MULTIMODAL BIOMETRIC IDENTIFICATION USING FINGERPRINT AND VOICE

Abstract

Biometric identification is a procedure of identifying a person according to his/her distinguishing characteristics. It comprises means for uniquely recognizing humans according to one or more intrinsic physical or behavioral characters. Nowadays fingerprint identification used in a different company in the world, include Ethiopia. The most challenge in the performance of a fingerprint identification system is heavily impacted by the fingerprint image quality. This study aims to design a multimodal biometrics identification system using fingerprint and voice, which is used to improve security, to address unimodal biometrics identification problems, and to increase the accuracy of fingerprint and voice identification. All work is done in MATLAB environment and voice recorded by TECNO Spark Mobile. The system is designed with two main components such as fingerprint identification and voice identification component. Researchers design those components to perform common activates such as preprocessing, feature extraction, training, and test. Finally fusion fingerprint with voice, and evaluate the performance of fingerprint, voice, and Multimodal Biometric identification. During voice preprocessing perform Silence removal and pre-emphasis activity additional to feature extraction and Hidden Markov Model used for train and test voice. The researcher use minutiae extraction and minutiae matching for fingerprint identification. To integrate fingerprint and voice, first normalized by min-max approach then integrate using matching scores level fusion by sum rule. For the experiment 30 fingerprints and 30 voices for training and testing were used, the fingerprint data are obtained from the Fingerprint Verification Company database from an online source and the voice was recorded from male and female on 16KZ frequency. The multimodal biometric identification using fingerprint and voice is evaluated using 10 voice data with 10 fingerprints, and the performance result shows fingerprint and voice give 85% and 75% accuracy respectively, and their combination gained 90 % accuracy result. Multimodal biometrics identification system using fingerprint and voice is very important to reduce fingerprint attacks such as; Spoofing, Exploit similarity, zero-effort attempt, Replay attack, Denial of Service attack, and Hill-climbing types of attacks. The main contribution of this thesis is to increase the accuracy of the Multimodal Biometric identification is a procedure of identifying a person according to his/her distinguishing characteristics. It comprises means for uniquely recognizing humans according to one or more intrinsic physical or behavioral characters. Nowadays fingerprint identification used in a different company in the world, include Ethiopia. The most challenge in the performance of a fingerprint identification system is heavily impacted by the fingerprint image quality. This study aims to design a multimodal biometrics identification system using fingerprint and voice, which is used to improve security, to address unimodal biometrics identification problems, and to increase the accuracy of fingerprint and voice identification. All work is done in MATLAB environment and voice recorded by TECNO Spark Mobile. The system is designed with two main components such as fingerprint identification and voice identification component. Researchers design those components to perform common activates such as preprocessing, feature extraction, training, and test. Finally fusion fingerprint with voice, and evaluate the performance of fingerprint, voice, and Multimodal Biometric identification. During voice preprocessing perform Silence removal and pre-emphasis activity additional to feature extraction and Hidden Markov Model used for train and test voice. The researcher use minutiae extraction and minutiae matching for fingerprint identification. To integrate fingerprint and voice, first normalized by min-max approach then integrate using matching scores level fusion by sum rule. For the experiment 30 fingerprints and 30 voices for training and testing were used, the fingerprint data are obtained from the Fingerprint Verification Company database from an online source and the voice was recorded from male and female on 16KZ frequency. The multimodal biometric identification using fingerprint and voice is evaluated using 10 voice data with 10 fingerprints, and the performance result shows fingerprint and voice give 85% and 75% accuracy respectively, and their combination gained 90 % accuracy result. Multimodal biometrics identification system using fingerprint and voice is very important to reduce fingerprint attacks such as; Spoofing, Exploit similarity, zero-effort attempt, Replay attack, Denial of Service attack, and Hill-climbing types of attacks. The main contribution of this thesis is to increase the accuracy of the Multimodal biom