Modelling aggregate heterogeneous ATM sources using neural networks

The simulated waveforms are shown in Fig. 2, while the corresponding experimentally observed waveforms are shown in Fig. 3a-c. The quality of the experimental synchronisation can be appreciated looking at Fig. 36 that shows the variable x, against the variable x5, each one corresponding to the other in the master and slave systems. Finally, Fig. 3c shows the observed phase portrait in the x,-x, plane when the forcing signal is applied. In Fig. 3a, the presence of a small ripple superimposed onto the decoded signal can be observed. This could lead to the wrong conclusion that the two circuits are not exactly synchronised; however, Fig. 3b excludes this case. The correct explanation of this fact is immediately obtained by measuring the actual current fed by the block B3. In fact, this current is slightly corrupted by the master itself due to the nonideal features of the realised current source generator. Apart from the triangular wave, other different waveforms have also been taken into account (e.g. sine waves, square waves, speech and musical signals, and so on), with successful results. Finally, it has been noted that the above cited ripple is always present independently from the nature of the considered transmitted signals. Therefore it could be significantly reduced by using a better current source or by filtering the decoded signal by means of a lowpass filter. Conclusions: In this Letter it has been shown how two SC-CNNbased circuits, generating the dynamics of the Chua oscillator, are used successfully for signal transmission. The SC-CNN approach for generating complex dynamics is quite general; in fact, simply changing the templates of a fixed structure system, it is possible to choose a different chaotic attractor. Alternatively, the architecture can be easily augmented by adding new cells in order to generate higher order dynamics [6]. Even if, in the reported case, the inverse system method has been considered, the modularity of the SCCNN approach allows an easy implementation of other synchronisation methods [l, 21. In this way, we stress the concept that the SC-CNN framework represents a suitable platform for chaotic system synchronisation with potential applications in secure communication systems.