Personal Authentication using Face and Palmprint
In this work a new method of personal authentication using face and palmprint images is investigated. The facial and palmprint images can be simultaneously acquired by using a pair of digital camera and integrated to achieve higher confidence in personal authentication. The proposed method of fusion uses a feed-forward neural network to integrate individual matching scores and generate a combined decision score. The significance of the proposed method is more than improving performance for bimodal system. The experimental results also demonstrate that Sum, Max, and Product rule can be used to achieve significant performance improvement when consolidated matching scores are employed instead of direct matching scores. The fusion strategy used in this paper outperforms even its existing facial and palmprint modules. The method proposed in this paper can be extended for any multimodal authentication system to achieve higher performance.
Figure 1: Sample face images from the database employed in the experiments.
Figure 2: Region and feature extraction from a typical palm image.
Figure 3: Convergence of training error from the palmprint and face matching scores; Distribution of genuine and imposter scores from the two biometric.
Figure 4: Comparative performance for user authentication using face and palmprint; Variation of FAR and FRR scores with decision threshold for combined decision.
Figure 5: Distribution of combined decision scores from the two classes.