Sie sind auf Seite 1von 4

IEEE Int. Symp.

on Circuits and Systems (IEEE-ISCAS05), Kobe, Japan, May 2005

A Complete Receiver Solution for a Chaotic Direct Sequence


Spread Spectrum Communication System
M. B. Luca1,2 , S. Azou1 , G. Burel1 , Member, IEEE and A. Serbanescu2 , Member, IEEE
1

Laboratoire dElectronique et Systmes de Tlcommunications (UMR CNRS 6165), Brest, France


2
Military Technical Academy, Bucharest, Romania

Abstract This paper is devoted to receiver design in a Chaotic


Direct-Sequence Spread Spectrum (CD3S) digital communication system. The demodulation is achieved through chaos synchronization in
an efficient manner thanks to dual Unscented Kalman Filtering. The
problem of carrier phase recovery, frequently neglected in chaos-based
communication systems, is addressed either through Costas loop prior
to demodulation or by dual estimation of the code, symbol and phase in
baseband. The input signal power fluctuations, which often causes large
errors in synchronizing chaotic waveforms, is also taken into account.
Numerical simulations are provided to show the significance of the
proposed receiver for both Additive White Gaussian Noise (AWGN)
channel and a nonstationary channel.

I. I NTRODUCTION
Over the last decade, a great research effort has been devoted
towards the development of efficient chaos-based modulation techniques (see [2]-[4]). This motivation originates from theoretical
results about the synchronizing capability of two identical chaotic
systems that start from different initial conditions [1]. Due to
its random-like behavior, chaos not only spreads the spectrum of
the information signal, thus providing robustness against channel
distorsions, but also acts as an encryption key. Hence, covertness of
transmissions can be ensured and due to intricate dynamics of the
signals, it is extremely difficult for the unauthorized user aware of
the transmission to access the information. Other potential benefits
have to be noticed, among others the sharing of channel resources
via Code Division Multiple Access (CDMA), resulting from weak
crosscorrelation of chaotic signals, and reduced complexity of
transmission devices.
Use of chaotic signals as spreading codes in a Direct-Sequence
Spread Spectrum system (DS-SS) has been shown to be a promising
way of applying chaos for digital communication purposes [5].
Use of chaos in a DS-SS system consists in replacing the standard
Pseudo-Noise (PN) generator by a chaotic dynamical system. One
motivation for such a chaotic modulation is that a growing number
of communication applications rely on the DS-SS scheme to achieve
CDMA. DS-SS systems also offers a good immunity to channel
impairments through processing gain adjustment.
This paper focuses on receiver design in a Chaotic DirectSequence Spread Spectrum (CD3S) system, the objective being to
bring a complete solution for practical applications, while using
the original chaos synchronization property underscored in [1]. In
particular, we address the problem of carrier phase recovery, which
is usually not considered in the related literature. Leung and Zhu
have recently derived important results about chaos synchronization
through Extended Kalman Filtering (EKF); the authors showed that
the EKF-based technique is a generalization of two conventionnal
schemes (unidirectionnally coupled and drive-response methods). It
is also shown that the EKF-based synchronization approaches the
averaged Cramr-Rao Lower Bound at high Signal-to-Noise-Ratios

