Research Journal of Engineering and Technology

Mandeep Singh Walia ¹

Affiliations
1 Panjab University SSG Regional Centre, IN

Abstract

Biometrics is constantly evolving technology which has been widely used in many identification applications. Biometric modalities are divided into two parts as unimodal and multimodal. In the real world applications, unimodal biometric modalities have many problems and under highly unconstrained conditions, it is not possible to get desirable results. Thus the researchers have moved towards the development of multimodal biometric system. In this regard, multibiometrics is used in high security systems. Fusion in multibiometrics is required to improve the performance of biometric system. Multibiometric is a challenging problem as it is very difficult to predict the optimal fusion strategy due to correlation and fusion complexity. In many multibiometric systems, the different biometric sources (speech and lip movement, two impressions of a person’s fingers) are correlated. In this paper, the limitations of unimodal biometric modalities and the types of multimodal biometrics are discussed. In this work, the fusion strategies in multibiometrics are analyzed with the implementation of speech and iris.

Keywords

Biometrics, Fusion, Iris, Multibiometrics, Speech.

Full Text

     

Mandeep Singh Walia ¹

Affiliations
1 Panjab University SSG Regional Centre, IN

Abstract

Iris patterns are very complex and the combination of complexity with randomness confers mathematical uniqueness to a given iris pattern. Once the image is captured, the iris elastic connective tissue is analyzed, processed into an optical “fingerprint,” and translated into a digital form. The fundamental computing concepts at the core of modern biometrics include image processing, pattern recognition, statistics, basic signalling, and some machine learning models such as knowledge based systems and neural nets. In this paper, methods employed for segmentation as Hough transform with methods employ for iris feature extraction are Hough transform, discrete cosine transform and discrete fractional transforms. In order to extract iris features a normalized iris image is divided into patches. The method is effective compared to existing methods. Performance analyses of different feature extraction methods are proposed. For verification, a variable threshold is applied to the matcher and the False Accept Rate (FAR) and False Reject Rate (FRR) are recorded. Experimental results show that the proposed method can be used for personal identification in an effective manner.

Keywords

Feature Extraction, Iris Pattern, Pattern Recognition, Receiver Operating Characteristic, Segmentation.

Full Text