In the next Machine Learning Group seminar, we have a talk by Dr. Nikos Deligiannis who is an Assistant Professor at the Department of Electronics and Informatics at Vrije Universiteit Brussel, Belgium.
Conventional data communication systems abide by a two-step approach, where first Shannon-Nyquist data acquisition is performed and then efficient data compression is applied. Compressed sensing (CS) is a new signal acquisition paradigm that offers the means to simultaneously perform sensing and compression. By recognising that signals are sparse or compressible in a given basis, CS enables signal acquisition using far less measurements than the classical Shannon-Nyquist acquisition scheme. CS has found many applications, including medical imaging, compressive image/video processing, and wireless sensor networks.
Inspired by the distributed source coding theory, which investigates the problem of data compression with decoder side information, we will address the problem of compressed sensing with side information at the reconstruction side. The side information, which is a signal similar to the one that is sensed, is integrated into CS via ℓ1-ℓ1 and ℓ1-ℓ2 minimization. We will provide lower bounds on the number of measurements that these problems require for successful reconstruction of the target signal. We will show that, provided that the side information has good quality, the number of measurements is significantly reduced via ℓ1-ℓ1 minimization compared to conventional CS.
We will present an application of our theoretical framework in low-power wireless medical imaging. Wireless capsule endoscopy in particular, which is essentially video capturing and transmission by a pill, can be a promising application domain. Furthermore, rooted in our theory, we will present an online algorithm for reconstructing time-varying signals from a limited number of linear measurements. The signals are sparse, with unknown support, and vary according to a generic nonlinear dynamical model. We will show the application of the algorithm to compressive video background subtraction: given a set of measurements of a sequence of images with a static background, we can simultaneously reconstruct each image and separate its foreground from the background. The presented method, which shares similarities with distributed video coding systems, will be shown to provide for a dramatic reduction in the number of measurements with respect to state-of-the-art compressive background subtraction schemes.
Nikos Deligiannis received the Diploma in electrical and computer engineering from University of Patras, Greece, in 2006, and the Ph.D. in applied sciences (awarded with highest distinction and congratulations from the jury members) from Vrije Universiteit Brussel, Belgium, in 2012. From June 2012 to February 2015, he held postdoctoral research positions in the Department of Electronics and Informatics at Vrije Universiteit Brussel and the Department of Electronic and Electrical Engineering at University College London. Since March 2015, he is an Assistant Professor at the Department of Electronics and Informatics at Vrije Universiteit Brussel, Belgium. His research interests include multiterminal communications, sparse signal processing and machine learning, information theory, wireless networks, and multimedia systems. Dr Deligiannis has received the 2011 ACM/IEEE International Conference on Distributed Smart Cameras Best Paper Award and the 2013 Scientific Prize IBM-FWO Belgium.
 J. F. C. Mota, N. Deligiannis, and M. R. D. Rodrigues, “Compressed sensing with side information: geometrical interpretation and performance bounds”, accepted to IEEE Global Conference on Signal and Information Processing, Symposium on Information Processing for Big Data, GlobalSIP’14, Atlanta, Georgia, USA, Dec. 2014.
 J. F. C. Mota, N. Deligiannis, and M. R. D. Rodrigues, “Compressed sensing with prior information: Optimal strategies and bounds”, submitted to IEEE Transactions on Information Theory, Aug. 2014 (available on Arxiv:
 J. F. C. Mota, N. Deligiannis, A. C. Sankaranarayanan, V. Cevher, and M. R. D. Rodrigues, “Adaptive-Rate Sparse Signal Reconstruction With Application in Compressive Background Subtraction”, to be submitted to IEEE Transactions on Signal Processing, 2014 (available on Arxiv: