541
1 INTRODUCTION
Audio watermarking (AW) corresponds to digital
information imperceptibly embedded into the audio
signal.AWformaritimeterrestrialradiotelephonyis
inspiredfirstofallbytheabilityofimplementationan
automatic identification in the voice telephone
communication. Applied to the existing analog
radiotelephony in all frequency bands (VHF, MF,
HF), the watermarking system could overcome
existinglimitations, enhanceefficiency and clearness
of radiotelephone messaging while ultimately
decreasingsocalled“humanfactor”.
In the maritime and aeronautical VHF services
analogue channels with frequency/phase and
amplitudemodulationcorrespondinglyareutilized.
For the meanwhile the identification of the sea
vesselsisrealized
bymeansofverbalcallingofship’s
call sign or numerical identification. However on
accountofdifferentreasonssuchverbalidentification
maybeabsent,transmittedwithdelay,orunderstood
with errors. This problem, applied to VHF
communication is illustrated in Fig. 1. Motor vessel
“Arcona”transmits a certain messageto all
stations.
Butoneofthereceivingvesselsmissedthenameand
call sign of the transmitting ship, and another ship
interpreted the name of transmitting ship as
“Gargona”instead“Arcona”.
It is obvious that false, incorrectly interpreted or
delayed verbal identification negatively affects
maritime navigation. Automatic identification could
avoidmisidentification
andcallsignconfusion.
Radio Regulation claims obligatory identification
for every radio transmission. In a present maritime
terrestrialradiotelephonyidentificationiscompletely
dependent upon operator. This situation takes place
onthebackgroundofhightechnologyinstrumentsis
beingimplementedinthemaritimebranch.
Stealthy Information Transmission in the Terrestrial
GMDSS Radiotelephone Communication
A.V.Shishkin&V.M.Koshevoy
OdessaNationalMaritimeAcademy,Ukraine
ABSTRACT:Audiowatermarking(AW)technologyisconsideredforstealthyinformationembeddingdirectly
intoaudio signal.Inherent inanalog radiotelephone channel interferences against watermarks areanalyzed.
Robust encoding/decoding algorithms are presented and appropriate project of AW system named as
AutomaticRadioTelephoneIdentificationSystem(ARTIS)is
proposed.Experimentalresultsforthepractical
VHF radio channel are presented. Relating on processing complexity the designed system enables
imperceptiblytransmitdataonarateupto260bit/secinthestandardVHFradiochannel.ARTISprovidesthe
full compatibility with the existing radio installation, and doesn’t require replacement of standard
VHF
transceivers and operational procedures. Besides, automatic identification the system may be used in the
specialapplications,forexample,underthethreatofterroristattack;generallycontributestonavigationsafety
andinformationsecurity.
http://www.transnav.eu
the International Journal
on Marine Navigation
and Safety of Sea Transportation
Volume 7
Number 4
December 2013
DOI:10.12716/1001.07.04.09
542
It is well known an Automatic Transmitter
IdentificationSystem(ATIS)[1]formarineVHFradio
that is used and mandated on inland waterways in
Europe for identifying the transmitting vessel. In
ATIS the identity of the vessel is sent digitally
immediately after the shipʹs radio operator has
finished talking
and releases push‐to‐talk (PTT)
button. Identification is performed by appending a
shortdatamessageinDigital SelectiveCalling(DSC)
format. The main drawback of ATIS is post‐report
transmissionofidentificationdata.
In COMSAR proposal [2] the necessity of
automatic identification is grounded and quite
reasonably noted that
the identification should be
done immediately after pressing the PTT button on
the contrary of ATIS releasing PTT. Another
shortcoming of ATIS is principle limitation for
Medium/High frequency (MF/HF) applications. This
limitationresultsfromthesmallbitrate(100bit/sec)in
comparisontoVHFDSCrate(1200bit/sec).
In restricted navigation environment
the
immediate and clear automatic identification is
extremely necessary. Automatic identification would
excludethehumanfactorandincreaseanefficiencyof
VHFradiocommunicationandmaritimesafetyinthe
whole.