(SNR). For practical applications, however, it is better to employ a


more robust nonlinear filter, due to important noise levels, multipath
propagation and channel fluctuations. It is proposed here to make
use of Unscented Kalman Filters (UKF) instead of EKFs to perform
the synchronization. The UKF, recently proposed by Julier et al. in
the context of nonlinear control [7], addresses the approximation
issues of the EKF at a comparable computationnal complexity.
The state distribution is again represented by a Gaussian random
variable, but is specified using a minimal set of carefully chosen
sample points. We first propose an UKF-based estimation scheme to
retrieve the symbol and spreading code simultaneously, the problem
of carrier phase recovery being solved in a previous step through
Costas loop. This dual estimation approach is then generalized to
recover the carrier phase at the same time, hence achieving the
whole receiving process in baseband. The automatic gain control
is also taken into account. Bit-Error-Rate (BER) simulations are
derived to show the significance of the proposed receivers.
The paper is organized as follows. The principle of the CD3S
transmitter is described in section II. Then, the problem of chaos
synchronization via nonlinear Kalman filtering is discussed in
section III. Section IV summarizes the dual Kalman estimation
scheme and presents the solutions for carrier recovery. Finally, BER
performances of the proposed receivers are exposed in section V.
II. T HE C HAOTIC DS-SS T RANSMITTER
The general scheme of a CD3S transmitter is shown in figure
(1). The information bits are first modulated, through BPSK (Binary
Phase Shift Keying) in our case, to get the symbols bk that are then
spreaded using the chaotic code ck R evolving at the chip rate
Fc  Fb , where Fb = 1/Tb is the symbol rate. At this point we
make no assumption on the type of chaotic sequence, only that it is
given by the recursive monodimensional equation ck = f (ck1 ).

Fig. 1.

General scheme of a chaotic DS-SS system

The spread spectrum signal will then take the form xk =


bbk/Lc ck , where bc denotes the integer part of the enclosed
number and where L = Fc /Fb stands for the processing gain.
The choice of the processing gain depends upon the available
channel bandwidth, the desired data rate and bit error rate together
with the existence of any covertness constraint. The signal is finally
passed through a square root band limiting Nyquist filter and before
the transmission through the communication channel, a sinusoidal
carrier modulation is done.

IEEE Int. Symp. on Circuits and Systems (IEEE-ISCAS05), Kobe, Japan, May 2005

III. C HAOTIC S YNCHRONIZATION USING S TATE E STIMATION


M ETHODS
Leung and Zhu showed in [6] that synchronizing chaotic systems
through state estimation was a generalization of some classical synchronization methods, namely the unidirectionally-coupled method
and the drive-response synchronization method. We now consider
the problem of chaos synchronization in the same way as in [6],
using the following model:
xk+1 = f (xk ) + vk , yk = xk + nk
(1)
where f () stands for the nonlinear recursive function that defines
chaotic motion and where the term nk N (0, R) in the observation model is the noise that affects the chaotic signal, as it is passed
from the transmitter to the receiver. We have also introduced the
process noise vk N (0, Q), independent of the past and current
state xk , that reflects any model uncertainty.
Due to the high nonlinearity of the model, the Kalman Filter
(KF), minimizing the Mean Square Error (MSE) in the linear case,
is no longer applicable; the popular solution is then to employ
the Extended Kalman Filter (EKF), as is the case in [6]. But we
know that the main drawback of this filter is the model first order
linearization, which can conduct to significant errors for the means
and covariances used by the KF algorithm. To limit these errors
we have considered the Unscented Kalman Filter (UKF), recently
developped by Julier et al. in the context of nonlinear control [7].
This filter relies on the Unscented Transform (UT) which gives the
statistics of any random variable that has undergone a nonlinear
transformation. Instead of linearizing the nonlinear model, as in
the EKF approach, the UT leaves the system functions intact and
approximates the probability density function using a set of sigma
points; each of these samples is then passed through the exact
nonlinear function to compute the resulting mean and covariance. In
comparison with the first order linearization which can provide an
exact estimate to the first order term in the Taylor series expansion,
the Unscented approach can give an exact estimate up to the third
order terms in the Taylor series of the transformed distribution.

Fig. 2.

