The need for personal safety and security is rising
everywhere: on the street, at home, in school. If we only had a
"watchdog" - one that would be everywhere and see everything, one that
would raise the appropriate alarm in real time – yet a watchdog that is
unobtrusive, doesn't compromise people's privacy, doesn't generate false
alarms, and doesn't cost a lot. Can we dream of such a watchdog? In the
HuSIMS project we aim to show that this is indeed feasible.
Concept A dense mesh of visual sensors ("a sensor on every
lamppost") that is sensitive to irregular behavior patterns in the
monitored area, and that upon detection of anomalous events can trigger
the appropriate emergency action and provide visual notification- all
this without compromising personal privacy.
The sensors will be simple, low-resolution, low-cost, low
power consumption, easy to install, integrated sensors. They will
include camera, processor and communication interface. After a short
learning period, each sensor will autonomously identify dangerous or
irregular situations: people running, commotion, one person chasing
another etc. Once a sensor identifies an alert situation, it will
communicate the alert and, upon request, send images for verification.
The sensors transmit images only upon demand or when an alert situation
is determined, minimizing the bandwidth requirements. Low resolution
ensures appropriate trade-off between security and privacy.
A wireless mesh network will connect the sensors to one or more
servers, and from there to control rooms and communication devices of
police, emergency medical centers, fire departments, etc. The network
will be self-organizing and self healing, allow quick addition/removal
of sensors and will naturally cope with non-functioning or sabotaged
sensors. The network will be based on WiMAX technology. For indoors
applications such as hospitals and hotels, we will switch to zigbee +
any wire (coax, phone line, power line) approach, to guarantee that the
information is transmitted.
A backend control application will handle the inputs from the
sensors, perform decision algorithms to produce alerts from multiple
sensor inputs, interface with operators and existing software/hardware
systems, and forward alerts to other servers / mobile units. The backend
application will also interface with emergency units and first
responders, will be able to update sensor software versions, demand
on-line video streaming, and control additional devices (street
illumination, public address systems).