Figure1. Automatic identification in the VHF maritime
radiotelephony
Only verbal identification doesn’t protect against
illegal radio transmission. Illegal transmissions are
especiallyharmful ontheVHFdistresschannel16.Of
course, unauthorized transmissions are performed
anonymously. Reliable automatic identification of
such transmissions could avoid the violation of
radiotelephoneregulation.
Another advantage of automatic identification
follows from the ability of digital
information
inputting to another ships’ navigational and
information systems, for example Electronic Chart
Display and Information System (ECDIS). ECDIS
makes visualization of neighboring vessels in the
range of VHF radio (i. e. approximately 30 nautical
miles).Howeverthetransmittingvesselbynomeans
is marked in an electronic map. Automatic
identificationwouldimplementthevesselmarkingat
electronicchartdisplay.
One more application of AW is a covered
information transmission in the special applications
(forexample,facingthethreatofterroristaggression).
ItisessentialthatAWbasedidentificationdoesn’t
require altering an existing radio installation and
operational procedures. To introduce
AW‐
identification function only new telephone receiver
withthe embedded processorat the transmitter side
andprocessorwithmini‐displayswitchedtocommon
audiooutputatthereceiversidehavetobemounted.
Automatic identification starts right away PTT
pressing and runs continuously during all
transmitting period independently from voice
signal
occurrence. No additional time and frequency
channelrecoursesarerequired.
AW‐identification provides the full compatibility
withtheexistingtransceiversandmakespossibilityof
step‐wiseimplementation.
SimilarapplicationofAWmaybeimplementedin
the aeronautical (118…136) MHz mobile service. In
paper [3] watermarking based on speech unvoiced
phonemes
recognizingandreplacingthembycertain
noiselikesequenceisproposed.
Maritime AW based identification is proposed in
[4],[5].Inthesepaperswatermarkingprocessdoesn’t
requireanyrecognizingalgorithmsandbasedonlyon
statisticalsignalproperties.
In this paper we present stealthy AW forming
algorithm based on signal energy
saving and
complete algorithmic ARTIS scheme for maritime
application.
2 INFORMATIONEMBEDDINGALGORITHM
2.1 Audiowatermarkingsystemanditscharacteristics
AW for maritime analog radiotelephony has some
important features. First, watermarks are inaudible.
Second, watermark technology doesn’t demand any
additionaltimeandfrequencyresourcesbeyondthat
used in the basic telephone
channel. Thirdly,
watermarks besides identification may convey also
another digital. Finally, watermarking doesn’t
demandalterationofstandard radioinstallation and
operational procedures. Moreover, looking forward
towardsdigitalcommunications,voicewatermarking
straightawayinthemicrophonecouldalreadysolve
identificationproblem.
Watermarking system is characterized by some
competitive properties: fidelity, data
payload and
robustness [6]. Fidelity defines an audible similarity
betweenanoriginalvoicemessageandwatermarked
message after information embedding. Data payload
referstonumberofbitsawatermarkencodeswithina
unitoftime.Robustnessistheabilityofwatermarksto
survivethevoicesignalprocessingoperations.
543
2.2 Modelofwatermarkedcommunicationchannel
Therearetwoapproachesforinformationembedding
inaudiosignal:a)spreadspectrum(SS)methodand
b)quantizationmethod.Thefirstmethodisbasedon
independentfromtheaudio(host) signalforming of
watermarking signal. Embedded information bit is
spread by multiplying on certain
pseudo random
sequence (PRS) like in mobile communication.
Herewith duration of elementary chip comes to
inversevalueofsamplingfrequency.Onthecontrary
quantizationmethodrelaysonthehostsignal.
Model of communication channel with additive
watermarkingispresentedinFig.2.Theencodermay
berealizedintheform
ofa)blind(ornoninformed)or
b)informedencoderdependingonignoringorusing
theinformationaboutthehostsignal
x
.Thevariant
b) is reflected with dashed line. The watermarked
signal
s
is additively formed from signals
x
and
w :
sxw
(1)
Figure2.Modelofwatermarkedcommunicationchannel
Power
2
w
of watermarked signal w is limited by the
acceptable level of introduced distortions of carrier signal
becausewatermarkpresenceshouldbenotaudible(orquite
tolerable)onthebackgroundofcarriersignal
x
.
