Performance of Biometrics Recognition System Using Multiple Scales Analysis

Biometrics trait can recognize individual based on the special characteristics unique to his/herself. New soft biometrics trait was introduced such as androgenic hair to overcome problem where the face of individual was hidden by the camera. In this work, performance of biometric recognition system was analyzed further using multiple scales methods such as the difference of Gaussian and discrete wavelet transform using Haar and Daubechies order2. The difference of Gaussians was built using 6 octaves and 4 levels in each octave while discrete wavelet transform was built using nine levels of decomposition. There were two databases of biometric characteristics, androgenic hair pattern and face. Total of 400 images were studied from each database and performance results were calculated. The best performance of accuracy recognition was obtained, 85,25% for androgenic hair database and 92,5% for face database.