6 1. Introduction
But it is difficult to recover the information when the chaotic signal is corrupted
by noise. The second technique mitigates the effect of noise by making use of
an adaptive filter to demodulate the information. In the third approach, an RBF
neural network at the receiver is trained to continuously estimate the state of the
nonlinear dynamical system at the transmitting side. Based on the estimation re-
sults, the analog information is recovered. Unfortunately, existing analog modu-
lation schemes still do not possess sufficient noise immunity, hence limiting their
practical use in communication systems.
1.3.2 Digital Modulation
Compared with analog modulation schemes, digital modulation schemes are more
practical and quite a number of chaos-based digital modulation and demodula-
tion techniques have been proposed [Schimming and Hasler, 2000; Jáko, 1998;
Kennedyet al., 1998; Kennedy and Kolumbán, 2000; Kiset al., 1998; Kolumbán,
1998; Kolumbán, 2002; Sushchiket al., 2000]. Typically, in a chaos-based digital
communication system, the digital symbols are mapped to nonperiodic chaotic
basis functions. For instance, inchaos shift keying(CSK), different symbols are
mapped onto different chaotic attractors which are produced either by a dynami-
cal system for different values of a bifurcation parameter or by a set of completely
different dynamical systems [Kennedy, 1998; Kennedy and Kolumbán, 2000; Kis
et al., 1998].
A number of coherent systems have been suggested and studied [Dedieuet al.,
1993; Kocarevet al., 1992; Kolumbán, 2000a; Kolumbánet al., 1998]. For a co-
herent system, an exact replica of the chaotic signal needs to be reproduced at the
receiving end. Because robust synchronization techniques are not yet available,
coherent systems are still not realizable in a practical environment.
Noncoherent communication schemes, which do not require the reproduction
of the chaotic signals at the receiving end, are more feasible in practice. The
first noncoherent chaos-based digital communication scheme, namelydifferential
chaos-shift-keying(DCSK) scheme, was proposed by Kolumbánet al.[1996]. In
DCSK, a reference chaotic signal is sent, followed by the same signal modulated
by a binary symbol. At the receiver, these two pieces of chaotic signals are cor-
related. The binary symbol is then decoded based on the sign of the correlator
output [Kolumbán and Kennedy 2000; Kolumbánet al., 1998; Kolumbánet al.,
2002b]. Correlation delay shift keying (CDSK) is another noncoherent modula-
tion scheme in which a reference signal is transmitted to the receiver [Sushchiket
al., 2000]. Unlike DCSK, the reference chaotic signal is added to the information-
bearing signal with a certain time delay. In consequence, the bandwidth require-
ment is lower but the performance is degraded compared with the DCSK scheme.
Another noncoherent detection technique, which is applicable to CSK modula-
tion scheme, has been proposed by Hasler and Schimming [2000]. The technique