Space division multiplexing in standard multimode optical fibers using speckle pattern recognition

Data-communication using optical fibers is an essential part of our current ICT society. Telecommunication using single-mode optical fibers forms the backbone of the internet, while huge amounts of data are send around in data-centers over multi-mode optical fibers. The explosive growth of data- and telecommunication demands at different link distances all ask for the development of techniques that offer ever higher data capacity. It is expected that existing techniques will soon not be enough to meet this challenging demand.

In order to solve this capacity crunch, there is a huge interest in further boosting the transmission capacity by using other, not yet fully exploited approaches: the data capacity can be further increased by sending data in parallel from sender to receiver. In its simplest form, one can increase the transmission capacity by installing in parallel multiple optical fibers. However, this is not a very cost-effective technique as the required number of network components just scales linearly with
the number of parallel fibers. Therefore, there has been extensive research in the past few years on how to send bits in parallel through a single fiber. We propose to do this in a novel way, such that conventional fibers and fiber components can still be used. The detection of the transmitted data than becomes a challenging computational task, which we will tackle using modern developments from the field of deep-learning in neural networks.