Rhythmic Gait Signature from Video without Motion Capture
Author(s)
Sadikali, Akil
Boyd, Jeffrey
Advisor(s)
Editor(s)
Collections
Supplementary to:
Permanent Link
Abstract
The goal of gait biometrics is usually to identify individual
people from a distance, often without their knowledge. As such,
gait biometrics provide a source of data that ties a visible pattern
of motion to an individual. We describe our work to convert one
particular biometric gait signature into a rhythmic sound pattern
that is unique for different individuals. We begin with a camera
viewing a person walking on a treadmill, then extract a phase configuration
that describes the timing pattern of motions in the gait.
The timing pattern is then converted to a rhythmic percussion pattern
that allows one to hear differences and similarities across a
population of gaits. We can also hear phase patterns in a gait independent
of the actual frequency of the gait. Our approach avoids
the inconvenience and cost of traditional motion capture methods.
We demonstrate our system with the sonification of 25 gaits from
the CMU Motion of Body database
Sponsor
Date
2010-06
Extent
Resource Type
Text
Resource Subtype
Proceedings