Research Of Modulation Classification Based On Duffing Oscillators Array

Modulation recognition has always been an important part of the process ofcommunication. No matter in the areas of military or in the areas of civilian areas orareas modulation recognition is of great significance. In today’s increasinglycomplex communications environment, modulation recognition is also facing somenew problems. For instance, in the non-partner communication of military field, theuseful signal is often transported in a strong background noise. And in addition, therecipient or the side who intercepts the transmitted signal will is know little aboutthe various parameters. These above require modulation recognition to work in alow SNR and scarce circumstance, with a little priori knowledge.On the other hand, the rapid development of nonlinear science, as to Chaostheory, stochastic resonance theory and so on, has brought fresh blood into thetechnology of modulation recognition. The researchers achieved some impressiveresults. Duffing chaotic system for example, has been demonstrated useful in weaksignal detection. Accordingly, this paper attempts to use the Duffing Chaos Theoryapply into modulation recognition method, and therefore discuss the possibility ofcombining engineering applications of modulation recognition and chaos theory.Firstly, this paper briefly describes the relevant knowledge of chaos theory,focusing on analysis through the Melnikov method about the dynamics of Duffingoscillator, which will work as the theoretical basis of this modulation recognition.Then we focuses on the influence of state of Duffing system when the externalsignal is involved, thereby designed a Duffing oscillator arrays pretreatment systemto distinguish different kinds of modulation. After that we propose a method basedon the power spectral entropy to quickly and efficiently determine the Duffingsystem status and make it work as the characteristic parameter. Finally combiningthe carrier estimation method based on the principle of intermittent chaotic, BPneural network classifiers, and Duffing pretreatment system and feature extraction,we construct a complete binary modulation recognition system. The simulationresults show that this method can effectively identify a binary signal modulationmethod in the low SNR conditions, and does not rely on a priori knowledge of thechannel advantages.