/Kinect-Speaker-Detection

Kinect Speaker Detection

Primary LanguageC#

Project done at CEERI, Pilani.

Works with 3+ Microsoft Kinect sensors simultaneously, to detect the speaking person.

Include MS Kinect.Toolit and Kinect.Toolkit.Facetracking in the VS Project Solution.

Things Done

  • Initialise multiple Kinects and run them concurrently.
  • Checking combined audiostream, and monitoring audio levels.
  • Switching to the camera with highest associated input sound intensity (w/o background noise).
  • Display the correct person on screen.

Possible enhancements

  • Create a wrapper so that a new Kinect can be initialised with one command.
  • [Optional]: Automatic addition of sensor if a new Kinect is connected to the computer.
  • [Optional]: Improve live camera switching to make it seamless, and instantaneous. (done!)
  • [Optional]: Lip tracking on face for better match.
  • Extract faces of people facing the different Kinect sensors and display their profile view.

Bitdeli Badge