Adaptive Recursive Multi-Sensor Systems (ARMSS) with Error Pattern Recognition

Bal N. Agamata

Abstract:	A multi-dimensional signal processing network, comprised of an adaptive recursive multi-sensor system (ARMSS) with error pattern recognition, is structured as a three- layer neural net. The first layer consists of parallel smoothing adaptive recursive filters. Data fusion is accomplished by a second layer of adaptive linear combiners. A parallel set of filters and one combiner represent one backpropagation network (BPN). The last layer, an error pattern recognition machine (EPRM), contains counterpropagation networks (CPN), which are nearest-neighbor pattern classifiers. The ERPM assesses error patterns, derives system weight vectors, and estimates noise and error variances. The BPN excitor-y synapses are modeled by the recursive filters, with one inhibitory or excitory synapse provided by the EPRM. Initially trained off-line, ARMSS is an on-line self-learning neural net.
Published in:	Proceedings of the 5th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GPS 1992) September 16 - 18, 1992 Albuquerque, NM
Pages:	1134 - 1144
Cite this article:	Agamata, Bal N., "Adaptive Recursive Multi-Sensor Systems (ARMSS) with Error Pattern Recognition," Proceedings of the 5th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GPS 1992), Albuquerque, NM, September 1992, pp. 1134-1144.
Full Paper:	ION Members/Non-Members: 1 Download Credit Sign In