Tracking Facial Features Using Cluster of Point Distribution Models

Invention Summary:

There is a need to address automobile driver alertness and driver fatigue aiding towards safe driving solutions. There is also a demand for higher levels of security to combat any threats to safety of our homes or our nation.

Rutgers researchers have developed a technology for real-time tracking of facial feature shapes, and expressions on a non-linear manifold applied to pose prediction, expression recognition, and eye tracking. The technology proposes a novel framework for tracking faces across large head rotations at near real-time processing rates. It also provides an integration of shape registration and tracking frameworks for shapes lying on any manifold by approximating non-linearities as piecewise linear surfaces..

Market Application:

• Monitoring driver alertness.
• Homeland security applications.
• Medical applications.
• Games and entertainment.
• Product placement applications in stores and customer-related analysis of interest in products.

Advantages:

• Speed and accuracy. Technology uses a dynamic analysis of the features (as opposed to static, which all other methods do).
• The technology tracks the face across any generic movement.
• The framework runs in real-time, and the tracking is robust to full head turning and for any head shape.

Intellectual Property & Development Status:

Issued US Patents: 8,121,347 & 9,014,465. For any business development and other collaborative partnerships contact marketingbd@research.rutgers.edu

Publications: Metaxas, D., & Kanaujia, A. (2013). Tracking Facial Features Using Cluster of Point Distribution Models.