Comparative analysis of some neural network architectures for data fusion

In this paper data fusion is considered within the general framework of perception, its different characterizations are exhibited and an implementation of data fusion with neural networks is proposed. In this setting, the various characteristics of fusion algorithms yield, in a natural way, different design alternatives for the architecture of the neural network. Finally, these alternatives are summarized together with comparative results. This paper validates the use of neural networks for data fusion, and provides a design framework for future work.