Abstract
The goal of sensory data fusion is to combine the complementary and/or redundant information from many sensors to gain a better description of the environment. Also, in pattern recognition, one is faced with the problem of selecting the best features or to generate a new feature vector of lower dimension, from complementary and/or redundant features. There are many different approaches in operation research to assigning weights, or priorities, confidence levels, etc., to a set of elements. It is worth investigating how these methods could be applied to the sensory fusion problem. This paper introduces a new method for sensory data fusion, and for feature selection, using the Analytic Hierarchy Process (AHP). AHP is well suited to the sensory data fusion problem, since the fusion problem is inherently a hierarchical process. The hierarchical structure in AHP allows multiple aspects of sensors to be compared, where each aspect can be given different weight, and allows multiple levels of comparison of sensors and multiple agents decision making on sensory fusion. Fusion of the same type of sensors, fusion of different types of sensors, and sensor selection can all be dealt with in the same framework.