In the channel according the Fig. 2 two interferences act
against watermark w : the first interference is itself the
carrier signal with the power
2
x
, and the second
component–anoise
n withthepower
2
n
.
Watermarkingchannelischaracterizedbyitscapacity–the
maximum achievable code rate. Assuming the both
interferences are white Gaussian noises the capacity
C
(bit/sample) of watermarked channel with noniformed
encoder, when host signal is not available to encoder, is
definedbyShannon’sformula:
2
22
1
1
2
w
xn
C
. (2)
Practically
22
x
n
, and capacity is limited
mainlybythehostitself.
At the same time using information about the
carrier signal
x
, it is possible to increase C for
informedencoder.Inthepaper[7]itwasshownthat
assuming the host is known at the transmitter, the
capacity of watermarked channel is defined by the
formula:
2
2
1
1
2
w
n
C
. (3)
Formula (3) shows that appropriately considered
carrier signal doesn’t influence on watermark
transmissionandthe capacityis determinedonly by
the second noise, which is unknown at the encoder.
Capacityforsuchchannelisincreasinggreatly,that’s
why an informed encoding (i.e. “writing on dirty
paper”) is attractive method
for watermarking on
accountofitspotentialcapacity.
2.3 Onechannelencodingalgorithm
Let
12
( , , ... , )
L
x
xx
x and
12
( , , ... , )
L
uu uu are
ahost signal vector and binary PRS:
(1,1)
i
u
in
Euclidian
L ‐dimensional space with norm and
scalarproductdefinedasfollows
22 2
12
=++...+
L
x
xxx , (4)
11 2 2
( ) = + +... +
LL
x
uxu xux,u . (5)
ItshouldbenotedthatEuclidiannormphysically
presentssquarerootfromthesignalenergy.
The task of encoder and decoder in the general
schemeinFig.2maybeexpressedinthenextmanner.
Encoder must embed one bit of information
(1,1)m
, i.e. it must watermark the vector
x
,
forming a certain new vector
12
(, ,..., )
L
ss ss
underprovisionthefollowingconditions:
=minx-s
, (6)
x=s
, (7)
() (,,)Qxm
s,u
, (8)
where
=( )x
x,u , ()Q
‐quantizationfunction.
The target function (6) is grounded by
minimization of introduced distortions because of
watermarking process. Actually, under the equal
conditions the less distortion, the stealthier is the
watermarkingsignal
w=x-s.
Limitation (7) results from containing the energy
of watermarked signal. Keeping the signal energy
gives an ability to apply quantization function with
adaptive step proportional to the power of the host
signal in the encoder and received signal in the
decoder. Adaptive quantization eliminates influence
of amplitude scaling on watermarks.
Note that such
adaptivequantization is equivalent to normalization
by signal norm and quantization with the constant
step.Watermarkingschemebasedonquantizationof
normalizedcorrelationbetweenthehostvectoranda
randomvectorisproposedinthepaper[8]forimages
watermarkingproblem.
Optimizationproblem(6)subjecttolimitations
(7),
(8)ingeneralneedsratherdifficultanalyticalsolution
and sizeable computations. The desired optimal
solutionmaybeeasilyobtainedaccordingtothenext
argumentations.
544
Having the host signal
x
it is possible to
compute
x
and then its quantized version
(, , )sQxm
subjectedtoinformationbit
m
and
quantization step
. Quantization function ()Q
has adaptive step proportional to the host signal
norm:
k x , where
(0.5 ... 2)k
assigns
distortions level. To be specific
(1,1)m agrees
with odd and even quantization levels
correspondingly. Quantization step adaptation
provides immunity from amplitude scaling and
contributes to watermarks audio imperceptibility in
thespeechsilentintervals.
From the well‐known relation
(cos()
s,u) = s u angle
between vectors s
and
u (seeFig. 3)maybeobtainedas:
1
cos ( )s
x
u . (9)
Figure 3. Geometric interpretation of norm saving data
embeddinginthreedimensionalspace
The last formula takes into account relations
x
=s
and =( )s
s,u .Requiredvector s lieson
asurfaceof
L ‐dimensionalconethathascenteraxis
u andsubtendingangle 2
.Theconeisshownby
dashedlines.
