Distribution Estimation Applied to Face Recognition as a Simple and Robust Solution

In this paper a simple and robust solution for face recognition in small groups of people is proposed. After acquiring the characteristic points, describing eyes, nose, lips and the shape of the face, the density function is estimated from them. When the function is known, its values are calculated in the characteristic points. These values are used to compose the feature vector for person description. This vector is then used for classification using Euclidean norm. Both methods gave satisfactory results and the second one even gave less Error Rate. Authors propose a solution to minimize False Acceptance Rate to the level where the algorithm may be applied in human identification and verification solutions where this factor is more important than the reduction of False Rejection Rate.