On-line adaptive chaotic demodulator based on radial-basis-function neural networks

Abstract

Chaotic modulation is a useful technique for spread spectrum communication. In this paper, an on-line adaptive chaotic demodulator based on a radial-basis-function (RBF) neural network is proposed and designed. The demodulator is implemented by an on-line adaptive learning algorithm, which takes advantage of the good approximation capability of the RBF network and the tracking ability of the extended Kalman filter. It is demonstrated that, provided the modulating parameter varies slowly, spread spectrum signals contaminated by additive white Gaussian noise in a channel can be tracked in a time window, and the modulating parameter, which carries useful messages, can be estimated using the least-square fit. The Henon map is chosen as the chaos generator. Four test message signals, namely, square-wave, sine-wave, speech and image signals, are used to evaluate the performance. The results verify the ability of the demodulator in tracking the dynamics of the chaotic carrier as well as retrieving the message signal from a noisy channel.