(Apr 9) Exploring and learning structure in signal processing

This seminar discusses recent research directions of signal processing to understand and utilize signal structure. In order to obtain desired information from the acquired signal or to perform a desired operation, the signal structure should be effectively modeled and an optimum signal processing algorithm should also be designed based on the model. In the field of signal processing for the past two decades, research has been actively conducted on the compression sensing that designs an efficient signal processing system using the signal sparsity structure. In recent years, compression sensing techniques have developed to utilize various signal structures in addition to the rare structures of signals. This seminar introduces recent results on compression sensing and introduces the application results to 5G millimeter band communication and large capacity MIMO communication technology. On the other hand, in order to deal with signals having a complicated structure which is difficult to be expressed by a conventional signal model, or when a large amount of computation is required for signal processing, research on signal processing technology utilizing a machine learning technique has been actively conducted recently. Deep-learning technology, which is getting a lot of attention these days, can be effectively applied to understand the surrounding environment and context, and to model complex signal structures to design efficient signal processing systems. This seminar presents the basic research directions of machine learning technology for high performance signal processing and introduces recent research results on communication field and autonomous navigation recognition field.