Sparse random projection for efficient cancelable face feature extraction
Youngsung Kim, Kar‐Ann Toh · 2008
Based on a recently proposed framework for cancelable biometric template generation, this paper focuses on boosting the computational efficiency using a sparse random projection. Comparing with a non-sparse random projection, we show empirically that the verification accuracy of templates generated by sparse random projection do not degrade while enjoying a more efficient feature extraction process than before. This work contributes to establishment of an algorithm for effective cancelable face template generation.