The MM3101, launched later this week at the CES show in Las Vegas, supports image pipeline processing, image enhancement and embedded vision features, including video stabilizer, face detection, gesture recognition and augmented reality and is the first in a family of cores for machine vision applications.
CEVA is aiming the first core at consumer applications including smartphones and smart TVs to add gesture control.
Instead of using fixed-function engines or running these functions on the main application processor, the MM3101 is specifically architected to support all of these functions on a single processor, resulting in a 20X reduction in power consumption when compared to CPU-based solutions. The core is capable of processing video streams of 1080p, or 8 megapixel images at 12 frames per second, simultaneously with video stabilization, colour correction, wide dynamic range (WDR), face detection and gesture recognition.
“The critical role that high-performance DSP technology plays in mobile and digital home applications is being brought to the forefront by the advent of new video, image and embedded vision applications,” said Gideon Wertheizer, CEO of CEVA. “Traditional combinations of hardwired blocks and general-purpose CPUs are no longer capable of delivering the required flexibility, power efficiency and performance for these fast-changing market trends. The MM3101 specifically targets these highly-complex applications and offers customers the ability to deliver cost-efficient, low-power and differentiated solutions through software."
CEVA is also collaborating with eyeSight on Human Machine Interface (HMI) software based on the MM3101 platform. eyeSight’s offering includes hand gesture recognition and virtual mouse user interface technology using finger tracking for multiple users in various conditions. CEVA and eyeSight will demonstrate a range of real-world applications for this technology at CES 2012. CEVA made an investment in the Israeli company last month as part of this move.
"Next-generation mobile and consumer devices are rapidly adopting embedded vision capabilities including gesture-based user interfaces and object recognition for applications such as augmented reality games, security, and navigation,” said Jeff Bier, founder of the Embedded Vision Alliance. “Implementing this type of visual intelligence in consumer products requires a combination of high processing performance, programmability, energy efficiency, and low cost. Fielding a licensable processing subsystem specifically targeting embedded vision applications is an excellent advancement for CEVA and the industry."
The platform is supported by a Software Development Toolkit that includes an optimizing C-compiler, IDE, debugger, simulators and profiler. The IDE helps developers implement, debug, optimize and run code on the platforms. To aid Vector Processor optimization, the IDE provides a view into all core and memory resources. As part of the solution, theCEVA-MM3101 includes a comprehensive library of video and imaging functions, including image pipeline kernels, linear and non-linear filters, pre and post-processing functions, face detection and video codec kernels.