Due to biometric template characteristics that are susceptible to non-revocable and privacy invasion, cancellable biometrics has been introduced to tackle these issues. In this paper, we present a two-factor cancellable formulation for speech biometrics, which we refer as probabilistic random projection (PRP). PRP offers strong protection on speech template by hiding the actual speech feature through the random subspace projection process. Besides, the speech template is replaceable and can be reissued when it is compromised. Our proposed method enables the generation of different speech templates from the same speech feature, which means linkability is not exited between the speech templates. The formulation of the cancellable biometrics retains its performance as for the conventional biometric. Besides that, we also propose 2D subspace projection techniques for speech feature extraction, namely 2D Principle Component Analysis (2DPCA) and 2D CLAss-Featuring Information Compression (2DCLAFIC) to accommodate the requirements of PRP formulation.
Bibliographical noteFunding Information:
This work was supported by the Korea Science and Engineering Foundation (KOSEF) through the Biometrics Engineering Research Center (BERC) at Yonsei University (Grant No. R112002105080020 (2009)).
All Science Journal Classification (ASJC) codes
- Modelling and Simulation
- Language and Linguistics
- Linguistics and Language
- Computer Vision and Pattern Recognition
- Computer Science Applications