Performance enhancement of speaker identification systems using speech encryption and cancelable features

Biometric systems based on speech features constitute a new and evolving trend in security. This paper is concerned with speaker identification systems. It studies traditional speaker identification systems based on cepstral analysis and neural classification. The paper develops the idea of remote access systems with speaker identification concepts by introducing efficient cryptosystems to achieve a large degree of security in these remote access speaker identification systems. The proposed approaches depend on chaos theory to maintain a low sensitivity to noise effect. Moreover, the concepts of cancelable biometrics are developed in this paper for more secure speaker identification. As known in the literature, cancelable image biometrics are used to save the features of the users from being stolen. If a similar approach is adopted in wireless access speaker identification systems as in this paper, the security can be enhanced to a great extent.