|
|
|
| Abstract Title:
|
| Advancements in motion analysis through Sciences and Arts integration
|
| Graduate Student Presenter:
|
Stjepan Rajko
|
| Name of the Author(s) and Affiliation(s):
|
Stjepan Rajko, Arts, Media and Engineering
|
|
During my IGERT traineeship, I had the opportunity to work in teams that span the sciences and the arts. This poster presents two bodies of work that demonstrate how the integration between the sciences and the arts has led to advancements in state-of-the-art motion analysis in areas such as gesture recognition and motion capture. In one body of work, unique combination of computational pattern analysis knowledge from computer science and movement domain knowledge from dance led to improvements in real-time gesture recognition. These improvements reduced the training requirements of gesture recognition systems and increased run-time speed without sacrificing performance (such as the modeling precision of the gestures). In particular, we are able to obtain reliable gesture recognition using only one or two training samples per gesture. We have successfully used such gesture recognition for both gestures expressed using the body and gestures expressed using tangible objects (such as a ball). In the other body of work presented in this poster, we are exploiting probabilistic graph models to overcome challenges presented by applications involving movement rehabilitation and creativity to develop novel motion tracking algorithms that reduce the costs of motion capture sessions. The costs are affected by reducing the time needed to prepare a subject for a motion capture session, and by allowing the use of less expensive off-the-shelf motion capture systems.
|
|
|
