Over the last two years, we have been working on software
infrastructure with our minds bent toward building an effective
software intercom that
might be deployed using ubiquitous computing paradigms in a house or
small office. This intercom system would facillitate conversations
across traditional house or office boundaries. Conversations might be
initiated to connect any people as they move throughout the environment
while one-way audio connections might allow a person to listen to a
sleeping baby as he or she moved throughout the house. In our current
implementation of the Family Intercom, we leverage
existing locationing infrastructure in Georgia Tech's Aware Home. While we
have been
building the software behind such an intercom, colleagues here at
Georgia
Tech have been exploring the social semantics of such interaction. Figure 1 is an example screen from the current GUI
implementation.
People and intercom audio nodes (icombos) are displayed in the home
based on where they are logically (master bedroom). The user can then
actuate the GUI to create one or two-way connections between people and
rooms. In this way, a person might listen or announce to a
room/person/people, or rooms of people might communicate across the
house's boundaries.
Figure 1, The current GUI for the
Family Intercom allows simple visualization of where in the home users
are. The users can then be connected via audio channels provided by
nearby Microphone/ Speaker media nodes.
Research
Over the course of three implementations of the Family Intercom, we
have
learned some significant lessons about the deployment and
development of ubiquitous computing infrastructure. Early in the
process, we decided to use simple microphone/speaker combination nodes
that would provide feedback-free full-duplex audio by using a
specialized proprietary audio card. Our initial implementation was a
very ad hoc deployment of singular programs allowing the connection of
audio inputs to audio outputs through the MediaBroker.
This
implementation relied on a single control interface and required the
direct specification of the network locations for all nodes. As the
MediaBroker evolved our Family Intercom evolved into a second
implementation. This second implementation still lacked the scalability
and autonomic tendencies we wished for. Our third implementation again
leveraged the MediaBroker to do what the
MediaBroker was very good at; transfering media from sources to sinks
while transcoding the media to suit all parties. This time, we have
added UPnP to the implementation in order to allow autonomic creation
of
connections to other audio nodes as well as the now distributed control
nodes. UPnP interaction has also allowed interaction with the new Aware
Home location system based on transient UPnP People devices.
The current audio node architecture and the current control node
architecture can be seen in Figure 2.
Figure 2
After each iteratively more successful implementation of the Family
Intercom we have examined our success and where we have still fallen
short. Our experience with the Family Intercom has led us to look to
what current academic and industry infrastructures exist and how they
might be leveraged in the deployment of future applications. This
examination has roughly emerged as the search for a standard ubiquitous
infrastructure.