October 10, 2013
ECE Ph.D. student Stephen Siena was awarded Best Paper honors at the IEEE Sixth International Conference on Biometrics: Theory, Applications and Systems (BTAS 2013). The paper, "Maximum-Margin Coupled Mappings for Cross-Domain Matching," was co-authored by Postdoctoral Fellow Vishnu Naresh Boddeti and ECE Professor Vijayakumar Bhagavatula.
The paper considers the problem of recognition across different domains of data. Typical pattern recognition techniques struggle to match different sets of data, even if the different sets of data represent the same thing. The paper proposes a formulation for learning projections that map different domains of data into a common space, where matching across the different domains can be done more effectively. The work demonstrates applications for low-resolution face recognition and cross-sensor ocular biometric recognition.