Novel approaches to signal transmission based on chaotic signals and artificial neural networks

A novel chaotic-based coding/decoding strategy that exploits radial basis function (RBF) artificial neural networks (ANNs) in a dynamic feedback (DF) configuration is reported. The ANNs are used as pseudochaotic carrier generators and as estimators for the received signal. The dynamics approximated were those of the logistic map (LM). This approach is compared with established methods that employ inversion, dynamic feedback, and least mean square (LMS) and recursive least squares (RLS) estimation. Our RBF-ANN-DF approach is shown to outperform these methods in terms of the recovered signal SNR at various channel SNRs with a speech information signal used as an example. In particular, the RBF-ANN-DF method is shown to outperform DF approaches by about 33 dB at all channel SNRs. Moreover, the proposed RBF-ANN-DF approach offers a recovered signal SNR improvement between about 15.1 and 27.4 dB for channel SNRs between 10 and 50 dB as compared to an LMS-based chaotic receiver. As a by-product, we have also shown that, for the logistic map, LMS- and RLS-based chaotic receivers are equivalent and, hence, the use of LMS-based receivers can result in implementation savings