Pattern Analysis & Applications (1999)2:292–311  1999 Springer-Verlag London Limited Serial Combination of Multiple Experts: A Unified Evaluation A. F. R. Rahman and M. C. Fairhurst Electronic Engineering Laboratory, University of Kent, Canterbury, Kent, UK Abstract: Multiple expert decision combination has received much attention in recent years. This is a multi-disciplinary branch of pattern recognition which has extensive applications in numerous fields including robotic vision, artificial intelligence, document processing, office automation, human-computer interfaces, data acquisition, storage and retrieval, etc. In recent years, this application area has been extended to forensic science, including the identification of individuals using measures depending on biometrics, security and other applications. In this paper, a generalised multi-expert multi-level decision combination strategy, the serial combination approach, has been investigated from the dual viewpoints of theoretical analysis and practical implementation. Different researchers have implicitly utilised various approaches based on this concept over the years in a wide spectrum of application domains, but a comprehensive, coherent and generalised presentation of this approach from both theoretical and implementation viewpoints has not been attempted. While presenting here a unified framework for serial multiple expert decision combination, it is shown that many multi-expert approaches reported in the literature can be easily represented within the proposed framework. Detailed theoretical and practical discussions of the various performance results with these combinations, analysis of the internal processing of this approach, a case study for testing the theoretical framework, issues relating to processing overheads associated with the implementation of this approach, general comments on its applicability to various task domains and the generality of the approach in terms of reevaluating previous research have also been incorporated. Keywords: Character recognition; Generic serial framework; Multiple expert configurations 1. INTRODUCTION In the field of pattern classifier design, it is now accepted that the application of more than one classifier (expert) in a framework which allows their combination in such a way as to emphasise the strengths of the cooperating experts and counter-balance their deficiencies and weaknesses can be a very powerful approach in implementing robust and more efficient systems. There are numerous ways in which multiple experts can be combined to bring about recognition enhancement, but a topic that remains to be addressed in this context is the matter of deciding on an optimum combination for implementing a system for a particular task domain [1]. In early work on multiple expert combination techniques, the main topic of debate was the conflict between ‘multiple-expert versus multiple-level’ [2–4] con- figurations. More recently, a broader view has emerged, and many combination frameworks to categorise the various Received: 9 October 1998 Received in revised form: 5 March 1999 Accepted: 29 April 1999 approaches have been proposed (see, for example, Rahman and Fairhurst [5], Xu et al [6], Powalka et al [7], Kittler and Hatef [8], etc.). All of these frameworks attempt to present a unified and comprehensive view of the way in which multiple expert decision combination works, in order to enrich and enhance the quality of the combined decision with respect to the individual decisions of the participat- ing experts. In this paper, one particular multiple expert multiple level decision combination strategy, the serial combination approach, is investigated. No comprehensive, coherent and generalised presentation of this approach from both a theor- etical and implementation perspective has been attempted to date, yet this approach has been implicitly exploited in many different application domains. 2. GENERALISED MULTI-LEVEL MULTI- EXPERT APPROACHES Before examining the serial multiple expert decision combi- nation approach in detail, a brief introduction to the differ-