Fromthe other hand to minimize the embedding
distortionweneedfindtheclosestvectoronthiscone
tothehostvector
x
.Itisclearthatvector
s
liesin
the plane which is formed by vectors
x and u .
Therefore vector
s may be expressed by the
combinationof
x
and u,namely
sxu
, (10)
where
and
are embedding factors to be
found.
To do this let us multiply scalar wise by
0
u
the two parts of relation (10). After evident
transformationsweget
2
=( - )/sx
u . (11)
Thenagainmultiplyingrelation(10)scalarwiseby
0s ,andsolvingquadraticequation,weget
2
2
1,2
2
2
s
x
2
2
xu
xu
. (12)
The nearest to
x
vector s must be taken with
the positive sign for
in (12). This vector
corresponds to point 1 in Fig. 3. On the contrary
negative sign in formula (12) gives the most distant
vector
s (point2).
Generallyembeddingmayberealizedinthetime
or frequency domain after Fast Fourier
Transformation (FFT). The last choice is more
preferable because of less watermark sensitivity to
timeshiftswithinsamplinginterval.Inthefrequency
domain the amplitudes of FFT coefficients must be
usedforinformationembedding.
Mathematically amplitudes are expressed by
nonnegativenumbers.Butifundertakenothingabove
presentedalgorithmmightgivenegativecoordinates
in
s especiallyforlowpowervector
x
.Toprevent
this we should add to previously introduced
limitations(7),(8)onemoreinequalitylimitation
0, 1, 2,...,
i
si n . (13)
For this additional limitation suboptimal
algorithm for minimizing
x-s is derived.
Appropriatepseudocodeisgivenbelow
whilemin(s
i)<0,i=1...n
x=abs(s);
Getnew
,,,x
s
;(according(10)–(12))
end
Practically this algorithm already converges just
onthesecondstep.
Informed encoding eliminates interference on
watermarkingsignalfromthehostsignalbutneedsa
certaintimedelayatthetransmitterside.
2.4 Decodingalgorithm
Receivedvectorisgivenintheform:
y
xwn (14)
where
n –additivenoisevector.
Decodercomputesscalarproduct
=( )y
y
,u
and
thenextractstheembeddedbitbyapplying
(1,1)
ˆ
arg min ( , )
m
myQym
. (15)
For binary quantization decision (15) may be
reducedto
ˆ
msigny
.
3 INTERFERENCESAGAINSTWM
Besides host signal interference there are another
interferences which are essential for watermarks
restorationat thereceiver. These interferences inthe
watermarking theory are identified as attacks. The
mostharmfulattacksare:
545
1 intersymbolinterference(ISI);
2 amplitudescaling;
3 additivenoise;
4 nonlineardistortions(clipping);
5 resamplinganddesynchronization.
Hofbauer et al. [9] proposed to take into account
Doppler effect that is actual for aeronautical
applications, but in the maritime communication it
mightbeneglected.
3.1 Intersymbolinterference
In the radio
channels ISI is usually caused by
multipath propagation. The transmitting medium in
VHF radio communication is the atmosphere, in
which radio signal is transferred by means of
electromagneticwaves.The received electromagnetic
signalisusuallyasuperpositionofaline‐of‐sightpath
signal and multiple waves coming from different
directions. This phenomenon is known as multipath
propagation. The received signal is spread in time
andthechannelissaidtobetimedispersive.
AnotherphysicalcauseofISIisnonuniformityof
frequency response of a channel. Analog low‐
frequency circuits of the transceiver are composed
from reactive elements. Frequency dependent
elementscausenonuniformityof frequencyresponse
within audio signal spectrum. When frequency
response is explicitly nonuniform within signal
spectrum output signal is highly differs from input
one. Distortions caused by bandlimited low‐
frequencychannelalsorepresentISI.
Fromthesignalprocessingpointofviewthetwo
physically different causes (presence
of reactive
elementsinaudiocircuitsandmultipathradiowave
propagation)leadtothesamefinalresultintheform
ofISI.
The most effective measure against ISI is
orthogonal frequency division multiplexing (OFDM)
technology that is commonly used in numerous
communicationsystems.
