00544024

7/23/2019 00544024

http://slidepdf.com/reader/full/00544024 1/8

vances

in Si

for

Electronicwar

James

P.

Stephens

Wright

hboratory

ABSTRACT

Th e denia l of effective communications by enemy

forces during hostile military operations has been a

primary concern for military comm andcrs since the

inception of radio comm unica tions on the battlefield

before World War

11.

Since then, the electromagnetic

environmen t has been in

a

cons tant state of evolution

toward more sophisticated jam-resistant and convert forms

of modulation. For example, exotic modulation tech niques

employing spread spectrum

SS)

signaling are routinely

used by our adversaries to provide their communication

links an advantage over US and Allied jammers. Mo re

recently, these same spread spe ctrum modulation

techniques a re being refined t o provide convert, low

probability-of-intercept LPI) features

to

the unintended

interceptor. The thrust of this paper focuses on

developm ents in the theory and algorithms for detection,

characterization, a nd exp loitatio n of advanced waveforms

using new mathematical signal processing tools introduced

within the past decade. Specifically, quadratic

time-frequency signal representations, wavelet transforms,

and cyclostationary signal processing ar e introduc ed. This

overview demonstrates the im portance of these advanced

techniques in a clear and concise ma nner . Applications

Author’s

Current

Addicss:

Wright

Laboratory Avionics

Directorate, Electronic

Warfare Division.

Based on a presentation at NAECON ‘96.

U S Government

work

not

protected

by US

Copyright.

and f utur e research activities are des cribe d in this

significant area that is gaining much a ttention in a variety

of tcchnical fields.

INTRODUCTION

In the early days of electronic warfare a n opera tor

would tune his radio across the band listening for threat

signals, likely in the fo rm of voice mo dula tion spok en in a

foreign language. Upon detection, h e would simply place

his noise jamme r at the same carrier fre quency with

as

much power

as

possible. Th e ope rator, using his ears and

brain, comprised the signal processor. Today, it is not

possible for manua l ope rators to identity threat signals

efficiently and effectively because of t he pr olifer ation of

complex waveforms used for voice, data , rada r, navigation,

and image transmission.

the tasks of detectio n classification, ide ntifica tion, and

exploitation in a complex envir onm ent of high noise

intcrfcrcnce a nd multiple signals. Some waveforms a re

intentionally designed

to

makc the de tection process

ncarly impossible. Such signals are referred to as LPI or

LPD waveforms, meaning they offer

a

low probability

interc ept o r low probability-of-detection. Af ter the signals

are detected, the task of classification requires sorting into

groups having similar characteristics. Parameters such as

carrier frequency, modulation type, data rate , and time o r

angle-of-arrival ar e just a few of the fundam ental features

that distinguish one signal from ano ther . Da ta bases ar e

used to configure these signal para me ters into arrays that

are compared to existing knowledge o r to estab lish new

Mo dern electronic intercept systems must perform

IEEE AES Systems Magazine Noverriber 996

31

7/23/2019 00544024


Fig.

IA.

Stationary

Fig. 1B. Non-Stationary

Fig. 1. Stationary and Non-Stationary Signal

records. T he sorting a nd cataloging of signals leads to the

process of identification, a critical step when effective

electronic countermeasures are

to

be initiated an d the

jamming of ones’ own resou rces is to be avoided. Finally,

the pro blem of effectively and efficiently jam min g the

signal is met. T he electronic attacke r must select a strategy

that requires the least amo unt of resources, but yet offer

the most effectiveness. Gettin g feedback regarding the

success of your jammin g is extremely helpful, but not

always possible. A tech nique called “look-through’’ allows

the jamm er to observe his effectiveness on th e targ et. This

may be accomplished by stoppin g to listen periodically or

listening thro ugh t he jam min g by using special filters.

Overall, electronic attack can be thought of as a game and

many of the strategies of gam e theory can be applied to

the overall problem. Each of these initial processes:

dete ction, classification, identification, and exploitation

require advanced signal processing techniques. Many of

the theo retical signal detectio n concepts in use today were

advanced in the

1920’s

and

1930’s

during the early

development of radar technology. But, in the last 20 years

the deve lopment of exotic modulation schemes

implemented through advances in digital signal processing

gave rise to signals supp orting higher infor mation rates,

greate r channel capacity, a nd improved noise immunity.

