thus, for every kiloyard travelled, the sound intensity gets
diminished due to the absorption by an amount dB. It is known that
absorption of sound in sea water is high compared to that in distilled water
due to the dissolved minerals in the sea.

1.4.3 Sound Speed Profile

The sound speed profile (SSP) represents a plot of the speed of

sound with depth from the surface to bottom. The Sound Speed Profile
may be constructed from any number of actual data points and then
subjecting it to a fitting algorithm to produce a smoother graph. Given the
Conductivity, Temperature, Depth (CTD) data, the Sound Speed Profile
can be computed using the Leroy’s formula given by
* = 1492.9 + 3 − 10 – 6 × 1012 − 10 (
− 4 × 101( − 18 (

+1.2 4 − 35 − 101( − 18 4 − 35 +
5
6
, (1.5)

8
 Sparse Signal Processing for Undersea Acoustic Links

where C is the sound velocity in m/s, T is the temperature in degree

centigrade, S is salinity in parts per thousand and z is the depth in meters.

The sound speed varies non-linearly with depth, due to the variation
of temperature, salinity and pressure from surface to bottom. At lower
depths, temperature determines the sound speed, but as the water depth
increases, pressure becomes the main determining factor in the speed of
sound propagation at larger depths.

When the measurements are not available, standard models like

Munk profile, can be used for many ocean regions. Walter H. Munk
introduced an equation for sound velocity as function of depth, with the
exponential increase dominating in the upper waters and the linear term
dominating in deep waters. Figure 1.6 shows a Munk profile and the Munk
equation is given as

7 8 = 1492.091.0 + : ; 1< + = − 1 >, (1.6)

Fig. 1.6 Sound speed profile using Munk equation {Courtesy: [6]}

9
Chapter 1 Introduction

where

==
( 51 2%%
2%%
, (1.7)

: = 7.4 x 10-3 is the Perturbation coefficient, the fractional adiabatic

velocity increase over a scale depth and z is the water depth.

A typical deep sea profile may be divided into several layers such as
the surface layer, seasonal thermocline, main thermocline as well as the
deep isothermal layer as shown in Figure 1.7. In the surface layer, the
velocity of sound is susceptible to daily and local changes of heating,
cooling and wind action. In the seasonal thermocline, temperature, and
hence velocity decreases with depth. The main thermocline, which lies
below the seasonal thermocline, is affected only slightly by seasonal
changes. Close to the sea bottom, there is the deep isothermal layer with
positive velocity gradient due to the effect of pressure on sound velocity.
As a consequence of the characteristic velocity profile of the deep sea, a
sound channel also called the SOFAR (sound fixing and ranging) channel
or deep sound channel (DSC) occurs in the deep sea. This sound channel
offers low transmission loss, and hence, very long ranges are achievable.

1.4.4 Ambient Noise

Noise in underwater communication scenario consists of ambient

noise which is always present in the background of the sea, and site-specific
noise which exists only in certain regions. The sources of ambient noise
include tides, seismic disturbances, oceanic turbulences, thermal noise,
biological sources, etc. At low frequencies (0.1 - 100Hz), earthquakes,
underwater volcanic eruptions, turbulence in the ocean and atmosphere, etc.
are the main noise sources. In the frequency band 50 - 300Hz, underwater
noise is mainly due to remote shipping traffic, while in the frequency band

10
 Sparse Signal Processing for Undersea Acoustic Links

Fig. 1.7 Layered structure of ocean {Courtesy: [4,5]}

0.5 - 50kHz, underwater noise is directly associated with the state of the
ocean surface and the winds in the area of interest. At frequencies greater
than 100kHz, thermal noise is dominant. Biological noise produced by
marine animals for communication with each other, locating the prey,
frightening enemies and so on, is seasonal and spatial in nature.

1.4.5 Multipath

Multipath effects occur due to the reflection of sound by the sea

surface and sea bottom as well as due to the refraction of sound in the
water. As a result, the receiver gets a bewildering mix of signals from the
transmitter in direct path as well as multipath as shown in Figure 1.8.

