This paper is concerned with a generalized p-feature M-class classification problem. A new algorithm for sequential pattern classification, which is based on projections of the p-dimensional feature data onto each feature axis, is developed. During the learning phase, the p-demensional feature data is projected onto each feature axis. For each feature axis, the optimum Binary Decision Tree is constructed using this projected data. During the classification phase, an unknown pattern is classified as follows. In the first place, the selection of the feature to be measured at each stage is predetermined by feature ordering. Each Binary Decision Tree is searched according to the predetermined order. After receiving the result of each search, the classification procedure, by classification rule, either classifies an unknown pattern or searches the next Binary Decision Tree. The error bound of the proposed algorithm is derived and some computer simulated experiments in character recognition are presented. The results are compared to conventional other schemes and show the effectiveness of the proposed algorithm.