Thes e same techniques a re now being used to exploit thesc

modern waveforms.

TECHNIQUES

A good understanding of the a dvances in signal

processing technology requires a discussion of the

funda men tals of F ourier Analysis. From this basic tool,

mo re complex processes have evolved. Som e

of

the more

significant techniques currently being researched for

Electronic W arfare app lications are described below.

The F ourier Transform

examination of time and frequency separately through the

use of the Fourier Transform. The F ourier Transform and

its digital im plementation, the FFT Fast Fourier

Transform ), allow the decom position of a signal into

individual frequency compon ents and their am plitudes.

How ever, the major drawback of these

tools

is that time

and fre quency information cannot be com bined to tell how

frequency con tent is changing in time. Fo r examp le, if you

look at the light coming from the s un above the earth’s

atmosphere, it is steady sta te and its frequency content is

the sam e over millions of years. If we loo k at th c sun’s light

at the surface of the Ear th, the frequency conte nt changes

dramatically during sunrise and sunset. We refer to the

first case as a stationary event and the second a

nonstationary event.

A

stationary signal is independ ent of

time, wh ereas, a non-stationa ry signal changes over time.

Figure 1 llustrates stationary and non-stationary signals

through the use of the Short-Time Fourier Transform

STF T). The STF T is also known by the na mes

“Windowed Fourier t ransform” an d “spectrogram.”

Almost

all

digital signal processing systems use the STFT

since the e nvironmen t is typically sampled ov er some

time-interval, processed i.e., FIT), and the n outpu t to its

intend ed fun ction. This process is continually repe ated.

But what

is

important

to

realize is that

a

only a por tion of

the RF environme nt is analyzed during a small time

segment.

signal in both time and frequency simultaneously. Th e

basic idea is to Four ier analyze a small part of the signal

aroun d the time of interest to determine the frequencies at

that tim e. Since the time interval is short com pared

to

the

whole signal, the process is called taking the short-time

Fourier t ransform. In implementing the ST FT, researchers

began to experime nt with the window. How large should

the time interval be? W hat if we sha pe th e window to give

mo re weight to the cen tral points and less weight

to

the

end points? Differe nt windows will prod uce different

short-time distributions. U nfortunately the estimates of

the pro perties of the signal are window dep end ent making

interpretat ion

of

the results difficult.

Traditional signal analysis deals with the

The STF T was the first tool devised for analyzing a

Time Frequency

Distributions

The motivation for the study of time-frequency

distributions is to improve upon the STF T. Th e basic

concept is to devise a joint function of time and frequency

3

I E E E ES

Systems Magazine November 1996

7/23/2019 00544024


0 0 5 0 1

0 1 5 0 2

0 2 5

0 3

35 4 0 4 5

5

F i e q u e n q

Hz)

Fig.

2.

Time-Freqeuncy D istribution for M ultiple Signals

that

will

describe the energ y of the signal accurately in

both time and frequency. Th e word “distribution” may be

puzzling to some. On e should think of it as a 3D surface

plot of how the energy

is

“distributed” in the time

frequency cells. For exam ple, in Figure

2

the

time-freq uency distribution for se veral signals is shown .

Pres ent in this plot is a linear chirp mo ving down in

frequenc y, a frequency hopping signal increasing in

frequenc y, and a frequ ency varying signal having

non-linear properties.

A

time-plot of the sum of these

three waveforms is toward the left running up the page.

Th e power spec tral density i:; shown below the ma in figure.

Th e nature of these signals is not obvious from either the

time-plot or the power spectral density. Tha t

time-frequency distribution clearly provides a clearer

representation of the characteristics of these signals. The

time-freq uency distribution tells not on ly what frequen cies

exist, but at what time e ach existed m aking multiple signals

much easier

to

separate and identify. In other words, the

power spec trum density tells us the frequencies that

existcd for the whole duration of the signal. The

time-frequency distribution allows

us

to determine the

frequencies at a particular time.

