Microsoft's RoundTable, which was demonstrated at the Nortel and Microsoft
Unified Communication seminar yesterday, is a table-top device, not much
bigger than a speaker telephone. It can be connected to a standard PC giving
synchronised voice and video conferencing.

The device creates a 360-degree panoramic video of the images of people
taking part in the teleconference and tracks the flow of conversation,
enabling the image and voice of the person speaking to be spotlighted.
Participants across multiple locations can participate.
RoundTable works with Microsoft's Office Communications Server 2007 and
Office Live Meeting 2007, allowing companies to integrate virtual
presentations, and share whiteboards and information. Conference calls can
be recorded and viewed later.
What differentiates RoundTable from other conferencing technologies is that
most videoconferencing systems are expensive and complicated to install and
operate. As a result, they are usually confined to specially configured
meeting rooms. RoundTable is plug-and-play.  Users need almost no training
to set it up and use.
One of the challenges of video conferencing is to determine the dominant
speaker in a conference room when several people are talking at the same
time.  Developers also had to deal with background noises and surfaces that
reflect sound a weak point of many video conferencing systems.
The challenge was to remove background noise, enabling users to focus on
what the speaker and not the sound someone is making, such as typing notes.
To make the conferencing experience as realistic as possible, the developers
solved other challenging technical problems, such as how to equalise the
light levels when one part of the conference room is darker than another.
Six separate microphones are built into RoundTable to help differentiate
between different sounds and isolate the dominant speaker through a process
called "beamforming".
Each of the microphones is built into a different part of the device.
Although sound travels fast – roughly 350 meters per second – there is a
slight time difference from the time a microphone on one part of the device
picks up a speaker's voice compared to a microphone on another part of the
RoundTable calculates this difference to identify the source of the voice.
It then uses visual cues to pinpoint, enlarge and emphasise the face of the
One of the challenges was how to differentiate between reflective and
non-reflective sound – something that most speakerphones cannot do. In some
cases, the sound that bounces off a white board has the same energy level as
a person's voice. This can create noise and static for people who are on the
other end of the speakerphone.
Microsoft came up with a solution to this problem and used RoundTable's
"eyes".  The sound-isolating technology relies on the cameras to isolate
where each person is in the room. This way, the device knows whether a
person or a white board is doing the "talking".
RoundTable is ideal for the two most general types of conferenced business
meetings. Firstly, meetings with two groups of people in two different
conference rooms in different locations. With a RoundTable in each room, the
devices will capture a 360-degree view of each, and combine images of each
person on the monitor.
Empty spaces around them are discarded. If someone has a PowerPoint or other
document to share with the group, it appears on the screen alongside the
images of the people.
The second type of meeting is also common.  One conference rooms is full of
people and several other people are dialed in from their office, home or a
hotel. The remote attendees can use a RoundTable device or a standard Web
cam to include their image in the meeting. If not, they can dial in from a
standard phone, and their voice will be added to the session if they have an
online connection. They will then be able to see everyone in the conference