3.2 Amplitud escaling
Amplitude scaling refers to uncertainty
of incoming
signal amplitude and its slow variations. Coming
back to multipath propagation, one can analyze a
variant when the different path lengths are very
similar compared to the wavelengths of the signal
components. Then the phase variations between
components will be small and they will all undergo
verysimilar
amountsofcancellationorreinforcement.
This case is usually termed flat fading. In
watermarking flat fading is simulated by amplitude
scalingattacks.
In general quantization based methods are very
sensitive to such attacks. Counteractions against
amplitude scaling are adaptive step quantization or
double signal normalization before and after
quantizationtothe
signalnorm.Applicationofthese
operations demand saving the signal norm in the
watermarkingprocess.
The proposed algorithm is exactly designed for
saving the signal norm (or squared root signal
energy).
3.3 Additivenoise
Additive noise is imposed onto the signal during
transmission.Thenoiseresultsfromthermalnoisein
electronic circuits, from atmospheric noise or from
otherradiostations.Quantizationnoisefromanalog‐
to‐digital converter may be attributed to additive
noise. Commonly recognized model of an additive
noise is additive white Gaussian noise, denoted in
Fig.2by
n
.
Additivenoiseismeasuredbysignal‐to‐noiseratio
(SNR)whichpracticallycomesto(15…17)dB.
3.4 Nonlineardistortions(clipping)
Nonlineardistortionsappearinamplitudelimitations
caused,forexample,bytheoverloadinaudiocircuits.
Overload arises from redundant power of
transmittingstation.Thesimplest
modelofnonlinear
distortions is clipping. Radiotelephone AW in any
caseshouldberesistantagainstsuchdistortions.
3.5 Resamplinganddesynchronization
At the transmitter and receiver sampling processes
arenotsynchronized.Itmeansthatsamplinginstants
in the receiver are shifted relatively corresponding
samples in the transmitter. Consider sampling
frequenciesare
equalatthetransmitterandreceiver.
So resampling comes to arbitrary shift of sampling
points. Desynchronization is here understood as
uncertainty in the starting of watermark in the
receiver. For correct watermark separation the
beginningofwatermarkshouldbefirstdetectedand
thenalldecisionpointsarecountedfromthe
starting
point. Analog radiotelephone channel by all means
leads to resampling and loss of the watermark
beginning.
Watermarksensitivitytoresamplingconsiderably
dependsonchoosingofembeddingdomain–timeor
frequency. Assume signal samples are independent.
Correlation time in this case may be accepted to be
1(2 )
cs
F
, where
s
F is sampling frequency.
Frequency domain embedding is based on FFT
computation which uses, for instance,
N
consecutive samples. FFT amplitudes in every
frequency channel have in this case
N
time
enlarged correlation period:
(2 )
cs
NF
.
ThereforefrequencydomainWMisconsiderablyless
sensitivetosamplingshiftsthantimedomainWM.
4 PRACTICALREALIZATION
4.1 OFDMbasedAWembedding
AW are multiplexed directly into audio signal and
thereforesubjectedtoinfluenceofcommonforaudio
signal transformations and interferences. Standard
transformations are: amplification, modulation,
546
filtering.Inthestandardbandlimitedaudiochannel
300 – 3000 Hz ISI impacts destructively on
watermarks.
ISIiscausedmainlybynonuniformityofchannel
frequency response and nonlinearity of phase
response in low frequency transceiver electronic
circuits. The most effective measure against ISI is
orthogonal frequency division multiplexing (OFDM)
technology that is commonly used in numerous
communicationsystems.
OFDM approach for AW was proposed in [10]
and. The main idea lies in embedding every
watermarking bit completely into a certain
narrowband component of host signal. AW
information jointly with host signal form OFDM
symbol.
Construction of OFDM symbol for
AW is
illustrated in Fig. 4. For simplification the figure is
drawnforthenextnumericalvalues:FFTdimension
8N ;lengthofspreadingsequence 5L ;number
ofwatermarkedsubchannels
2
B
.