W hat exactly is wron g with the

STET

you might

ask? The STFT is easily unde rstan dable and it gives a good

time-frequency representation for m any signals. However,

it can be shown mathematically that the

STFT

does not

satisfy what are called “m arginal energies.” Hen ce,

something is being added or subtracted from the

representation. If the joint density of the time-frequency

distribution satisfies the individual inten sities in time and

frequency, “marginal energies” are satisfied. But, for the

STF T this condition is not satisfied.To d o so would

require an arbitrarily small window in both time and

frequenc y. This is contradictory. A small window in time

results in a wide frequency window. The concept known as

the U ncertainty Principle states that good time and

frequency resolution cannot be simultaneously achieved.

On e must be sacrificed at the expense of the other.

600

400

200

12

0

200

frequency

Fig.

3A.

STFT

loa:

0

I

5 0 4

I

1

0

frequency

Fig. 3B. Wigner-Ville

Fig. 3.

Cross-Terms Produced from

Wiper-V ille Distribution

To

satisfy the m arginal conditions, other

distributions such as the Wigner Distribution W D) have

been developed. The W D is a quadrat ic non-l inear)

distribution that will produce interference terms, also

called cross-terms, when multiple signals are analyzed.

Although the WD provides improved time a nd frequency

resolution, the presence of th e cross-terms is a

disadvantage. A variant of the W D, called the

Wigner-Ville Distribution W VD ) incorporates smoothing

to decrease the effect of cross-terms by using indep end ent

windows in time and frequency. The W VD ) is also a

qua dra tic distribution but through the choice of the length

of the time and frequency windows, reduced cross-term

suppression is obtainable. Figure

3 illustrates the

generation of cross-terms from two chirp signals through

the use of the WV D. Oth er dis tributions have been

developed both to minimize the effects of cross-terms and

because they ar e simpler to implement in software. Th e

main stumbling block in attempting to use th e wide variety

of time-frequency analysis methods available is the fac t

that the ir behavior is dramatically different from one

problem to the next and each has peculiar properties.

It

is

IEEE

ES ystenis

Magazine November I996

33

7/23/2019 00544024


time

Fig.

4A.

Wigner-Ville

“0 5 1 15 20 25 30

35

40 45 50

lime

important t o recognize that even though

a

distribution

may not behave properly in all respects or interpretations,

it may still be useful if a particular property is to be

exploited. This po int is emphasized in Figure 4 using

topdown plots of the W igner-Ville, Choi-Williams, and the

Rihaczek -Marge nau distributions on identically the sam e

signal environm ent. Althou gh, each distribution is

different in ap pea ranc e, they are equivalent in the sense

that each can be obtained from the othe r and they each

contain the sam e am oun t of information. They are very

different, but n onetheless each has been used successfully

for particular applications. These are just thre e

possibilities out of a large number of choices, all with

different behavior. Ther e has been considerable

controversy in t he past few years regarding the choice of a

quadratic time-frequency distribution for the analysis of

non-stationary signals. T he numero us distributions which

have been propose d may be interpreted as smoothed

versions of the W D, w ith the type of smoothing

determining the am oun t of attenuation of interference

terms, loss of time-frequency resolution, and mathem atical

properties. H er e ag ain, the choice of th e best distribution

depends o n the na tur e of the signals to be analyzed and on

additional issues such as the mathematical pr operties

required a nd limitations in compu tation and storage. A

successful application of time-frequency distributions

presupposes so me de gree of expertise on the part of the

user. It is seldom possible to view time-frequency analysis

as a

“black box” where the signal is input and som e clear

and m eaningful result is automatically obtained

as

the

output. Som e prior knowledge about the signal must

generally be known in order t o select the most suitable

distribution and a da pt the param eters to the signal, [l

1

are outstanding sources for

a

description of many or t he

Fig.

4B.

Choi-Williams more common~t ime-f requencyistributions.

“0 5

10

15

20 25

30 35

40 45

50

time

Fig.

4C.

Rihaczek-Margenau

Fig.

