Research Of Signal Separation In MIMO Multimode Fiber Communication Based On ICA

Wireless MIMO(Multiple-Input Multiple-Output) communication is familiar to us. As we all know, it can significantly improve the capacity of the system. By comparison, optical MIMO communication creates many parallel channels in one multimode fiber. When applied to short distance communication net which uses multimode fiber as physical layer, it can significantly improve the capacity of the system, too. Of course, this benefit is not without a price, what the photo receivers have received are different signals, and anyone of these signals is mixed by all the source signals transmitted. Then, the critical question is how to recover the source signals from the received signals.ICA(Independent Component Analysis) is a new data processing method developed in recent years. Its purpose is to recover the statistically independent source signals from the observed mixed signals without any prior knowledge. It is in fact an optimization problem, which has not only one solution. So what we can do is only search an approximate solution to make the criterion which is a certain measurement of the independence of the separated signals to realize its optimization. Since ICA can be thought of a kind of calculation whose purpose is to search the solution to realize optimization of the chosen independence criterion, it actually consists of two parts. The first one is which independence criterion to choose. The second one is what method to use in order to realize its optimization. In other words, ICA equals independence criterion and optimization method.This paper uses mutual information as the independence criterion. When the mutual information of the various components of the output reaches its minimum point, ideally it is 0, the components are independent. In this paper, the natural gradient method has been used to achieve this goal. Since this method can only separate linear mixed signals, this paper quoted two demonstrators that take the channel matrix of the system as a linear matrix to prove that this method can be used in MIMO multi-mode optical fiber communication system With this assumption, the method studied in our paper based on mutual information of the output signals can recover the source signals successfully. In order to estimate the separation performance of the method when it is used in the MIMO multi-mode communication system, firstly, we use the channel matrix given by the demonstrators done by researchers of Eindhoven University to theoretically analyze the performance of the algorithm. Secondly, we have designed an experiment of a 2 * 2 MIMO multi-mode optical fiber communication system based on MGDM , and we analyzed its performance applied to separate the data received by the detector in the experiment.