Random translational transformation for changeable face verification

Abstract

Producing changeable biometric templates from limited number of human biometric traits is important for the deployment of biometric technology in a wide variety of applications. This paper introduces a new method for changeable face verification using random translational transformation. The proposed method is based on translating the original feature vectors by adding a randomly generated vector. The sorted index numbers of the resulting vector is stored as the template for verification. The random translational transformation in conjunction with the sorted index number approach constitutes a non-invertible transformation, and hence the privacy of the users can be protected. It is shown that the proposed method is computationally simple, and is capable of generating templates with strong changeability. The effectiveness of the proposed method is well supported by both the detailed analysis and extensive experimentation.