Vector
x of host audio samples in the time
domain, a) is buffered by columns into
NL
matrix, b); FFT along each column is performed, c);
B
watermarkingbitsareembeddedinthefrequency
coefficients independently along rows, d). In Fig. 4
samples in the time and frequency domains are
shown by squares and circles accordingly; virgin
samples are shown white and modified ones due to
watermarking are grey. As it seen watermark
distortions are distributed along
the whole
watermarkedsequence
s .
Figure4.FormingofOFDMwatermarkingsymbol
Embedding algorithm for each subchannel are
presentedinsection2.
General system for informed OFDM
encoding/decodingispresentedinFig.5.Signal
x is
splitinserial‐to‐parallel(S/P) demultiplexerinto
N
slowflowswhicharesubjectedtoFFT.Messageflow
m is split into some, suppose
B
, /2BN slow
flows. Channel encoders use
B
Fourier coefficients
for watermarking independently in each channel. In
the simple case one bit may watermark one
coefficient.Formorecomplexvariantsoneembedded
bit may be distributed among
1L coefficients. In
general
NL
samples of
x
are needed for
embedding
B
message bits. Algorithms for each
channel encoders are identical. Watermarked
coefficients and all the rest undisturbed coefficients
arethenusedforinverseFFT(IFFT).Againparallel‐
to‐serial (P/S) block combines slow flows into one
sequence
s inthetimedomain.
Figure5.Multichannelrealization: a)encoder,b)decoder
At the receiver signal
y
is split into
N
flows
which are transformed in Fourier coefficients. These
coefficients are processed according demodulation
algorithm for extracting a watermark message bits
ˆ
m
.
Parameters
,,,NLB
are subject for optimal
tradeoffdecisionfidelity–datapayload–robustness.
Datapayloadisexpressedinbit/secbytheformula
/( )
s
R
B
FNL
. (16)
Parameter
L
influences on watermarking
fidelity. The greater
L , the higher fidelity is.
ObjectivelyfidelitymaybeexpressedbyWatermark‐
to‐SignalRatio
20 /
wx
WSR lg
. (17)
Embedding and detection processing for
simultaneous multichannel processing according to
OFDM principle is well realized in the matrix form
thatdoesn’tneedusinganylongdurationcycles.
Watermarking signal
w in OFDM presentation
occupies the whole frequency audio band and
simulatesanordinaryadditivenoise.Byappropriate
choosingofparametersintheembeddingalgorithmit
is possible to eliminate audibility of
w on the
backgroundofhostsignalandexternalnoise.
It should be noted very good stealthy
characteristics for truncated vector
x of length
2L
. For this case embedding distortions,
expressedby
WSR ,maybemuchmorenumerically
increasedcomparativelywithlarge
L .Itisexplained
by concentration of one bit watermark signal in the
short time interval (about 16 msec under conditions
8kHz
s
F
,
64N
). Within this interval the total
audio energy is preserved due to embedding
algorithmandenergydifferencesbetweenhostsignal
x
and watermarked signal s inside the short
speech interval are not perceived. Therefore
watermark energy may be gained under the same
audible similarity. At the same time
L reduction
comes to increasing data payload according to
formula(16)andsimplifiestechnicalrealization.
Variant of ARTIS constructer design is presented
inFig.6.
547
Figure6.ExampleofVHFARTISdesign
4.2 AWdetectionandsynchronization
Estimation of watermarking bits (15) needs proper
synchronization the correlation process.
Synchronizationandmarkerfieldsconsumeacertain
time and decrease an effective information bitrate.
Additionalresourcesarespentalsoforchecksum.
Tosettleallthesetaskssimultaneouslywepropose
tocomputeslidingcorrelationand
detectwatermark
by detecting the whole OFDM symbol using hash
function.Ingeneralhashfunctiontransformsaninput
data of variable size to a fixed‐size string, which is
called hash value. Commonly hash function is
employedforcheckingintegrityofinputdata.Weuse
hash function for the purpose
of OFDM
watermarking symbol detection. Integrity of the
detectedsymbolisensuredautomatically.
To implement this idea let us compose
watermarking data containing information bits
M
itself and its hash function
HM:
[ , ]
M
HM .
On the receiving part decoder is permanently
processingblocksofsamplesuntilhashfunctionwill
satisfy.