Synchronization performance MSE/R as a function of R

To analyze how the estimation of the Taylor series terms affects


the synchronization, we will evaluate the synchronization MSE
normalized to the observation noise variance, for the case of a
nonlinear function given by the zero mean logistic map f (xk+1 ) =
1 2x2k . The use of a second order nonlinearity assures us that the
Unscented method yields the exact statistics (mean and variance)
of the transformed distribution. Numerical simulations were made
using sequences of 104 samples length, and calculating the MSE
after a transient sequence of 500 samples. The observation noise

variance was considered in the interval 106 , 101 . Using the


results presented in figure (2), we can say that for small observation
noise variances, the two estimation methods achieve almost the same
MSE, but for larger ones (R & 102 ) the influence in the statistics
of higher moments increases, and so the EKF fails to offer a good
estimation.
In conclusion we can say that for a more realistic communication
channel, with usually large observation noise and particularly for
spread spectrum systems, the Unscented method will have better
performances. This is why for the rest of the article we will employ
only the Unscented method.
IV. S OLUTIONS FOR C ARRIER P HASE R ECOVERY
As suggested recently [8], the chaotic synchronization and the
symbol detection can be achieved at the same time thanks to parallel
Kalman filtering (Code/Symbol dual estimation). The carrier phase
recovery is usually considered as a solved problem, the analysis
being made for the baseband signals. For those papers where this
problem is not totally neglected, usually it is employed the well
imposed Costas loop. The methods for the phase recovery presented
in this section will consider the Costas loop as the departing point.
A. Carrier recovery prior to dual estimation of the code and symbol
As mentionned above, the proposed receiver structure relies on
dual Kalman filtering. As a part of the state model employed, the
non linearity submitted by the chaotic signal plays a major role, as
the information is not retrieved by a classical correlator but by a
chaotic synchronization method. The general scheme of such a dual
Kalman filter receiver is given in figure (3).

Fig. 3.

Code/Symbol dual estimation-based CD3S receiver

In what follows, no complex values will be considered at the dual


Kalman filtering level. This means that a separate carrier recovery
block (e.g. Costas Phase Locked Loop) is needed to compensate
any carrier phase deviation. As shown by figure 4, the dual Kalman
filtering scheme enables the estimation, at chip rate Fc , of the
original chaotic spreading code ck together with the data symbol bk
from noisy observations yk . Each of the two filters uses last estimate
of the other as a parameter, the general model being presented in
equation (2). The dynamical model and the observation model used
for chaotic code estimation take the following form:
ck+1 = sgn(bk1 )f (ck ) + vkc , yk = sgn(bk1 )ck + nk (2)
where the noise sequence vkc N (0, Qc ), independent of the past
and current state ck , reflects the model uncertainty due to channel
imperfections; the observation noise nk N (0, R) will be mainly
dependent upon the SNR at the receiver input.

Fig. 4.

Code/Symbol Dual Estimation Block

IEEE Int. Symp. on Circuits and Systems (IEEE-ISCAS05), Kobe, Japan, May 2005

Similarly, the symbol will be estimated at chip rate through the


following dynamical and observation models:
ck1 ) + nk
(3)
bk+1 = bk + vkb , yk = bk f (

`
where the gaussian noise sequence vkb N 0, Qb , independent
of the past and current state bk , will influence the adaptability of the
symbol filter; a low value Qb will result in slow changes whereas a
larger value will result in rapid variations of the symbol estimates.
It must be emphasized that thanks to the set of noise terms {vkc ,
vkb , nk } and due to use of the Unscented implementation this dual
Kalman receiver can still operate on frequency selective channels
[9], whereas the multipath propagation is not explicitely taken into
account in equations (2) and (3).
B. Dual estimation of the Code, Symbol and Phase
Opposed to the previous receiver structure, relying on a Costas
loop to solve the phase recovery prior to demodulation, we are
now going to develop another solution considering a possible phase
error in the recovered baseband signal. As shown in figure 5, the
frequency transposition is considered with a fixed carrier to get the
in-phase and quadrature signals.

