COMPSENSE

COMPRESSIVE DATA ACQUISITION AND PROCESSING TECHNIQUES FOR SENSING APPLICATIONS

 Partecipanti

Mappa

 Word cloud

Esplora la "nuvola delle parole (Word Cloud) per avere un'idea di massima del progetto.

 Obiettivo del progetto (Objective)

'This is a proposal to support Dr. Ali Cafer Gurbuz’s reintegration as a full-time faculty member at TOBB University of Economics and Technology (ETU) in January 2009, leaving his post-doctoral position at the Georgia Institute of Technology, USA. Dr. Gurbuz’s expertise is in the areas of signal, image and array processing, with emphasis on radar, subsurface imaging, and compressive sensing. The technical objective of the proposed research is to develop novel data acquisition and processing techniques for newly emerging compressive sensing systems. Dr. Gurbuz brings unique experience to ETU as he has collaborated with one of the pioneers of compressive sensing, Dr. Romberg, and has an invention disclosure in the area of compressive wireless sensor networks. The three-year research plan outlined in this proposal builds upon his expertise acquired over his last six years in the US and continues his successful collaboration with researchers in the US while also ensuring transfer of that knowledge to potential collaborators in Europe. Our research plan is comprised of three sub-thrusts: 1) compressive remote sensing, with applications to subsurface imaging using GPR and EMI sensors; 2) multi-sensor and distributed compressive sensing; and 3) detection and estimation algorithms based directly on compressive data. These are timely research projects since compressive sensing is a newly emerging technology, especially in the area of remote sensing, where there are many unexplored areas of research and since initial results have shown its potential in a variety of applications. ETU is in a good position to lead his project as one of the fastest growing universities in Turkey with modern facilities and a dynamic faculty. ETU is very confident of Dr. Gurbuz’s potential to become one of the leading researchers in Europe in his field and is ready to do its part for his successful reintegration. But support from this grant is extremely critical to achieve success.'

Introduzione (Teaser)

Scientists developed novel techniques for remote sensing in a bid to acquire big data volumes and process them efficiently.

Descrizione progetto (Article)

Compressed sensing (CS) is a new approach to digital data acquisition. Analogue signals are digitised for processing not through uniform sampling, but through measurements using more general test functions. The EU-funded project 'Compressive data acquisition and processing techniques for sensing applications' (http://acgurbuz.etu.edu.tr/publications.html (COMPSENSE)) developed novel data acquisition and imaging techniques for emerging CS systems.

With surface reflections being a major problem in CS sub-surface imaging, scientists provided a simple method using compressed measurements for non-planar surfaces. This imaging method proved to be more reliable and could find shallow targets by using removed data from surface reflection.

The project presented a novel sparse ground-penetrating radar imaging method that is very effective for off-grid targets. The proposed technique was based on an iterative orthogonal matching pursuit (OMP) algorithm that uses gradient-based steepest ascent-type iterations to locate the target.

Compared to standard sparse reconstruction techniques, more reliable results were obtained for multiple off-grid targets and with substantially fewer reconstruction errors. In addition, a novel perturbed OMP algorithm performed controlled perturbation of selected support vectors to decrease the orthogonal residual at each iteration.

Regarding multi-sensor and distributed compressive sensing, the project team developed another new technique for estimating direction of arrival based on compressed measurements. Other methods for detection and estimation involved direct acquisition of CS data, extracting features or generating data without first reconstructing the signal.

Work was also geared toward analysing CS energy efficiency on wireless sensor networks (WSNs). Energy dissipation models for both CS and conventional approaches were built and used to construct a mixed integer programming framework that jointly captured energy costs. Numerical analysis showed that CS prolonged network lifetime for sparse signals and was more advantageous for WSNs with a small coverage area.

The project team posted several articles in journals and conference proceedings, in addition to giving presentations at workshops or seminars. A list of publications can be found on the project website.