There's a group at the VU Amsterdam working on this. Maybe there's something interesting in their list of publications: http://amsterdamgesturecenter.com/publications/
I know they're doing studies using corpora of gesture-annotated video data, and work in the cognitive linguistics framework.