Rate

 

2 Comments / 416 Views

visit the Parallel Programming community

This demonstration shows an innovative solution to a classic problem: how computer speech recognition can determine whether the user is talking to the computer, or to someone else. Face-tracking software processes video from a camera atop the PC screen to determine whether the user is facing the PC, or looking away. While facing the PC, audio-input and speech-recognition is enabled; while looking away, it's muted. This technique also helps provide coarse end-pointing in a noisy environment.

This demo highlights the use of two compute-intensive and highly-parallelizable workloads: image-processing and speech-recognition, and is currently only feasible on multi-core PCs. It's also best-suited as a client-based workload (rather than cloud/server-based), due to video bandwidth.

To learn more about the research that Intel is involved with please visit: www.intel.com/research