An Independent Evaluation of Subspace Face Recognition Algorithms
An Independent Evaluation of Subspace Face Recognition Algorithms
Dhiresh R. Surajpal and Tshilidzi Marwala
Abstract— This paper explores a comparative study of both the linear and kernel implementations of three of the most popular Appearance-based Face Recognition projection classes, these being the methodologies of Principal Component Analysis (PCA), Linear Discriminant Analysis (LDA) and Independent Component Analysis (ICA). The experimental procedure provides a platform of equal working conditions and examines the ten algorithms in the categories of expression, illumination, occlusion and temporal delay. The results are then evaluated based on a sequential combination of assessment tools that facilitate both intuitive and statistical decisiveness among the intra and inter-class comparisons. The ‘best’ categorical algorithms are then incorporated into a hybrid methodology, where the advantageous effects of fusion strategies are considered.
human face is an extremely complex, dynamic and deformable object with features that can vary considerably and rapidly over time. Skin coverage offers a non-uniform material that is often difficult to model [1] and that can change in response to the effects of emotion, temperature, reflectance properties and perspiration levels, thus creating a large variety and variability within the configurations of facial expression. Another avenue includes time-varying changes by measure of growth, facial hair, effects on the skin due to aging and skin colour changes attributed to ultraviolet exposure. A further complexity is introduced by artefact related changes such as change due to injury and fashion-related issues such as cosmetics, jewellery and hairstyles [1]. Appearance-based analysis, which is one of the oldest approaches, is still said to give the most promising results [2]. Among the most popular publicly available subspace approaches are the classes of Principal Component Analysis (PCA), Independent Component Analysis
Dhiresh R. Surajpal and Tshilidzi Marwala
Abstract— This paper explores a comparative study of both the linear and kernel implementations of three of the most popular Appearance-based Face Recognition projection classes, these being the methodologies of Principal Component Analysis (PCA), Linear Discriminant Analysis (LDA) and Independent Component Analysis (ICA). The experimental procedure provides a platform of equal working conditions and examines the ten algorithms in the categories of expression, illumination, occlusion and temporal delay. The results are then evaluated based on a sequential combination of assessment tools that facilitate both intuitive and statistical decisiveness among the intra and inter-class comparisons. The ‘best’ categorical algorithms are then incorporated into a hybrid methodology, where the advantageous effects of fusion strategies are considered.
human face is an extremely complex, dynamic and deformable object with features that can vary considerably and rapidly over time. Skin coverage offers a non-uniform material that is often difficult to model [1] and that can change in response to the effects of emotion, temperature, reflectance properties and perspiration levels, thus creating a large variety and variability within the configurations of facial expression. Another avenue includes time-varying changes by measure of growth, facial hair, effects on the skin due to aging and skin colour changes attributed to ultraviolet exposure. A further complexity is introduced by artefact related changes such as change due to injury and fashion-related issues such as cosmetics, jewellery and hairstyles [1]. Appearance-based analysis, which is one of the oldest approaches, is still said to give the most promising results [2]. Among the most popular publicly available subspace approaches are the classes of Principal Component Analysis (PCA), Independent Component Analysis