A novel clustering method for Chernoff faces based on V-system

Abstract

The Chernoff faces is a classical method to display multidimensional data graphically in Multivariate Statistics. Based on the V-system, a class of complete orthogonal function system on L 2 [0,1], a new method to quantify the overall features of Chernoff faces is proposed in this paper. Chernoff faces are firstly transformed into the frequency domain by the V-system, the overall features of Chernoff faces are then obtained. It offers a new programmed clustering method for Chernoff faces. Using it, self-adaptive clustering can carry out by computer. This self-adaptive clustering technique can avoid misjudgment caused by human eyes. Especially, when the number of data sets processing is very large, its advantage is more obvious. The results of the concrete experimental tests indicate that the novel method for clustering Chernoff faces is simple, fast and effective.