4.

Differences

in

Various

Quadratic Distributions

Wavelets

Th e Wav elet Transform WT) is of interest for the

analysis of non- station ary signals, because it provides still

another al ternat ive to the S TFT and many of the quadrat ic

time-frequency distributions. Th e basic difference is in

contrast to the STF T, which uses

a

fixed signal analysis

window. Th e W T uses short windows at high frequencies

and long windows at low frequencies. This helps to diffuse

the effect of the Uncertainty Principle by providing good

time resolution a t high frequencies and good frequency

resolution at low frequencies. Unlike many of the

quadratic functions such as the W igner-Ville and

ChoiWilliams distributions, the WT

is

a linear

transformation therefore extraneous cross-terms are not

generated. Ther e is on e other major difference between

the ST FT and th e W T. The STF T uses sines and cosines

as

an ortho gona l basis set to which the signal of interes t is

effectively correlated against. Th e W T uses special

“wavelets” which usually comprise an orthogonal basis set.

The W T then com putes coefficients, which represents a

measure of the similarities, or correlation, of the signal

34 IEEE

AES

Systems Magazine

November

996

7/23/2019 00544024


with respect to th e set of wavelets. In othe r words, the W T

of a signal correspond s to its deco mpo sition with respect

to a family of functions obtained by dilation’s or

contractions) an d translations moving window) of an

analyzing wavelet. A filter bank conce pt is ofte n used t o

describe the WT. The

W T

can be interpre ted as the result

of filtering the signal with a set of bandp ass filters each

with a different center frequency

f.

In the S TF T case, the

bandpa ss filter’s bandwidth is indepen dent of th e ce nter

frequency. In contrast, the bandwidth of th e W T is

proportional to

f

or equivalently, the filter’s quality factor

Q Q

f

/ bandwidth) is independent of f . In othe r words,

the W T can be viewed as a “consta nt-Q” analysis.

Another interpretation of the WT is associated with

multiresolution analysis, where the de compo sition, with

respect to an orthogonal basis set, is per form ed by an

iterative scheme based on high pass and

low

pass filtering

followed by dow nsampling. The signal

is

then decomposed

into a discrete set of orthogonal details from which the

signal can be exactly reconstructed. Thi s offers the

potential for signal and data compression. Still another

interpreta tion suggests that the ba nk of filters represents a

set of “matched” filters whose outputs represe nt the

degree o r correlation to a signal fea ture of intere st. It is

importa nt to note th at, within certain technical constraints,

the “moth er wavelet” may be chosen arbitrarily. This

me ans that a n analyzing wavelet with prop erti es especially

suited to the analysis of som e particular class of signals,

such as spread spectrum waveforms, may be chosen to

supp ort a given application.

Discrete Wavelet

Tramtwm-

Haar

Wavelet

om BPSK

0

5

10 IS

20 25

Fig.

5.

Signal Analysis with a Ha ar Wavelet

on a

BPSK

Signal

Figure

5

illustrates the use of a Fast Wave let

Transform FWT ) on a BPSK modulated signal. More

signal detail is visible at the lower end of the vertical axis.

Th e actual BPSK keying can be clearly extracted at the

1/64 scale and the carrier can be extrac ted U128 scale.

X t) x x

Fig. QA.Time D omain Implementation

of 2nd Order CyclostationaryProcess

exp -jxnt)

exp l

mt

Fig. 6B. Frequency Domain Implementation

o f 2nd Order CyclostationaryProcess

Cyclostationarity

by essentially all digital signals and so me naturally

occurring waveforms. As before, a stationary signal is on e

whose statistics do not vary with time. There fore, a

stationary signal can be sampled at periodic intervals

without being concerned that the signal may he changing

over time.

A

cyclostationary signal is periodically

stationary. Tha t is by delaying the signal by some am oun t,

the statistics do not vary with respect to the signal before

the delay. By processing signals as cyclostationary, we ca n

take advantage of the periodic feature s

of

a waveform.

These preiodicitites arise from modulating, sampling,

keying, scanning, coding, multiplexing, and o the r similar

operations, or from naturally occurring periodic events or

the motion in rotating machinery.

