Ear Recognition Using Kernel Based Algorithm
T Sudarson Rama Perumal, Shilpa Somasundar · 2014
A person's identity can be recognized and authorized using image preprocessing techniques. In traditional biometrics systems, a Person's face, iris and fingerprint are mostly used for recognition. The longevity of the data measured from these parts may change with age, health and makeup. It has been proved that no two person's ear is exactly alike. Human ears have specific characteristics and features that make them unique. The structure does not change much with individual age. Traditional biometric Systems lack in prediction accuracy and the images are with noise. So, to improve this condition a robust ear recognition system is proposed with the following phases 1) Preprocessing of ear image using Contrast Enhancement 2) Feature Extraction using Kernel Principal Component Analysis (KPCA) 3) Classification using Kernel Support Vector Machine Analysis (KSVM). The System will be tested for accuracy with various neural network configurations and with noisy images.