Fig. 5.

yk , nk being the observation state and noise vectors and vk


stands for process variance vector
iT
h
T

(7)
, nk = n1k , n2k , vk = vk1 , vk2
yk = ykRe , ykIm
As considered for the C/S dual estimator, the noises are supposed
zero mean gaussian distributed and independent with the other
parameters that construct the model:

vkc N
(0, Qc ) , vkb N 0, Qb
`
(8)
vk N 0, Q ; Q = diag Q1 , Q2
For the observation noise we have considered nk N (0, R),
2
with the covariance matrix R = 2 I2 .
To illustrate how the phase recovery works we have considered
the case of a demodulation with a carrier frequency offset and
a nonstationary channel which suffers from a periodical phase
alteration. This simulation was made for a Carrier-to-Noise-Ratio
(CNR) of 20dB at the receiver (Fig. 7).

Code/Symbol/Phase dual estimation-based CD3S receiver

Fig. 7. Representaion of the estimated phase parameters in case of a


frequency error on a nonstationary channel

Fig. 6.

Code/Symbol/Phase dual estimation block

We propose this receiver as a generalization of the previous C/S


dual demodulator, adding a model which will make the estimation
of the phase error using the code and informational signal as
parameters. We consider a second order model to express the phase
error. Another important aspect is that this time the observation
vector will consider also the quadrature signal to achieve the error
phase estimation, even if it is employed a biphasic modulation.
The models used to estimate the code, the symbol and the phase
are given in equations 4, 5 and 6, respectively.
8

< ck+1 = sgn bk1 f (ck ) + vkc

(4)
: yk = sgn bk1 ck g k1 + k1 + nk
(
bk+1 = bk + vkb

(5)
yk = bk .f (
ck1 ) .g k1 + k1 + nk
8

k+1
k + k
>
<
=
bk + vk
k+1
(6)
k
>
: yk = sgn bk1 f (
ck1 ) g (k ) + nk
where g (k ) = [cos (k ) , sin (k )]T ;

We recall that only the UKF was used for these simulations.
We add also that for a lower CN R case, with a more difficult
channel, estimated phase jumps of can occur, which for a
normal phase modulation will conduct in errors at the retrieved
informational signal. So differential phase modulation (DPSK) was
employed instead.
The performances of the C/S/P dual receiver, as for the C/S dual
estimation scheme, will now be presented in the next section.
V. N UMERICAL RESULTS
For a numerical communication system the best performance
criteria is given by the BER vs. a parameter of study. In our case,
as we consider a single user transmission, the parameters of interest
will be mean CNR and the considered channel characteristics. Robustness of the proposed receivers against noise and some channel
imperfections have been studied through Monte Carlo simulations,
our objective being to achieve a BER of about 103 .
A zero-mean logistic map (see section III) is used as it generates
sequences with a Dirac like autocorrelation function, and so qualifies
very well as a spreading sequence. Receivers simulated were both
C/S and C/S/P structures for the following channels:
1. stationary AWGN channel (no phase or transfer coefficient shift);
2. nonstationary gaussian channel with variations of phase and
of the channel transfer coefficient; As phase variations we have
considered the equivalent of a sinusoidal instantaneous carrier
frequency shifting with a 20Hz amplitude, and a period of 0.1s.

IEEE Int. Symp. on Circuits and Systems (IEEE-ISCAS05), Kobe, Japan, May 2005

The results are given with or without any gain control method
each time to increase the pertinence of our numerical simulations.
We showed in a companion paper that the input power fluctuations
yield significant performance degradations for the receivers that
ignore this problem while using chaos synchronization [10]. The
first approach that we proposed (the modified dynamics approach)
is based on a modification of the chaotic dynamics that has been
used by the transmitter; the second approach (denoted control loop)
relies on a feedback control in order to match the second order
statistics of the transmitted chaotic spreading code.

