Two biometric identifiers are better than one. Researchers fuse face, ear images
Researchers in a multinational team say they have created a biometric recognition system that uses three-dimensional images of faces and ears together that is 99.25 percent accurate with an 0.75 percent error rate threshold.
The two biometrics are fused to increase authentication accuracy, according to a paper from a team made up of scientists from India, South Korea, Saudi Arabia and Bangladesh. Their results indicate the method is competitive with the state of the art.
Three-dimensional facial images have their challenges, not least of which is file size, but they are better than two-dimensional images in that they are capable of solid face affirmation and work better in low light. And there is more biometric data with which to authenticate.
Three-dimensional ear images are stable over many years but have problems with low light levels and pose variations, and large file sizes can result in high computational costs.
Principal component analysis was used for three-dimensional facial recognition and independent component analysis was used for ear recognition.
For facial recognition, the team used the Face Recognition Grand Challenge database. Each of the chosen 30 subjects expressing anger, happiness, disgust, fear, surprise and sadness.
The University of Notre Dame collections F and G provided 30 pairs of ear images captured at various angles.
The researchers said they next will tackle biometric hardware. They are considering ways to cut the cost of a portable 3D scanner in a handheld or notebook device.