| Windows CE Shared Source Contest 1st-place winner: Smart Cameras for Smart Conference Rooms |
(Aug. 16, 2006)
 While many aspects of the typical office or university environment have adapted to today's fast-moving, information-rich, Internet-based reality, the conference room remains an anachronism. This first-place entry in the Summer 2006 Windows CE Shared Source Contest implements a "smart conference room appliance" using inexpensive, off-the-shelf components.
Among other problems, conference rooms can't easily link up with rooms in other buildings, let alone other countries, and after the meeting the only record is notes that someone may have taken, and they generally aren’t electronically searchable, the developers of the "Smart Cameras for Smart Conference Rooms" project maintain.
The team's project represents a first step toward the goal of "an easy-to-use networked device which transforms an everyday conference room into a self-aware collaboration space." Their Windows CE-powered smart camera appliance uses sensors placed in a conference room to detect the person speaking, find the best camera to get video of the speaker, automatically point the camera at that person, then stream the video over the Internet. Meetings proceed naturally because the system automatically adjusts to the participants' use of the space, according to the project team.
Hardware
The system uses a pair of LogiTech QuickCam Orbit MP USB cameras connected to "an old PC we had lying around." This PC runs Windows CE, and is dubbed the "CEPC." The 1.3 megapixel webcams have a robotic camera head that can swivel 189 degrees horizontally and 102 degrees vertically under computer control. With more than 180 degrees of horizontal rotation, two cameras placed back-to-back in the middle of a conference room are sufficient to cover the entire space.
Smart Conference room system architecture
(Click to enlarge)
"Motes" -- tiny wireless sensors developed at the University of California, Berkeley -- are used to detect sound information for the system. By placing several motes throughout a conference room, we can identify the one that senses the loudest sound source, and from that infer the location of the person speaking.
"Currently the motes relay their information to a standard Windows PC, which sends the information across the Ethernet to the CEPC. Long-term, it would be useful to port the mote management code to the Windows CE platform, so the CEPC could interact directly with the motes."
Software
The CEPC includes the shared source webcam driver and the shared source CamServ service to stream video over the network. The ViewRCam client, included with CamServ, displays network-streamed video on another Windows PC.
The team says they modified the the shared source webcam driver to add support for camera movement. Angle movement can be specified as either absolute or relative. The ViewRCam client was modified to be multi-threaded, which made the video much smoother. The CamTest application was modified to support interactive camera movement while viewing the video output. This feature is used to set up and calibrate the system.
Operation
The system requires some setup and calibration before use. For the initial demonstration, six motes were placed around a conference table such that each one covers approximately on sixth of the room, according to the project team.
For initial setup, a person sits near a mote and speaks. An operator uses the CamTest program to focus the appropriate camera on the speaker. The process is repeated for each mote. The end result is a mapping of camera positions to mote sound levels.
Following setup, the system can be tested by simply walking around the room speaking in a normal voice. If the system is working correctly, the cameras should be able to follow the speaker around the room.
The team's video (see below) demonstrates the ability of the CEPC to track the currently-speaking member of a video conference. It also indicates some of the problems that remain to be overcome.
This Project's Whitepaper and Video
Click on these two links to download this project's complete whitepaper and project video:
This Project's Code
Portions of this project's code are available here (direct link to rar file).
About the Contestants
 
Team member images, automatically transmitted by the CEPC
Xin Hu (top-left), a first-year PhD student in the EECS department at the University of Michigan, has gained embedded experience by working with sensor networks, and is currently working on a handset security project. Zhigang Chen (top-right), a fifth-year PhD Student in same department as Hu, has been working with TinyOS for 3 years and just started using Windows CE as result of this contest. Scott Gifford (bottom), a research programmer at the University of Michigan working on location awareness technologies, has written applications to interface a Pocket PC with a location sensing system.
About the Summer 2006 Windows CE Shared Source Contest
The first-ever Windows CE Shared Source Contest, co-sponsored by Microsoft's Mobile and Embedded Devices (MED) group and WindowsForDevices.com, ran from May 23 through July 28, 2006. Contest participants were challenged to create cool, practical projects combining Windows CE 5.0 along with either or both of two Shared Source components -- the Windows CE WebCam driver and/or the Windows CE DVR engine. Prizes included Xbox 360 game consoles and other goodies.
Further details about the contest -- including the complete list of winners, the prizes awarded, and the list of judges -- are available here:
Windows CE Shared Source Contest winners announced
Related Stories
(Click here for further information)
|
|
|
|
|
|
|
news feed