Fig. 8. BER performance of the C/S dual estimation-based receiver over


AWGN channel

For the nonstationary channel we can observe that the performances of the two systems are coming closer, but the C/S scheme
stands as the best method. In concordance with the results presented
in [10], the statistical gain control loop remains the best automatic
gain control (AGC) solution. Again as the number of joint state
estimation parameters increases (C/S/P case), the performances of
the model dependent AGC decreases.
VI. C ONCLUSIONS
The design of a Chaotic Direct Sequence Spread Spectrum
(CD3S) receiver has been studied in this paper. Thanks to dual
Unscented Kalman Filtering, the chaotic spreading code is synchronized in an efficient way while estimating the symbol at the same
time. Other advantages of the UKF method is its recursivity and its
limited computational cost, which make practical implementation
simpler. The role of carrier phase recovery, usually neglected in
the related literature, have been solved in two ways: the first
approach considers the phase recovery and Code/Symbol estimation
as a separate step and the second method solves this problem in
baseband using three parallel Kalman filters (Code/Symbol/Phase
dual estimation). The gain control is also addressed through either
chaotic dynamics modification at the receiver or statistical matching
between original and synchronized code. The proposed receivers
have been evaluated through Monte Carlo simulations of the BER
over a stationary Gaussian channel and in presence of channel gain
and phase fluctuations. Without any channel fluctuation, the solution
of carrier phase recovery prior to C/S estimation is better than
C/S/P dual estimation in baseband. The full baseband demodulation
becomes an interesting alternative in case of nonstationary channels.
R EFERENCES

Fig. 9. BER performance of the C/S/P dual estimation-based receiver over


AWGN channel

Fig. 10. BER performance of the two receivers over a nonstationary channel

The results obtained by the simulations are shown by figures 8, 9


and 10. We represent the C/S and C/S/P demodulators on separate
figures, as we observe a high difference in BER performances. In
particular the C/S structure has presented no errors for the first
channel type in the CNR interval [5dB, 12dB]. So we can definetly
say that this structure overpasses the C/S/P one. We observe also
no difference in performances for the C/S systems with both gain
control loops on the stationary channel.

[1] L. Pecora and T. Caroll, Synchronization in chaotic systems, Phys.


Rev. Lett., vol. 64, pp. 821-823, 1990.
[2] M. Hasler, Synchronization of chaotic systems and transmission of
information, Int. J. Bifurcation and Chaos, vol. 8, no. 4, pp. 647-659,
1998.
[3] G. Kolumban, M. P. Kennedy, and L. O. Chua, The role of synchronization in digital communication using chaos - Part II : Chaotic
modulation and chaotic synchronization, IEEE Trans. Circuits Syst.,
vol. 45, no 11, 1998.
[4] T. Yang, A Survey of Chaotic Secure Communication
Systems, International Journal of Computational Cognition
(http://www.YangSky.com/yangijcc.htm), vol. 2, no. 2, 2004.
[5] J. Kraus, J. A. Nossek, T. Yang, and L. O. Chua, Evaluation of a
continuous valued chaotic spreader used in a Chaotic Digital CodeDivision Multiple Access ((CD)2 MA) system, Int. J. Bifurcation and
Chaos, vol. 10, no. 8, pp. 1933-1950, 2000.
[6] H. Leung and Z. Zhu, Performance Evaluation of EKF-Based Chaotic
Synchronization, IEEE Trans. Circuits Syst. I, vol. 48, pp. 1118-1125,
2001.
[7] S. Julier, J. Uhlmann, and H. F. Durrant-Whyte, A new method for
the nonlinear transformation of means and covariances in filters and
estimators, IEEE Trans. Automat. Contr., vol. 45, no. 3, pp. 477-482,
2000.
[8] S. Azou and G. Burel, Design of a demodulator in a chaos-based
spread spectrum communication system using dual Unscented Kalman
Filters, IEEE-Communications, Dec. 5-7, 2002, Bucharest, Romania.
[9] S. Azou, C. Pistre, and G. Burel, A chaotic direct sequence spreadspectrum system for underwater communication, IEEE-Oceans02,
Biloxi, MS, USA, Oct. 2002.
[10] S. Azou, M. B. Luca, G. Burel, and A. Serbanescu The problem of
gain control in a Kalman filter based synchronization chaotic receiver,
IEEE-Communications, June 3-5, 2004, Bucharest, Romania.

Das könnte Ihnen auch gefallen