The Effect of Distance Similarity Measures on the Performance of Face, Ear and Palm Biometric Systems

Distance or similarity measures are essence components used by distance-based recognition techniques. Since the Euclidean distance function is the most widely used distance metric in PCA and LDA recognition systems , no empirical study examines the recognition performance based on these two methods by using different distance functions, especially for biometric authentication domain problems. The aim of this project is to investigate whether the distance function can affect the PCA and LDA performance over different biometrics datasets. This project helps the researcher to identify suitable distance measures for datasets. Our experiments are based on three different types of biometrics datasets containing face, ear and palmprint data with four different distance functions including Euclidean, Manhattan, Mahanoblis and Cosine similarity distance are used during PCA and LDA classification individually. The presence of statistically significant performance differences is assessed using McNemar's Test.