comparison and simulation of digital … · comparison and simulation of digital modulation...

CECOM Bottom Line: THE WARFIGHTERFile Name.PPT - Briefer�s Name - Briefed 3/11/031

COMPARISON AND SIMULATION OF DIGITAL MODULATION RECOGNITION

ALGORITHMS

Wei Su and John Kosinski

U.S. ARMY CECOM RDEC

Intelligence and Information Warfare Directorate

Fort Monmouth, New Jersey


Automated Signal Exploitation and Modulation Recognition

Automated Signal Exploitation and Modulation Recognition

signal Demodulated signal

Cooperative Communication modulation

typepulse

shapingsymbol

ratecenter

frequency

?Signal processing

Statistical estimation

?Modulation recognition

?

signal

Non-cooperative Communication ?


Factors Affect the Modulation Recognition Results

Factors Affect the Modulation Recognition Results

hResidual carrier frequency estimationhCarrier frequency and phase trackinghBaud rate estimation and pulse

timinghPulse shaping recoveryhChannel distortion


Center frequency offsetCenter frequency offset

PSK8 signal with phase shift produced by center frequency offset

A PSK8 signal


Pulse Timing

h Re-sampling will be necessary if the symbol frequency divided by sample frequency is not an integer

h Error is introduced in re-sampling

Pulse Timing

4 samples per symbol 10/3 samples per symbol

After re-sampling

Before re-sampling


Channel Distortion

Before 16-QAM After

Transmitter Radio Channel

RF Receiver Equalizer Decision Maker

An adaptive filter (lower figure) is used to eliminate the multiple channel distortion in wireless communication. A distorted 16-QAM signal (left figure) is blindly equalized to give a correct output (right figure) before recognition process

Channel Distortion


Literature SearchLiterature Search

hA web site search of �modulation recognition� gave 52,500 hits.

hMore than a hundred publications on modulation recognition.

hMany GOTS and COTS products available for various applications.


Some Well-known Approaches

Modulation feature extractionh I-Q AnalysishZero-crossing hPower-lawModulation classificationhPattern recognitionhMaximum likelihood


Problems in Algorithm Comparison

h Deferent modulation typeh Deferent assumptionh Deferent signal environmenth Deferent specification

PSK

PSK

FSK

PSK2

PSK8

PSK4

FSK8

FSK4

FSK2

ASK4

ASK8

ASK2

32QAM

CW

UNKNOWN

64QAM

16QAM

MSK


Azzouz and Nandi�s Method

Based on signal parameter variances and variance thresholds

CW PSK2PSK4ASK2ASK4FSK2FSK4

Base Band Signal

reco

gniti

on

tree

thresholds

xxix −)( s.t.d

.

abs

frequency phase

amplitude


Phase Variance is affected by the Center Frequency Offset

Phase Variance is affected by the Center Frequency Offset

BPSK with residual carrier after phase tracking and compensation

unwrapped phase trackingphase tracking


Phase Variance is Affected by Pulse Shaping (π/4 DQPSK)

Phase Variance is Affected by Pulse Shaping (π/4 DQPSK)


Phase Variance is Affected by Phase Wrap


Limitation in Azzouz and NandiLimitation in Azzouz and Nandi

hNot robust to center frequency offsethNot robust to pulse shapinghNot robust to phase wraphThresholds depend on SNR level hRestricted in modulation types (2 bits)


Liedtke�s MethodLiedtke�s Method

Based on delta-phase and histogram correlation

I

/

( )2

( )2

+

tan-1

amplitude

IF

sqrt

timing circuit

Q

t∆∆2φ

T∆∆φ

Freq.

delta phase

BPFbaud rate detector

hist

ogra

ms

CW PSK2 PSK4 PSK8 ASK2 FSK2

templates

reco

gniti

on tr

ee

s.t.d


Limitation in Liedtke�s Method Limitation in Liedtke�s Method

hManual tuning of center frequencyhManual tuning of the time-recovery band

pass filterhLimited in modulation types


Zero-crossing Detection Method

Based on zero-crossing phase, and delta phase histogram

IF

s.t.d.

delta-phase histogram

frequency estimation

phase estimation

histogram templates

zero-crossing counter

reco

gniti

on tr

ee PSK2 PSK4 PSK8 FSK2 FSK4 FSK8

templates


Limitation in Zero-Crossing Method

hNot robust to multiple frequencieshNeed high SNRhNeed high sampling rate


Power-Law Classification Method

Based on power spectrum analysis

signal ( )2

( )4

FFT peak detection

envelope s.t.d.

PSK2 PSK4 ASK2 FSK2

reco

gniti

on tr

ee


Limitation to Power-Law Method

hNeed very high sampling ratehAffected by pulse shapehLimited in modulation typeshNoise is amplified with high order


Modified Azzouz and NandiModified Azzouz and Nandi

CW PSK2PSK4PSK8ASK2ASK4FSK2FSK4

IF

I

/

( )2

( )2

+

tan-1

amplitude

down demodulation

sqrt

Q

frequency

phase


reco

gniti

on tr

ee

adaptive thresholds

xxix −)(

s.t.d.

abstiming

recovery

track and

correct phase

xxix −)(

s.t.d.

abs

xxix −)(

s.t.d.

abs

xxix −)(

s.t.d.

abs

X2

phasex2


Center Frequency Offset Correction

Before BPSK After

=

=

)()(

2

1

ximagxreal

xx

xsymbol x

Vxzz

z =

=

2

1

−

−=0110

)()( 22

2121 zzzzzG

2

)( 22

2121 zzzz −

=γ

VzGIVλγ

λ+

−=

1)(

Estimated phase

Method 1:

Center Frequency Offset Correction

The unwanted phase rotation (left figure) is corrected adaptively (right figure) by a modified blind carrier phase tracker (upper figure) so that the feature variation methods can be used robustly for modulation recognition

Method 2: CFO = angle( Σ(yky*k-1)fs/2/π k = 1, 2, �, N

yk is a sample at k fs is the sampling frequency


Modified LiedtkeModified Liedtke

I

/

( )2

( )2

+

tan-1

amplitude

down demodulation

sqrt

timing circuit

QT∆∆φ

frequency

delta phase

varia

nce

test

CW PSK2PSK4PSK8ASK2ASK4ASK8FSK2FSK4FSK8


peak detector

reco

gniti

on tr

ee

templates

IF


Summary

hThere are many publications but most of them are related to base band symbol recognition

hCenter frequency offset (CFO), pulse shape, timing, and channel fading can be estimated but the estimates will not be perfect

hSignal type UNKNOWN should be defined in all algorithms

hA good modulation recognition algorithm should be robust to CFO, timing error, and fading

comparison and simulation of digital … · comparison and simulation of digital modulation...

Documents