Do embedded processors shape applications, or is it theother way around?
In reality, it works both ways. This is particularly evident indigital-signal-processing-intensive applications, such as wirelesscommunications and video compression. These applications became feasible on alarge scale only after the emergence of processors with adequate performanceand sufficiently low prices and power consumption. And once those processors emerged, theseapplications started to take off. Then, the growing market attractedcompetition and investment. Processorvendors tuned their processors for these applications.
As a result, we got a generation of DSP processors withfeatures such as add-compare-selection instructions for Viterbi decoding, and ageneration of DSPs and CPUs with features like sum-of-absolute-differenceinstructions and single-instruction-multiple-data operations for videocompression.
A few years ago, after nearly two decades of evaluating andusing embedded processors for digital-signal-processing-intensive applications,my colleagues and I at BDTI realized that embedded computer vision applicationswere poised to benefit from the same type of "virtuous circle" that hadpreviously enabled the proliferation of wireless communications and videocompression algorithms.
Computer vision has been around for decades in applicationslike factory automation. But only veryrecently has vision begun to be incorporated into high-volume applications likevideo games and automobile safety systems. And, now that vision is starting to appear in volume applications,processor vendors are beginning to focus on embedded vision applications, andto tune their processors for these applications -- often by incorporatingspecialized coprocessors specifically designed for vision processing.
It's easy to see why processor suppliers are excited aboutembedded vision applications. "Machinesthat see" offer compelling value in many applications and markets. Takeautomotive safety, for example. Over onemillion people are killed each year in automobile accidents. By reducing thenumber and severity of collisions, vision-based safety systems may be able tosave many thousands of lives.
Embedded vision also promises to improve human-machineinteraction-- long the Achilles' heel of consumer electronics. Instead ofhunting for the right hand-held remote control, imagine a world where yousimply stare at your TV for a few seconds, and in response it turns itself onand offers you a personalized menu of options, which you can choose from viasimple gestures. Market research firm IMS Research estimates that by 2015,vision-enabled devices will be shipping at a rate of over 3 billion units per year.(Read about many more embedded vision applications here.)
In some applications, vision functions will be relativelysimple and will be able to fit into existing processors (perhaps with a modestboost in clock rate or an additional core). But many of the most compelling embedded vision applications use veryperformance-hungry algorithms. Implementing these algorithms at low cost and low power consumption willrequire specialized processors. As aresult, we expect to see processor suppliers introducing more processors thatare optimized for vision applications, and providing more applicationdevelopment support (such as optimized software libraries) for theseapplications.
Jeff Bier is founder,Embedded Vision Alliance and president, BDTI. This article was originally posted by EETimes.