11
Chapter 1 Introduction

Fig. 1.8 Shallow water multipath propagation: direct as well as reflected paths {Courtesy:
[7]}

A1

A2
Delay τ1

Power Multipath
Delay τ2
Input Disrtibution ∑ Channel
.

AN
Output
.

Delay τN

Fig. 1.9 Multipath Channel Model {Courtesy: [8]}

The channel impulse response of a multipath channel can be

modeled as in Figure 1.9 and expressed as

? = ∑E
BF AB C D − DB , (1.8)
where N is the number of multipaths, DB is the n path delay and AB is the
th

amplitude associated with nth path.

The different types of propagation paths that exist between a source

and a receiver in the deep sea are as shown in Figure 1.10. Sound
propagates in the sea by way of a variety of paths depending on the sound
speed structure in the water column and the source–receiver geometry.
These paths are direct path (A), surface duct (B), bottom bounce (C),
convergence zone (D), deep sound channel (E) and reliable acoustic path
(F), where x is the source and o is the receiver.

12
 Sparse Signal Processing for Undersea Acoustic Links

Fig. 1.10 Various sound propagation paths in deep sea {Courtesy: [5]}

1.5 Compressive Sensing

Nyquist sampling theorem forms the basis of nearly all the signal
acquisition, transmission and reconstruction techniques widely used in
almost all the systems. Sampling a signal at the Nyquist rate and
transmitting it at the same rate in situations where the bandwidth is a scarce
resource may adversely affect the system throughput and efficiency. This
warrants the need and requirement for devising systems and techniques for
transmitting a signal at a rate lower than the Nyquist rate. If the signal is
transmitted at par with the Nyquist criteria, it might result in too many
samples. This necessitates the need for compressing the signal before it is
stored or transmitted. The signals so compressed can be reconstructed at a
later stage following certain optimization techniques.

The available bandwidth of an underwater acoustic channel depends

on the transmission loss, which increases with range and frequency. This
implies that the transmission loss limits the available bandwidth for
underwater acoustic communication. The bandwidth available for
underwater acoustic communication is of the order of 1kHz, which is much
lower than that available for RF communication links. If the information
13
Chapter 1 Introduction

could be stored or transmitted by fewer numbers of samples than the

Nyquist rate, it would have been advantageous in underwater acoustic
communication scenario.

A new method called compressive sensing is used to represent and

reconstruct certain classes of signals at a rate below the Nyquist rate.
Compressive Sensing is a new paradigm that gained the attention of
researchers in signal processing, communication as well as mathematics. It
helps in reconstructing the signal from far less samples than that required
by the sampling theorem, which paves the way for saving memory and
catering low data rate requirements in communication applications.

1.6 Sound Propagation Models

This section discusses the Bellhop and Rayleigh models, which are
used widely for undersea acoustic modeling.

1.6.1 Bellhop Model

Bellhop is a highly efficient ray tracing toolbox for predicting the

acoustic pressure fields in ocean environments. It was developed by
Michael Porter from Heat, Light & Sound (HLS) Research, Inc. and is
available as free software. Bellhop model can generate a variety of useful
outputs including transmission loss, Eigen rays, Channel Impulse
Response, etc. for the environment specified by the user.

To utilize the ray tracing capabilities of the Bellhop model, a

precise description of the physical characteristics of the targeted
environment is important, which include the depth of the ocean, sound
speed profile in the location, information concerning the bottom contour
and roughness, information about the surface, etc. The environmental data
can be gathered from the real measurements or are available from known
data sets for the particular ocean environments. In general perspective, the
14
 Sparse Signal Processing for Undersea Acoustic Links

real measurement data are preferred so that the model will generate more
realistic results. The environmental file specified should include the
frequency of operation, depth-speed pairs, number of sources as well as
receivers and their depths. Bellhop model allows for range-dependence on
the top and bottom boundaries (altimetry and bathymetry), as well as the
sound speed profile. Additional input files allow the specifications of
directional sources and geo-acoustic properties of the bounding media as
well as the surface/ bottom reflection coefficients.

