DOA estimation of distributed sources using neural networks
Mehrez Hirari, Kaoru Gotoh, Masashi Hayakawa · 2002
We propose a neural network solution to estimate the DOA of distributed sources. We characterize a distributed source by its center to which corresponds the incident angle and by a set of parameters including its frequency, amplitude, and width. We use an artificial neural network-based algorithm to determine the direction of arrival (DOA) of the source. The problem is mapped onto an artificial neural network by discretizing the parameter spaces to a specified granularity and representing each possible combination of the parameters by a single neuron.