"Grand Challenges" in Signal Processing Research

There are several research problems in signal processing of such profound importance and huge potential impact (if they can be solved) that they will fundamentally impact and improve the lives of most people in the developed world. A few are listed below.

Universal Speech Recognition

The most natural and efficient way for humans to communicate is through speech, yet we generally must interact with our most advanced and powerful technology through cumbersome and unnatural interfaces such as keypads, keyboards, and mice. The recent emergence of speech recognition in certain applications, such as automated telephone-based query systems and voice dialing, gives a hint of the great improvements in convenience, efficiency, productivity, and user satisfaction that automatic speech recognition may ultimately provide. The tiny range of today's applications is due to the relatively poor performance of these systems, particularly in noisy or uncontrolled environments. Signal processing advances that extend both the accuracy of automatic speech recognition and the environments in which it can be used (e.g., the automobile, the home) would make almost all of our technology more useful and more appealing, would massively improve productivity in many industries, and would greatly enhance the quality of our modern technocentric life. This signal processing problem is incredibly challenging, but there are few technological advances that could have a larger human impact.

Distributed Sensing

We are rapidly entering an era with computers and sensors everywhere. (For example, a room with even a few people likely has several microphones and cameras attached to networked communication devices (they're called cellphones).) The technology will soon exist to deploy networks of sensors at reasonable cost and to gather many types of data from virtually anywhere. This could have revolutionary impact in potentially thousands of applications such as Almost all signal processing techniques known today work only with data from single sensors, or carefully (and expensively) positioned multiple sensors (such as an equally spaced linear array of antennas or microphones.) The challenge is how to process data from many distributed sensors arbitrarily or randomly placed, often of different types, to obtain situational awareness, reduce false alarms to tolerable levels, and to fill in the gaps. Relatively little is known about such signal processing problems, yet their solution will be required to take full advantage of the fruits of the electronics revolution.

Cognitive Radio

The telecommunication and wireless revolution has vastly altered, and generally improved, human life. Radio spectrum scarcity threatens to stifle this ongoing revolution. However, measurements in even busy urban areas such as Washington, D.C. have shown that even though almost all of the spectrum is allocated, at any given time only a small fraction is used. New, agile communication methods and signal processing techniques that can detect when spectrum is free and make efficient use of it while not disrupting primary services could provide ten times more overall spectral efficiency, thus enabling low cost, broadband wireless communications to continue to revolutionize our work and play. Achieving this will require fundamental new advances in communication theory, communication system design, and the signal processing technology with which it will be realized. Given that telecommunications directly represents over four percent of our gross domestic product (GDP) and indirectly sustains entire other industries making up a much larger fraction, perhaps no other research area offers as much promise for enhancing future economic growth.

Medical Imaging

Modern medical imaging technologies, such as magnetic resonance imaging (MRI) and computer-aided tomography (CAT), have revolutionized health care. Most people don't realize that these technologies actually collect raw data that is unintelligible to doctors or to anyone else; it is through signal processing that the data are unscrambled to produce recognizable medical images. The current signal processing is relatively simple, and much of the advances in resolution and quality in the past have largely come from improving the imaging hardware. However, the hardware advances are now reaching their limits, and future improvements will have to come largely from new signal processing techniques and from new combinations of imaging hardware and signal processing. With continued research in signal processing for imaging, the medical imaging revolution is still near the beginning; without such signal processing research, the revolution is nearing its end. Some of the most pressing problems in American healthcare, such as the early detection of breast cancer, will be solved only through sustained research in signal processing.

Return to Doug Jones's Homepage

Created by Douglas L. Jones on April 21, 2005; Last updated by D.L. Jones on April 21, 2005