PSD) function fully characterized the second-order

statistical behavior in th e frequency domain of a stationary

random signal, the spectral correlation density

SCD)

function fully characterizes t he second-ord er statistical

behavior in the freq uenc y domain of a cyclostationary

signal. That

is,

unlike stationary signals, such as therm al

noise, some spectral c om pon ents in cyclostationary signals

will correlate with each oth er. There ar e

two

intuitive ways

to view th e co nce pt of cyclostationary signal processing: in

the time dom ain and in the frequency dom ain. In the time

domain, consider a simple delay-and-multiply opera tion as

shown in Figure 6A. If the signal contains a periodic

component , and if the delay is chosen properly, a strong

sinusoid will be p resent at the output. Th e computation of

the

SCD

consists of perfo rmin g this operat ion over a wide

range of delays. Taking the Fo urier Transform of eac h of

these outp uts will produce the

SCD.

In the frequency

domain, Figure 6B, consider up-shifting the frequency

spectrum of interest by some small amount then

down-shifting the spectru m by the sam e amount and the

Cyclostationa rity is a statistical property exhibited

In th e sam e way that the power spectral density

IEEE AES SystemsMagazine November

1996

35

7/23/2019 00544024


I101

2 4

I

C M

Wl

h.ome HZI

Fig. 7A. 3 D View

SPECT RAL CORRELATION DENSITY 194ul 95

1 ’

’ 4 t

Fig.

7B.

End View

Fig.

7.

Spectral Correlation D ensity

compu ter correlation of the two spectrums. If th ere is

correlation between shifted spectral compone nts, spectral

lines will be gene rated. Rep eating this process o ver a

range of frequency shifts will also produce the S CD .

Figure

7

illustrates a typical

SCD

for a

BPSK

signal

showing both the carrier frequency 16 Hz) and data rate

0.5Hz).

Th e end view plot helps read the se rates. Th e

cycle frequency equates to t he am ount of frequen cy shift

in the frequency domain interpretation of th e

SA.

APPLICATIONS

Much work is underway to develop mo re advanced

signal processing techniques which will mo re effectively

and efficiently exploit modern digital co mm unication and

rad ar signals. These techniques are directed at improving

th e tasks of detection , classification, and identification.

Th e traditional STF T has been applied to signal

processing problems in many different area s including

electronic warfare. Some of the major applications for the

ST FT include time-varying signal analysis, system

identification and spectral estimation, signal detection and

par am eter estimation, speaker identification, speech

coding, estimation of th e grou p delay

or

the instantaneous

frequency of a signal, and complex demodulation. Besides

processing received signals, these same

STFT

algorithms are

used to synthesize signals using inverse transform techniques.

Some applications of

STFT

synthesis techniques are

time-varying iltering, non-linear noise removal, room

dereve rberation, time-scale modification, dynamic range and

bandwidth c ompression, and waveform d esign.

While many conventional statistical signal

processing metho ds treat random signals as stationary,

cyclostationary techniques take advantage of periodicities

associated w ith signals. C yclostationary signal processing

has been shown to b e very useful for signal processing

tasks such as the separation of spectrally overlapping

signals and reliable extraction of information from

spectrally overlapping signals. For example, information

such as em itter location, modulation type, and carrier and

clock frequencies c an m ore easily be removed in congested

R F environm ents through cyclostationary signal

processing. Th e presence of signals buried in noise and/or

severely masked by interferenc e can also be mo re easily

detected by exploiting the spectral redundancy associated

with cyclostationarity. Estimating paramete rs such as the

time difference-of-arrival at two reception platforms or th e

direction of arrival at a reception array on a single

platform is improved over conventional systems that

ignore cyclostationarity.

for th e analysis and processing of non-stationary signals for

which sep arat e Time-dom ain and frequency-domain

analysis are not a deq uate . Researchers have applied the

Wigner D istribution for signal detection, spectrum and

instantaneous frequenc y estimation, and pattern

recognition. Synthesis techniques have been used to

perform time-varying filtering, multi-component signal

separation, a nd window and filter design. A quadratic

time-frequency representation known as the ambiguity

surface has bee n used extensively in radar and

communications. In th e radar case, an estimation of the

distance a nd velocity of

a

moving target is made, where th e

distance and velocity correspond to th e “r ang e” and

“D opp ler shift” parame ters. T he cross-ambiguity surface

provides pert inent information about the performance of

the m aximum-likelihood estimator, thus aiding in the

design of the transmitted signal. Synthesis techniques can

also be used for isolating a desired comp onent of a

multicomp onent signal, provided the signal term of

interest do es not overlap significantly with oth er signal

terms.

