Computer Vision: Object Recognition and Human-Computer Interaction

Publication Year:
2013
Usage 46
Abstract Views 41
Downloads 5
Repository URL:
http://openworks.wooster.edu/independentstudy/709
Author(s):
Snyder, Benn
Tags:
computer vision; kinect; openni; libfreenect; gesture recognition; Graphics and Human Computer Interfaces
thesis / dissertation description
This thesis focuses on computer vision and gesture-based human-computer interaction. In examining computer vision, the project covers existing computer vision systems, including OpenNI/NITE and libfreenect. It explores topics such as identifying humans and objects in scenes, recognizing gestures and context-specific movement, and more general scene analysis. The results of the computer vision work are applied to human-computer interaction. The project examines different types of user interfaces and the applicability of gesture-based interaction to those interfaces. One goal is a generic system for controlling user interfaces using a vision-based gesture system. The included software, consisting of three separate projects, covers both high-level HCI and low-level interaction with vision sensor device drivers. HighNI is a set of Java modules for OpenNI and NITE 1.5. It abstracts the existing Java classes to a higher level and provides some skeleton examples for gesture callbacks. OpenNI2-FreenectDriver is a bridge driver that connects OpenNI2's driver interface to libfreenect's API. It allows OpenNI2 to use Kinect hardware on non-Windows platforms where Microsoft's SDK is not available. FreeNUI is a new framework for natural user interaction, based around libfreenect and OpenCV. It is an experimental project that explores interface design and C++11 language features.