We have described a static posture classification system based
on a sensing chair that monitors the pressure distribution patterns
on its surfaces in real time. The PCA technique, commonly
used in computer and robot vision, has been successfully
applied to the problem of posture classification from pressure
maps. A PCA-based algorithm has the advantage of being very
fast. Its disadvantage is the lack of physical interpretations associated
with eigenposture spaces. Our current system runs in
real-time (with an update rate of roughly 6 Hz) on a Pentium
PC in Windows 98 environment. Average classification accuracy
is 96% for subjects the system had felt before, and 79%
for those who are new to the system. We are currently investigating
ways to improve the system’s accuracy for new subjects
by taking into account the first three eigenposture spaces that are
closest to the test pressure map. Future work is aimed toward a
dynamic posture tracking system that continuously tracks not
only steady-state (static) but transitional (dynamic) sitting postures.