Wavelet theory provides

a

unified framework f or a

variety of signal processing applications. For example,

wide use is found in multiresolution signal processing and

speech and imagc compression and enhancement. While

conceptually , the W T is

a

classical constant-Q analysis

concept, applied m athematicians have recently recognized

the m any different views and applications stem from a

single theory. Still ano ther alternative to the STFT, the

Time-freq uency representations ar e powerful tools

36

IEEE AES Systems Magazine November 1996

7/23/2019 00544024


main ap plication of the W T iin signal processing will be in

non-stationary signal analysis. The zooming p rop erty of

wavelet analysis allows a very good rcp resen tation of

discontinuities in the signal. For example, as applie d to

image processing, the W T is very good at detec ting and

enhancing edges. This property is useful for pulsed radar

signals, detecting the frequency transitions of frequenc y

hopping radios, or any abrupt transitions characteristic of

the signal-of intere st. Perh aps the biggest potentia l use for

wavclet analysis is in signal compression, thus allowing

increa sed bandwidth efficiency.

FUTURE RESEAKC

As military comm unication, radar, and navigation

systems become mor e complex, frequency employing

sophisticated anti-jam

AJ)

o r

LPI

signaling schem es to

provide robustness or covertness, the perform ance

limitations of the traditional radiometric energy dctector

becomes significant. Featu re extraction techniques

beco me increasingly m ore difficult

as

the covert

commu nicator attemp ts to suppress and conceal his

features . Electronic countermeasures become more

difficult and less effective as .AJ techniques arc

implem ented. Classical radiometric me thods fo r energy

dete ctio n are highly susceptible to unknown and changing

noise level and interference activity. For e xample, spread

spectru m modulation techniques make it difficult for the

radiom eter detectors to function because th e signal is

spread in bandwidth to obtain som e processing gain in the

presence

of

interfcrence. Spread spectrum may also

cmploy spectral overlapping techniques such as

code-division multiple access CD MA ) which will confuse

the ra diom eter since multiple, similar looking signals,

share a common bandwidth. New signal detection and

fea ture s extraction systems are neede d to effectively

exploit these new waveforms in today’s complex signal

environment. Developing new signal counterm easure

techniques and assessing the performance of the se

techniques against a candida{ AJ/LPI waveform designs

requires new theoretical based approaches an d

computational tools and techniques.

with costly and c omplex instruimcntation or through

com puter analysis. Com puter algorithms have be en

developed which analyze and graphically display the results

of the data f or user interpretation. However, new transforms

are currently being developed

that

provide improved

graphical represen tations of time varying dat a beyond that of

conventional

FFTs.

Pattern recognition techniques will be

merg ed with artificial intelligence technology and neural

networks to develop automated analysis and interpretation of

signal data in real-time. By establishing a databa se of

signatures signals will ultimately be automatically recognized

and

an

optimal jamming strategy generate d.

Spectral correlation analysis instruments or

cyclostationary algorithms will become standard

Timc-series data analysis is presently performed either

equipment in com munications laboratories and production

facilities with either comm ercial or educational missions.

