Browsing by Author "Rotinwa-Akinbile, Mariam Olasunmbo"

Recently, contactless biometrics systems have emerged not only for protecting the users of the systems against transmittable infectious diseases but also to permit remote access. However, the existing spectra of biometrics devices are mostly unimodal which are susceptible to several problems such as high intra-class and interclass variations. It may be possible to enhance the multi-functionalism of these systems by integrating two or more biometrics. Hence, in this dissertation, multimodal biometrics system involving the integration of palmprint, teeth and voice biometrics is developed. The development of the multimodal biometrics system was conducted in two phases. In the first phase, palmprint, voice and teeth recognition algorithms were separately developed. A novel skin segmentation algorithm was developed using artificial neural network (ANN) technique and this was deployed to segment the palmprint and face images from the background. Similarly, a novel region of interest (ROI) extraction technique was developed from which the palmprint creases specifically the principal lines were extracted and teeth were then extracted from the segmented face images respectively. The extracted lines were characterized by discrete Fourier transform (OFT) coefficients of the detected K-endpoint distance matrix which was subsequently used for recognition. On the other hand, teeth recognition was based on the magnitudes of the extracted DFT coefficients of th~ teeth radii signature. In the voice recognition scheme, the subjects were requested to pronounce a uniform word. The recorded signals were pre-processed and the needed information extracted using linear predictive coding (LPC) technique. The derived coefficients were used as input into the ANN scheme for classification. Having successfully developed a unimodal system for each biometrics deployed, the features were then integrated in the second phase at the score level to obtain the proposed multimodal system used for home appliances control. In the unimodal systems, average recognition rates of 100%, 87.9% and 65.7% were attained for palmprint, teeth and voice recognition systems respectively. On the other hand, an average recognition rate of 99.1 % was obtained for the proposed multimodal biometrics system. The developed multimodal system was later applied for restricting unauthorized users from the operation of home appliances which was basically controlled by hand gesture recognition. The control mechanism achieved 100% accuracy. Experimental results of this scheme have demonstrated the possibility deployment in other areas such as mobile devices, computer security, forensic security amongst others.