| Outcome/accomplishment:
Researchers at the Quality of Life Technology Center (QoLT) are collecting
a large dataset of sensor readings from actors performing kitchen tasks,
and they are making the dataset freely available on the Internet for experiments
in activity recognition and monitoring.
The NSF-funded QoLT is an
Engineering Research Center (ERC) headquartered at Carnegie Mellon University
(CMU) and the University of Pittsburgh (UPitt).
Impact/benefits:
Researchers in computer graphics, computer vision and robotics
have begun to work with
very large collections of data to model human motion. These databases
have been used to create models of human movement for which researchers
from all over have found many applications in sports, science, medicine,
biomechanics, animation of avatars in games or movies, surveillance, better
strategies for humanoid robots, and human activity recognition. These
databases have facilitated research and provided standardized test datasets
for algorithms.
Explanation/ background:
Identifying human activities from sensors combines many simultaneous inputs:
there are numerous movements associated with each activity; there are numerous
ways to measure the movements; there are multiple ways to analyze the sensor
data; people perform the activities differently; and an individual may
not perform the same activity consistently.
The activities measured by
QoLT include several actors preparing several dishes, as well as some unusual
situations such as a pan fire, broken dishes and leaving the refrigerator
door open. The dataset includes video and audio from multiple
locations in the kitchen space, as well as a body-worn camera and audio
recording device and accelerometers. To support sharing of the data,
QoLT researchers have created a publicly accessible Internet repository
and several tools for uploading, analyzing and visualizing all the data. |