Analysis systems co uld bc designed f or quality con trol

monitoring, testing, system design, perfo rma nce

evaluation, teaching, an d research and de velopment. Fault

Testing and prediction are possible through the analysis of

time-series vibration data of rotating mechanical systems

such as engines. Higher-order cyclostationary moments

and cumulants are being researched

to

find new

techniques and algorithms for signal detection,

characterization, and exploitation against LPI w aveforms

or against signals in difficult environments.

place regard ing advanc ed signal processing techniques in

the last

15

years. C omp uter m odels and simulations have

shown the advantages of thesc techniques. Recently,

algorithms have be en developcd which speed the data

processing of these potential techniques

to

realizable

goals. Hardw are is being developed and integrated into

curre nt military systems using many of these, or related ,

techniques. Industry needs

to

be made aware

of

the

possibilities tha t a dvan ced signal processing techniques

offers, for cxample, new signal processing techniques a re

currently being devel oped w hich cxploit the periodic

natu re of natu rally oc currin g signals. Periodic, cyclic, or

rhythmic pheno mena arise naturally in many areas

of

disciplines. Som e

of

the fields whe re periodic, time-series,

dat a are analyzed includ e medicine, biology, meteorology,

climatology, hydrology, oceanology, and econom ics. The

techniques that are being researched for the detection

characterization, a nd identification of

LPI

waveforms

show great promise in these o ther scientific and

commercial fields. Quadratic time-frequency distributions

have served a s useful analysis

tools

in

fields as diverse as

qua ntum m echanics, optics, acoustics, bioengineering,

image processing, and oceanography. These techniques

have been use d to analyze spee ch, seismic data, and

mechanical vibrations.An exccllent example is usc

of

these techniques fo r recognizing cardiac patterns in the

fields of medicine and biology. Wavclets and

time-frequency distribution are being used for the

detection of elcctroencephalogra m EEG) spikes,

ventricular fibrulation in electrocardiograph EC G)

patterns, and a variety of o ther biomedical related

waveforms. Resear ch efforts have been m ade and continue

to

make impo rtant groundwork contribution to E W

programs, but also continue

to

provide research benefits to

other applications.

Considerable theo retical development has taken

REFERENCES

[ l ]

Cohcn, Leon, July 1989,

“Time-Frequency Distributions Review,”

Proceedings

of

the IEEE,

Vol. 77, No. 7

[2] Hlawatsch, F. and Boudreaux-Bartels, G.F., April 1992

“Linear and Q uadratic Time-Frequency Signal Representations,”

IEE E Signal Processing Magazine.

IEEE ES

Systems

Magazine

November

IN6

37

7/23/2019 00544024


[3]Riou l, Olivier

and

Vetterli, Martin, October 1991,

“Wavelets

and

Signal Proce ssing,”

IEEE

Signal Processing Magazine.

[4]

Gardner,

William

A., April 1991,

“Exploitation

of

Spectral Redundancy in Cyclostationary Signals,”

lEEE

Signal Processing Magazine.

Since January 1991,

James P. Stephens

has been employed as

a n

electronics engineering with W right

Laboratory , Avionics Directo rate, R F Technology Division, Electro nic Combat B ranch,

a t

Wright-Patterson Air

Force Base, Ohio. H e is responsible for the planning and exe cution of research projects which will exploit the

theory of communicat ions jamming concepts appl icable to future Air Forc e need s and objectives . H is research

activities involve both in-h ouse and contractual efforts in comm unications coun termea sures toward the

developm ent and ev aluation of con cepts, systems, and supporting techn ologies with emphasis

on

spread spect rum

and digital signal processing techniques. Mr. S tephe ns was previously assigned

to

Foreign Te ch no loa Division,

Directorate

of

Technolo gy and Th rea t, as an electronics engineering analyst from 1982 to 1991. H e was also

employed by the Fede ral Com munication Commission from 1969 to 1982. Mr. Stephens received

a n

M S E E f r om

the Air Fo rce Institute of Tech nology in 1990 and

a

BS EE from W est Virginia Institute of Technology in 1969.

