Home > About Us > Technology > Technical Report > Technical Report No.23 2014

Real-time Communication Scene Detection and Analysis Technology

Today, we often capture important meetings on video for those who did not attend to view later. It is easier to understand what happened in a meeting by watching a video than by reading meeting minutes in text form, but it is difficult to record the video if one is not a professional videographer, and it takes a long time for viewers to watch the video because they cannot skip directly to the parts they wish to view.
In order to address this, we developed real-time communication scene detection and analysis technology, which uses cameras, microphones, and depth image sensors to detect and analyze who discussed what in the meeting. Based on this detection and analysis, the technology uses multiple cameras that zoom in on and record the person currently speaking to automatically create a video of the meeting. In addition, this technology also creates an overall view of what occurred, indexed based on changes that occurred in the state of communication. Viewers can see a timeline of the entire discussion and easily select only the scenes they wish to view.

Author

  • Jun Shingu
    Incubation Center, Research & Technology Group
  • Takayuki Takeuchi
    Incubation Center, Research & Technology Group
  • Jim Vaughan
    FX Palo Alto Laboratory
  • Yoko Kurihara
    Incubation Center, Research & Technology Group
  • Tsutomu Abe
    Incubation Center, Research & Technology Group
  • Don Kimber
    FX Palo Alto Laboratory