Supposecurrent block
1
[, , ]
iiNL
ss
, containing
NL samples, gives estimations of information bits
ˆ
M
and received hash function
ˆ
HM. Decoder
calculateshashfunctionoverthereceivedinformation
bits
()
ˆ
HM . If
ˆ
(
ˆ
)()
HM HM then the decision
of watermark symbol detection is accepted and the
detected bits are believed
ˆ
M
. Otherwise next
incomingblock
1
[, , ]
iiNL
ss
entersforprocessing.
Letusestimateasufficientlengthofhash‐function.
“Good”hashfunctionmapseveryinputcombination
into unique output combination. Then lengths
sequencesofinformationbits
M
andhashfunction
HM should coincide. Suppose
length M length H l. False detection will
takeplaceifandonlyifsomerandom
l ‐lengthinput
combination will give
l
‐length hash function that
coincideswiththesubsequentbits.Theprobabilityof
such event is
2
2.
l
er
p
Taking 8l one can
obtain
16 5
21.510
er
p
for randomly
occurred sequence. Of course, total false detection
willincreaseproportionallythesearchingtime
s
t for
subsequent AW symbol. To reduce
s
t encoder is
blockedaftersuccessivedetectionatthetimeslightly
below the symbol duration (
NL samples in our
notation).
Practicallycyclicredundancycheck(CRC)codeof
length
8l
CRC‐8 with generator polynomial
82
1
x
xx
maybeused.
Theproposeddetectionmethodinourapplication
exceeds standard communication format of fields:
synchronization–marker–data–CRCintheuseful
watermarked information per host sample. This
method provides decoding OFDM watermarking
symbol in the whole and thus saves from the
necessityofsynchronizationand
markerfieldsatall.
Second advantage is the absence of synchronization
erroroncorrelationreceiveraccuracy.
These superiorities are achieved however at the
expense of considerable processing loading in the
receiver.But detection algorithmis based mainly on
FFT matrix operations and may be easily on‐line
realizedforaudiofrequency.
5 EXPERIMENTALANDSIMULATIONRESULTS
Computer simulation in MatLab environment was
held using for host signal the real voice frame of
length 27500 samples. Parameters of simulation are
presentedinTable1.
Table1.Parametersofsimulationandexperiment
_______________________________________________
Parameter,notationNumericalvalue
_______________________________________________
Samplingfrequency,
s
F 8.0kHz
OFDMsub‐channelwidth125Hz
Numberofsub‐channels,
B
15
WSR ‐16/‐14/‐12dB
SNR
12dB
AWrawrate,
R
60/125/268bit/sec
m‐sequencelength,
L
31/15/7
FFTdimension,
N 64
Delayinthetransmitter,
Tx
D 56/120/248msec
Identificationtime0.25/0.5/1sec
_______________________________________________
IntheTable1thefollowingnotationsaretaken:
20 /
wx
WSR lg
‐Watermark‐to‐SignalRatio,dB;
20 /
s
n
SNR lg
‐Signal‐to‐NoiseRatio,dB;
/
Tx s
DNLF
‐Delaytimeinthetransmitter,sec;
/
s
R
B
FNL
‐AWrawdatapayload,bit/sec.
Identification time is estimated taking into
consideration the total number of bits, needed for
representationof 9‐symbol decimal MMSI (Maritime
MobileServiceIdentity).Thisvalueisacceptedby30
bits.CRClengthistaken
8l
.
Simulation have being carried taking into
consideration all interferences and limitations,
discussedinSection2.
Experimental testing was carried out in the VHF
channel17ofmaritimemobileservice(156.85MHz);
emissionclassF3E/G3E(frequency/phasemodulation,
analog telephony) on the base of hardware
installation: maritime VHF transceivers RT‐2048
Sailor,
USB ADC/DAC (analog‐to‐digital
548
converter/digital‐to‐analog converter) moduleЕ14‐
140L‐CARD.
In the trials all watermarking symbols along the
voice frame were detected correctly and no false
detectedsymbolswereregistered.
6 CONCLUSIONANDFUTURE
INVESTIGATIONS
The addressed and properly understood
radiotelephone communication in VHF, MF and HF
bands plays an important
role in general maritime
safety. Significance of particularly VHF channel 16
was emphasized, for instance, in the paper [11].
