CogniMem’s Chris McCormick, applications engineer, demonstrates how the addition of CogniMem’s general-purpose and scalable pattern recognition technology can be used to add hand gesture control to applications like Google Earth, in combination with the Microsoft Kinect. Cognimem CM1K is a massively parallel pattern recognition hardware accelerator than can learn and recognize in real time, as well as work with noisy inputs. Here the hand is isolated (using Microsoft’s skeletal tracking algorithm) and captured using 3-D information, then compared to a knowledge base of hand gestures for recognition. The hardware technology is taught, not coded, and new examples can be learned in the field or updated after system delivery without impacting existing software, hardware or system performance.
For more, see this video in the Embedded Vision Academy.