Signal Processing - November 2017 - 16

would be important, putting intelligence
in systems, addressing diversified applications like safety, security, food, travel,
water, and similar resources including
those required in medicine, agriculture,
automotive, astronomy and astrophysics,
and so on. The context, would, however, change.
Ajay V. Deshmukh (ajay.deshmukh@
rediffmail.com) is a principal at the
Bajaj Institute of Technology, Wardha,
India.

Volker Lohweg
Signal processing is,
of course, in a mature
state. Many concepts
and theories are well
established. However, I believe that signal processing has
more to offer in the next ten to 20 years.
Two aspects: sensors will be the key
technology drivers and key enablers for
many applications, but signal processing has to be on board because of data
volume and necessary speed. These
facts will definitely create new signal
processing concepts.
Big data, which turn into smart data,
have to cope with speed in the context of
machine learning, classification, decision making, etc. Here we will also see
new concepts and maybe new theories
in the coming years.
Volker Lohweg (volker.lohweg@hs-owl
.de) is a professor at the Institute of Indus-
trial IT, Germany.

Abdelhak M. Zoubir
Signal processing will
continue to g r o w.
There are many challenges in rea l-life
applications where
signal processing is
much needed. These
areas include renewable energy, car
engine monitoring, autonomous driving,
synthetic aperture sonar, psychology, biomedical engineering, big data analytics,
and synthetic biology, to mention a few.
My view is that there is little room for
theoretical breakthroughs in signal processing. This does not mean that there
will not be emerging areas with great
16

potential for the advancement of knowledge in signal processing and solving
the many unanswered theoretical questions, such as in adaptation and learning
over sensor networks.
The future of signal processing and
its success lies, in my view, in answering
questions encountered in new application
areas, for example, cultural heritage, and
there is a trend to moving to new application areas outside the electrical engineering field. Indeed, many of my colleagues
share this view with me, and we are putting much effort into solving new problems. In short, the future and potential
of innovation in signal processing lies in
interdisciplinary research (see [1]).
Abdelhak M. Zoubir (zoubir@ieee.
org) is a professor at Technische Univer-
sität Darmstadt, Germany. He served as
editor-in-chief of IEEE Signal Processing Magazine from 2012 to 2014.

Kush R. Varshney
Signal processing theory and methods are
increasingly applicable to data science
applications outside in
society. With th is
emer-gence in sociotechnical (not simply technical) solutions,
new problems begging for research
breakthroughs are popping up. We have
started investigating some of these problems through the IBM Science for Social
Good initiative, but we are just at the tip
of the iceberg.
Kush R. Varshney (krvarshn@us.ibm
.com) is a member of research staff and
the manager of Data Science Theory and
Algorithms at the IBM Thomas J. Wat-
son Research Center, Yorktown Heights,
New York.

Muhammad Zubair Ahmad
Signal processi ng
has historically been
associated with the
communication systems and one-dimensional signals. The
rapidly developing
field of machine learning has been
perceived as the alternative to signal
processing. Evidence of this is the
IEEE SIGNAL PROCESSING MAGAZINE

November 2017

discussion held at Technion in 2014,
under the title "Is Deep Learning the
Final Frontier and the End of Signal Processing?" The computer science community and industry has successfully
marketed machine learning as an alternative to signal processing. A major factor
contributing to this is the fear of signal
processing among undergraduates, which
I have found common across countries.
The research community at large has
learned to fear signal processing and has
been told that machine learning and data
processing is a skill that can be acquired
with minimal training.
Thus, I would call for a rebranding of
signal processing as a paradigm fundamental to industrial applications and the
modeling of physical reality. Anything
ranging from the individual systems of the
human body to the machines and automata
developed by the humans can be modeled
using this framework. Different modeling
strategies based on statistical learning theory, analysis, differential geometry, topology, and group theory must be generalized
under this universal framework.
The signal processing community
must come up with a method of teaching that is alluring to young students. We
need to convince young (as well as old)
minds that the modern trend of sacrificing understanding for practicality may be
economically good but is directly conflicting with the driving principles of the
scientific community.
Muhammad Zubair Ahmad (ahmad
mz@myumanitoba.ca) is a Ph.D. degree
candidate at the University of Manitoba,
Canada.

Thomas A. Baran
This question reminds
me of a favorite say-
ing of my Ph.D. advisor,
Prof. Alan Oppenheim: "There will al--
ways be signals, they
will always need processing, and there will always be new
applications, new mathematics, and new
implementation technologies."
In October, we held a symposium titled
"The Future of Signal Processing" at the
(continued on page 25)

Table of Contents for the Digital Edition of Signal Processing - November 2017