Ear Recognition

Generic 3D-Driven Recognition System

We have developed a system which is capable of using 3D data as input, along with a suitable model to output metadata information. The metadata information is then used for recognition.

Motivation

• Ear’s similarities with the face
  • A human’s ear is unique
  • Strong geometric biometric characteristic
  • Commercial 3D Scanners can be used
  • Non interaction with the subject is required
• Ear’s advantage over the face:
  • Remains relatively undeformed under facial expressions

Main ideas

• Proposed method’s main idea:
  • Construct an Annotated Ear Model
  • Register each dataset with the model. Cost: O(n)
  • Extract metadata
  • Compare metadata directly (no extra registration)
• Most previous works:
  • ICP based
  • Register each dataset with all others.Cost O(n2)
  • Comparison based on ICP distances

Results

• Compared to Yan and Bowyer
  • Lower accuracy (less than 4%)
  • Higher efficiency
• Time needed for the full UND Ear database (415 gallery/415 probes):
  • 7 hours vs 276 hours!!!
  • Less than 3% of time needed
  • Extremely important for large databases

References

1. T. Theoharis, G. Passalis, G. Toderici and I.A. Kakadiaris. "Unified 3D Face and Ear Recognition using Wavelets on Geometry Images". Pattern Recognition, 2007 (In Press).
2. G. Passalis,I.A. Kakadiaris, T. Theoharis, G. Toderici, T. Papaioannou. "Towards fast 3D ear Recognition for real-life biometric applications". IEEE Advanced Video and Signal based Surveillance, London, United Kingdom, Sept. 5-7 2007.
3. I.A. Kakadiaris, G. Passalis, G. Toderici, N. Murtuza, and T. Theoharis. "Quo Vadis, 3D Face and Ear Recognition?" Multi-Sensory Multi-Modal Face Biometrics for Personal Identification, Eds. R.I. Hammoud, B. Abidi and M. Abidi, pp. 139-164, 2006.