1.6.2 Rayleigh Model

The effects of multipath in underwater acoustic channel causes

constructive and destructive interferences and phase shifting of the signal.
The shallow water medium range channel, where the multipath fading is
predominantly due to reflections from the sea bed and the sea surface,
exhibits Rayleigh fading properties. Rayleigh fading, which is used for
modeling the effect of propagation in a wireless environment, assumes that
the magnitude of a signal that has passed through such a transmission
medium will fade according to certain well known (Rayleigh) distribution.

The Rayleigh probability density function (pdf) is defined as,

G = HJ K ;GL H (J K,
I 1I
(1.9)

and the Cumulative Distribution Function (CDF) is expressed as,

M G = 1 − ;GL H (J K,
1I
(1.10)

where σ is the scale parameter and x lies in the range [0, ∞).

15
Chapter 1 Introduction

1.7 Scope of Work Done

This thesis addresses one of the emerging topics in Underwater

Acoustics, viz. the various modulation and coding techniques for realising
reliable OFDM for Undersea Acoustic Links. This thesis proposes a matrix
padding method for sparse recovery, which is robust even in the presence
of noise. Orthogonal Frequency Division Multiplexing and coded
interleaved OFDM as well as coded interleaved STBC have been
investigated. The synchronization and channel estimation issues in OFDM
systems have also been investigated.

Chapter 2 is devoted to the review of the works reported in open

literature in the areas of compressive sensing, propagation of sound in sea,
channel modeling, various modulation schemes, various coding schemes,
channel estimation techniques, synchronization issues in OFDM
(Orthogonal Frequency Division Multiplexing), etc. Compressive sensing
methods have been reported and various reconstruction methods have also
been consolidated by various researchers. Many compressive sensing
solvers do exist, which include l1-magic, l1-Regularized Least Squares
(l1_ls), Convex Optimization (CVX), Your Algorithm for L1 Optimization
(YALL1), Orthogonal Matching Pursuit (OMP), Compressive Sampling
Matching Pursuit (CoSaMP), etc. Various research works relating to
modeling of underwater channel using Bellhop and Rayleigh models as
well as OFDM, diversity techniques and channel estimation techniques
were also covered in the literature survey.

Compressive sensing has been evolved as a very useful technique

for sparse reconstruction of signals that are sampled at sub-Nyquist rates.
Compressive sensing helps to reconstruct the signals from few linear
projections of the sparse signal. Chapter 3 presents the background of

16
 Sparse Signal Processing for Undersea Acoustic Links

compressive sensing and the various compressive sensing algorithms that

are reported in open literature. A method for Sparsification and Recovery
has also been proposed with a block diagram. Undersea acoustic links have
numerous applications in pollution monitoring, environmental data
collection, offshore and coastal surveillance, etc. Chapter 3 also provides a
brief description of the various modulation techniques in use for undersea
acoustic links. Orthogonal Frequency Division Multiplexing, which is a
multicarrier modulation technique, is widely employed in undersea acoustic
applications because of its numerous advantages. OFDM requires an
estimate of the channel parameters at the receiver for undoing the channel
effects. The sparsity of the channel impulse response of undersea links can
be used for estimating the channel characteristics. A brief overview of the
dictionary based sparse channel estimation is also included.

Compressive sensing recently gained immense attention due to the

commendable advantages the technique offers in recovering certain target
signals from a few random measurements. With the help of compressive
sensing, the salient information in a signal can be preserved in a relatively
small number of linear projections. A new technique which converts the
signal into sparse domain by applying Discrete Cosine Transform (DCT)
and then compressing it using a modified measurement matrix, followed by
an LMS based adaptation, has been proposed in chapter 4. Making use of
this proposed matrix padding technique, a computationally efficient sparse
signal reconstruction has been achieved. Ambient noise is always present in
the background of the undersea channel and the four basic sources that can
model the ambient noise in the ocean are noise effects due to turbulence,
shipping, wind and thermal processes. A study of variation of power
spectral density (PSD) of the overall ambient noise has also been
performed. Generation of channel impulse response using the Bellhop

17
Chapter 1 Introduction

model utilizing the environmental data like the sound speed profile of the
water column, the number of transceivers and their depths has also been
described. OFDM is a good choice in underwater communication, because
of its numerous advantages. Chapter 4 also introduces the system models
for normal OFDM as well as coded and interleaved OFDM, for undersea
acoustic links. Space-time block coding (STBC) offers higher link
reliability and this thesis combines STBC with coding and interleaving for
improving the bit-error-rate performances. Synchronization and channel
estimation are critical in practical OFDM communication scenario. A
system model of an OFDM Receiver with Symbol Time Offset (STO) and
Carrier Frequency Offset (CFO) estimation as well as channel parameter
estimation has also been described.

