This paper introduces the development of a practical security system based on the individuals' biometric features. In such systems, problems emerge from either the individual him/herself or the surrounding environment. In the proposed system, possible problems were pointed out and solutions were suggested. During the process of logging the system, the voice sample passes through four stages: capturing, preprocessing, feature extraction, and verification. During the capture stage, the environmental problems are avoided as possible by using accurate sensitive microphone and powerful sound card. In preprocessing stage, the voice sample is supplied digitally by the sound card then this sample is normalized using scaling, trimming, interpolation algorithms ensuring that voice samples are of the proper dimensions without lose of intelligent information. In the verification process, the proposed system uses both algorithmic techniques and Artificial Neural Networks (ANN). The proposed algorithmic technique uses the accumulated search distance (ASD) to determine the dissimilarity between the target and reference samples. Moreover, a multilayer feed forward ANN determines the similarity between them. Finally, both results are integrated and verified against the predetermined threshold. |
