الفهرس | Only 14 pages are availabe for public view |
Abstract Speech conversations are very important part in our daily life. In the last few years, the fast and extensive growth of network technology increased speech conversations through Public Switch Telephone Network (PSTN), cell phones, satellites and the Internet. It is almost impossible to prevent unauthorized people from eavesDROPping, when an audio signal is broadcasted in air or through insecure wired connections. Therefore, it has become possible for any unauthorized person to receive the transmitted data with simplest receivers.Protection of speech conversations has recently become an important issuebecause of the great development in cryptanalysis activities.This thesis is concerned with speech cryptosystems. It presents a group of efficient metrics that can be used for performance evaluation of these cryptosystems. In addition, two approaches for speech encryption are presented.The first approach is based on chaotic Baker map. It comprises both randomization and masking to fill silent periods and change signal energy and pitch. The second approach is based on the permutation and substitution of audio samples depending on a secret key and the utilization of discrete transforms. Simulation experiments have shown that the proposed cryptosystems are robust to several cryptanalysis attacks with a good immunity to noise and a moderate execution time. Finally, the thesis presents a general framework for a multi-level securitysystem implementing speaker identification, speech watermarking, and speech encryption. The performance of this system has been studied with speech enhancement and deconvolution schemes, and the simulations have shown promising results. |