Orthogonal Frequency Division Multiplexing is widely employed

nowadays because of its advantages like resilience to intersymbol
interference, higher spectral efficiency, simpler channel equalization, etc.
Various modulation techniques like Quadrature Amplitude Modulation and
Phase Shift Keying are used in conjunction with OFDM. Coding
techniques like convolutional and BCH coding, as well as interleaving
techniques can be used along with OFDM for improving the error
correction capability of the receiver. Chapter 5 discusses OFDM as well as
some of the modulation and coding techniques which can effectively
improve underwater acoustic communication. Diversity techniques improve
reliability of data transfer by transmitting the same data on two or more
communication channels with different characteristics. The most
commonly used space-time block coding (STBC) technique, the Alamouti
STBC, has also been described. For undoing the channel effects and
demodulating the signal with fairly acceptable accuracies, the channel
parameters need to be estimated. Channel estimation is performed in

18
 Sparse Signal Processing for Undersea Acoustic Links

orthogonal frequency division multiplexing by inserting pilots into the

subcarriers of an OFDM symbol. Underwater channels exhibit sparse
channel impulse response which implies that very few of the channel taps
have a nonzero value and the channel can be estimated using the
compressive sensing techniques to achieve better performance. The
dictionary based sparse channel estimation technique has also been
analytically explained.

The performance of the matrix padding method is compared with

other existing compressive sensing algorithms like l1_ls, l1-magic, YALL1,
OMP, CoSaMP, etc. and the results of comparison in terms of signal-to-
noise ratio, correlation and mean squared error, have been investigated in
chapter 6. OFDM has been simulated using QAM (Quadrature Amplitude
Modulation) and PSK (Phase Shift Keying) based modulation techniques
for undersea acoustic links. The performances of various orders of QAM
based OFDM systems for undersea acoustic communications have been
studied and the bit-error-rates under various Signal to Noise Ratio
conditions have also been compared for Additive White Gaussian Noise
(AWGN) and Underwater Channels for 16-QAM based OFDM. The bit-
error-rate performances of normal as well as coded OFDM with and without
interleaving schemes have been simulated for various signal-to-noise ratio
levels for both convolutional and Bose, Chaudhuri and Hocquenghem
(BCH) codes. The performances of Alamouti STBC for a 2x1 and a 2x2
system were studied. It has been observed that the proposed STBC system
with coding and interleaving offers better performance compared to the
normal 2x1 and a 2x2 STBC systems. The performances of various STO as
well as CFO estimation methods have been compared and it has been
observed that STO estimation using difference method and CFO estimation
using pilot based method guarantee acceptable performances. The undersea

19
Chapter 1 Introduction

channel has been estimated using dictionary based sparse channel

estimation technique for various ranges and it has been found that by using
a minimum number of pilots, acceptable bit-error-rate (BER) performance
can be achieved. Chapter 7 brings out the salient highlights of the research
work undertaken for realizing an underwater acoustic communication
system and the general inferences gathered. This chapter also enlists the
scope and direction for future research in this area.

1.8 Summary

The importance of underwater acoustic communication as well as

various topologies for underwater acoustic sensor networks were covered.
The factors affecting the propagation of sound in water as well as modeling
of undersea channel using the Bellhop and Rayleigh models have also been
discussed. This chapter also gives an introduction to Compressive Sensing,
which helps in efficient use of the available scarce bandwidth of the
undersea channel. Sparsity, an inherent characteristic of many natural
signals, enables the signal to be stored in a fewer samples than that has
been dictated by the Nyquist rate and subsequently be recovered, by means
of compressive sensing. The scope of the work carried out in this thesis has
also been briefly outlined.

20