About SPM e-Newsletter

Since April 2007, the IEEE Signal Processing Magazine has introduced a new form of publication - a Monthly Electronic Newsletter. The e-Newsletter will complement the bi-monthly Magazine to serve the members in the IEEE Signal Processing Society (SPS).  Through email notification and expanded coverage on its website, the e-Newsletter will provide members with timely updates on:

• society and technical committee news,

• conference and publication opportunities, new books, and Ph.D. theses,

• signal processing related research opportunities, and

• activities in industry consortiums, local chapters, and government programs.

The e-Newsletter is a gateway to reach out to signal processing professionals around the world.  We invite you to contribute and share your news with tens of thousands of SPS members through this monthly electronic publication with fast turn-around cycle. IEEE members may manage their subscription of the email notification of the eNews and related SPS announcements at this page.  Please bookmark <http://enews.ieee-spm.org> for current and archived issues of eNews.

Submission Instructions  - Contribution for the August-September '07 Issue Due August 15, 2007

Please contact the Associate Editors of the corresponding sections as listed below to provide your input or if you have questions. Make sure that you include your name, affiliation, and email and phone contact information. Contributions submitted by August 15, 2007 will be considered for inclusion in the next issue of the SPM e-Newsletter.  

Contact Information of the SPM e-Newsletter Team

  Min Wu, SPM Area Editor for e-Newsletter, University of Maryland, College Park, USA (minwu AT umd.edu)

  Huaiyu Dai, Associate Editor, North Carolina State University, Raleigh, USA (huaiyu_dai AT ncsu.edu)
     Conference and publication news

  Alessandro Piva, Associate Editor, University of Florence, Italy (piva AT lci.det.unifi.it)
     News and activities in local chapters and research groups (including new Ph.D. theses)

  Mihaela van der Schaar, Associate Editor, University of California, Los Angeles, USA (mihaela AT ee.ucla.edu)

     News and activities of SPS Technical Committees, industry consortiums and international standards

  Nitin Chandrachoodan, Digital Production Editor, Indian Institute of Technology – Madras (nitin AT ee.iitm.ac.in)
     Online submission and production system

  Shih-Fu Chang, SPM Editor-in-Chief, Columbia University, New York, USA (sfchang AT ee.columbia.edu)

  * Please replace "AT" in the email addresses with @.

Back to Top

In-Depth Articles of July 2007 SPM eNews

Reproducible Research in Signal Processing Contributors:  Patrick Vandewalle, Jelena Kovacevic, and Martin Vetterli

As described on Wikipedia, reproducibility is one of the basic principles of the scientific method. It refers to the ability of an experiment to be replicated by another researcher working independently. While such a definition is applicable to science in general, what does it mean for signal processing research specifically?
Should all the research presented in our journals be reproducible? For which types of signal processing research is reproducibility most important? What do we do when working on a project with industry? And what does reproducibility require: do the code and data have to be made accessible, or is it sufficient to describe an algorithm accurately in a paper?

One way of approaching the issue is the reproducible research setup developed by J. Claerbout
and D. Donoho in their labs at Stanford in the early nineties. With a publication, members of these labs also publish their code and data that were used to produce the results presented. This allows other researchers to easily reproduce results, and compare them to their algorithms using their own data.

At the latest ICASSP conference, a special session was held on this topic, and raised interesting
discussions. Papers were presented on issues ranging from standardized data sets, practical setups for reproducible research, publishing aspects, to case studies of actually reproducing certain results. We would be very interested in continuing these discussions online in a discussion forum. Here are some issues to discuss: Would you consider most of the current signal processing research as reproducible? Do you have personal experiences on making research reproducible? Does your lab have a particular setup for sharing code or data? Do you have a personal opinion about the current state of reproducibility in signal
processing research and the need (or the absence thereof) to take particular actions? We would be happy to hear your opinion on the discussion forum!
 

Return to New Initiatives and Trends

Signal Processing for Future Radio Telescopes Contributors:  Amir Leshem and Alle-Jan van der Veen

Radio astronomy forms an interesting application area for array signal processing techniques. Traditionally, radio telescope design was in the forefront of electrical engineering technology. Technological advances in the last decade have created possibilities for large distributed interferometric radio telescopes with very large receiving areas and a sensitivity which is one to two orders of magnitude better than the current generation. Increased sensitivity implies receiving more interfering signals, and therefore RFI detection and
removal is now an important topic in radio astronomy. Fortunately, massive digital phased array technology has also greatly advanced during this period and can provide increased flexibility to filter out
interference as well as he possibility of directing multiple beams simultaneously.

Several major international research groups are working on next generations of phased array instruments. The most ambitious one falls under the framework of the Square Kilometer Array (SKA) program, with a target commissioning date of around 2020. A second instrument which is a distributed phased array radio telescope is the Low Frequency Array (LOFAR) which is currently under construction in The Netherlands, and slated for 2008.

Prior to this, the Allen radio Telescope Array (ATA) being built in California is serving as a prototype for one of the SKA concepts. It is based on large number of small steerable dishes phased together to form several stations. In contrast, the LOFAR design calls for an instrument consisting of about 13,000 "simple" omni-directional antennas (10--240 MHz), grouped in about 70 stations spread in spirals over an area with a diameter of about 300 km, as well as in a more densely populated central core.

The 200 antennas in each remote station are used as a phased array and are combined in such a way that a beam is formed into a desired look-direction. The resulting output of each beamformer is similar to the output of a telescope dish pointing into the same direction, but is obtained without the use of any moving parts. LOFAR can be seen as a stepping stone for SKA, which should have an effective aperture area of one square kilometer, in the frequency range from 100 MHz up to 25 GHz. Just as as LOFAR, it will be a large distributed telescope with many individual elements. The telescope concept for SKA is not yet defined, but several designs are currently being worked out.

In terms of signal processing challenges for the ambitious design of these radio telescopes, we identify three main problems that should be solved in order to meet the design goals:

- RFI mitigation. The frequency bands of interest to radio astronomers contain many sources of RFI (radio frequency interference). RFI mitigation techniques will (necessarily) have to form an integral part of the system design. Interesting issues arise because of the hierarchy in the new generation of
telescopes: RFI mitigation is possible at the station level (beamforming) but also at the central level (before or after correlation) [ref].

- Calibration. Initially the locations and frequency-dependent gains and phases of each receiver unit are unknown and need to be estimated. Additionally, the disturbance due to the propagation through the earth ionosphere (time- and space-varying) has to be measured and compensated for. For large distributed arrays, this is a challenging task [ref].

- Imaging. In its simplest form, image formation consists of a spatial Fourier transform of the received correlation data, followed by deconvolution to compensate for the subsampling of the spatial domain. Accurate array calibration parameters are needed to perform this step correctly. After initial image formation, iterative deconvolution algorithms are used to find the locations of the point sources and subtract their effect in the image, so that the more subtle structures become visible. This step can be combined with a gradient search to improve the calibration parameters. Current techniques such as self calibration need to be extended to the case of distributed arrays with millions of unknown parameters. Also the fact that calibration parameters change within the beam of each station introduces a space varying beam that needs to be considered in the imaging process. Finally new insights coming from real time RFI mitigation can be used to improve the quality of the image formation, by considering strong sources as spatially located interference. This leads to a new generation of image formation techniques [ref].

     

 

Return to New Initiatives and Trends

IEEE Signal Processing Magazine 2007 - Hosted by the Digital Video and Multimedia Lab at Columbia University
SPM e-Newsletter Hosted by the Communications and Signal Processing Labs at University of Maryland