الفهرس 
AbstractOnly 14 pages are availabe for public view
 |  | from | 105 |  |  |
| | | from | 105 | | |
Abstract
Biometrics is getting more and more attention in recent years for security and other purposes. Biometrics is a scientific discipline that involves automated methods of recognizing people on the basis of their physiological and behavioral characteristics. One of the widely adopted behavioral biometrics is “voice recognition” where the person is identified and/or verified through the characteristics of its voice (voice biometrics). The main strength of this technology is its dependency on a signal which is natural and not annoying to generate. Also, this signal can be easily obtained from almost anywhere with no dedicated person equipment or training. However, there still exist many problems that pertain using the individual’s voice as a verification tool such as improving the identification accuracy in harsh environment and reducing the time required to perform the recognition process to make the voice recognition technology more suitable for real time applications or the applications which have been developed to be used by a large number of users. In this thesis, we have designed and implemented two systems. The first one is based on Support Vector Machine SVM classifier while the second is using the Particle Swarm Optimization PSO algorithm to optimize the parameter values of the SVM classifier. In other words, the second classifier can be considered a hybrid classifier (SVM-PSO). Generally, our proposed work has been designed and implemented with two main objectives in mind, specifically, maximizing the accuracy of recognition and the minimizing the required time in order to verify the identity of some one.
The experiments have applied on a data set which consists of 450 voices for both genders (80 train – 370 test). The proposed systems are robust, good verification results have been achieved, and the verification time has been reduced.