DARSAT-1

Observation Satellite:

adar SAR) Image

Th e f i rs t RADA RSA T image was acquired un der condi t ions

of darkness, overcast skies, rain, and s trong wind, at 5:41 p.m.

local time on the evening of November 28,1995. T he sa te l l i te

was in the ascending, east-looking pass of its

348th

orbit and its

imaging mode was a Standard 1beam , with an in cidence angle of

23

degrees. The image shows a portion of Cap e Breton Island ,

Nova Scotia, Canada, and is centere d at lati tude N 46” 27’ 05”

and longi tude W 060” 18’50 ”.

t

covers

a n

area of 132 km

x

156

km with

a

spatial resolution

of

about

25

m, and was obta ined

from an alti tude of close

to

800 km.

lines, harb our structures, railways, and m uch of the city of

Sydney, including the thre e wharves in Sydney Harbo ur,

are

evident. Local roads and runways at Sydney Airport ar e readily

identified. These are indicative of RAD ARSA T’s util i ty in land

use mapping and urban mapping.

Coastal Delineation:

Shorelines and coastal features, which

are

often obscured by fog o r clouds, are clearly visible. Th e stee p cliff

of Ca pe Sm okey, the coastal inlet

of

Ingonish Harbou r and i t s

barrier beaches along the northeastern sh ore

of

Nova Scotia

are

evident. Middle He ad peninsu la, which divides North a nd S outh

Ingo nish Bays, is also seen. RADARSA T’s coasta l moni tor ing

capability is not only impor tant

to

Canada but

also

to many other

par ts of the world, especially the predomina ntly cloud covere d

tropics. The identification of coastal fe atures

is

also useful in

environmental monitoring programs.

Surface Landf orm :

This image reveals the l inear fea tures of

the drum lin fields elongated hills shaped by glaciers) in the

Island’s Mira region. Th e prominent Aspy Fau lt and oth er major

faults and bounda ries between rock types are easily identified.

Th ese il lustrate R ADARS AT’s sensitivity

to

surface topography

and landforms, making RAD ARS AT

a

valuable tool for

detectin g surface mineral deposits and geological features .

Cultural Features:

Appear ing as bright spots, buildings, power

hip Detection and Monitoring:

Ship detection a nd surveillance

of activities in shipping lanes highlight another R AD AR SA T

capability.

I n

this image, the ship Portland Carrier and its

V-shaped wake can be seen. Th e shape of i ts wake indicates that

the ship was inbound towards Sydney Harbour.

sensitivity

to

wind effects on the oce an surfac e, especially where

the water is exposed to t he ful l force of the winds. The mapping

of currents , sea s ta te , and many o ther ocean features for weather

forecasting, ship routing and fisheries resource man agement is

possible.

Forestiy:

West of Ingonish, south of the Che ticamp reservoir

white), i t is possible to see the forested areas darke r) dama ged

by the spruce budworm outbreaks

of

the 1970s and 1980s.

The pu rpose

of

t he R A D A R SA T p rogram of the Canadian

Space Agency is

to

provide ra dar imagery

on

an ope rationa l basis

to com mercial, government and scientific users worldwide.

RADARSAT’s onboard tape recorder and RSI’s in ternat ional

network of distributors and gro und stations ensure that

users

around the world will have access

to

high-quality im ages

of

the

Ea rth, regardless of weather and light conditions.

RADARS AT-1, funded by the Fede ral and Provincial

Gov ernme nts, was built by the prime contractor S par Aerospace

Ltd. The sa te l l i te launch on

4

November

1995

was sup plied by

the Un ited Sta tes in return for radar imagery for use by NAS A

Nat ional Aeronaut ics and Space Adminis t ra t ion) ,

NOAA

National Oceanic and Atmosp her ic Adminis t ra tion) and o ther

US

Government agencies .

Comm unications Branch, 6767, route d e I’ACroport,

Saint -Huber t , Quebec J3Y 8Y9; Telephone 514) 926-4351, Fax

514) 926-4352, WWW Site http://radarsat.space.gc.ca/, Site

R A D A R S A T s u r

WWW

http://radarsat.espace.gc.ca.

Ocean Su8ace:

Th e br ight patches show RADARSAT’s

Fo r more inform ation, contact: Canadian Space Agency,

38

I E E E

AES

Systems

Magazine

November

1996

http://radarsat.space.gc.ca/

http://radarsat.espace.gc.ca/

http://radarsat.espace.gc.ca/

http://radarsat.space.gc.ca/

Documents

00544024