Automatic identification, in turn, would provide an
effectiveandclearlyunderstoodcommunication.The
considered automatic identification solves a number
problemtowards VHF radiotelephone improvement,
eliminationahuman factorand finallyenhancement
ofmaritimesafety andsecurity.Asopposed toriver
system[1],ARTISisgroundednotontheappending
a certain digital sequence to the radiotelephone
transmission, but realizes just its identification from
the very beginning of the transmission, and
permanentlyrunsduringthewholetransmission.
AW identification doesn’t require an additional
frequency and time resources, alteration standard
transceivers and radio communication procedures
andappearsonlyinadditionalnoise,thatcanbeset
tominimallevel(equalorbelowofthechannelnoise).
Implementation the proposed ARTIS function
needsonlyphonereceiverreplacementbyanewone
with built‐in processor at
the transmitter side and
switchingadecoder tostandardaudio outputatthe
receiver part. It easily provides compatibility of
standard equipment and the equipment with
identification function. Moreover, AW identification
worksaswellindigitalchannelsifthelastwillstart
operatingin maritime radiotelephony. This ability is
ensuredby
theAWresistanceagainstADC/DACand
voicecompressionprocedures.
ARTIS implementation makes possible a further
integrationofVHFradioandnavigation(AIS/ECDIS)
equipment.
ThenexteffortsinARTISdevelopmentshouldbe
stepsinhardwarerealizationforrealtimeprocessing
and investigations of ARTIS adaptability in the
MF/HFbandsofmaritime
communication.
REFERENCE
[1]ETSI EN 300698‐1. Radio telephone transmitters and
receivers for the maritime mobile service operating in
the VHF bands used on inland waterways; Part 1:
Technicalcharacteristics and methods of measurement.
50p.
[2]Automatic Transmission of the Identification of the
Radiotelephone Station. COMSAR 16/7, 15 December
2011.
[3]K. Hofbauer, G. Kubin “High‐Rate Data Embedding in
UnvoicedSpeech”.
http://www.eurocontrol.int/eec/public/standard_page/D
OC_Conf_2006_012.html
[4]Shishkin A. V. “Identification of radiotelephony
transmissions in VHF band of maritime radio
communications”, Radioelectronics and Communication
Systems,November2012,Vol.55,No.11,pp.482‐489
[5]A.Shishkin, V.Koshevoy “Audio Watermarking in the
Maritime VHF Radiotelephony” // Weintrit A.
(ed.):
Navigational Problems. Marine Navigation and Safety
of Sea Transportation. A Balkema Book, CRC Press,
Taylor & Francis Group, Boca Raton – London‐New
York–Leiden,2013,ISBN:978‐1‐138‐00107‐7.pp.293–
298.
[6]Cox,I.J.Miller,M.L., Bloom,J.A.,Fridrich,J.,KalkerT.
2008.Digital
watermarkingand steganography.Second
Edition–MorganKaufmannPublishers.594p.
[7]M.H.M.Сosta, “Writing on dirty paper,” IEEE
TransactionsonInformationTheory,vol.IT‐29,pp.439‐441,
May1983.
[8]Xinshan Zhu, Shuoling Peng “A Novel Quantization
Watermarking Scheme by Modulating the Normalized
Correlation”,IEEE Int.Conf. Acoustics,
Speech andSignal
Processing(ICASSP),March2012,pp.1765‐1768.
[9]Hofbauer,K.etal.“Speechwatermarkingforanalogflat‐
fadingbandpasschannels”,IEEE Transactions on Audio,
Speech, andLanguage Processing, 2009,Vol.17,No.8,pp.
1624–1637.
[10]Shishkin A.V. “OFDM‐based audio watermarking for
electronic radiotelephone identification”,
EWDTS’2010.
St.Petersburg,Russia,2010.pp.190‐194.
[11]Brzoska S. “Advantages of Preservation of Obligatory
Voice Communication on the VHF Radio Channel 16”,
TransNav‐InternationalJournalonMarineNavigationand
Safety ofSea Transportation,Vol.4,No.2, pp.137– 141,
2010.
