Abstract
A great deal of research in mode space source localization processing exists. One technique proposed by Shang performs mode filtering of the observed pressure field on a vertical line array, using the calculated mode amplitude distributions and phase differences to estimate source location. Ozard et al. proposed a feed forward neural network using input vectors of mode energy distribution for source localization, a dedicated neuron for each possible source location was employed in their study. Instead of employing a dedicated neuron for each possible search location, the interest of this study is to determine an efficient setup of neurons to properly sample the search space. In MATLAB neural network toolbox, the self-organizing-map function learns to classify input vectors according to how they are grouped in the input space, specifically their distribution and topology. We used this function to examine how input vectors of mode energy distribution for signals at different depths are grouped to determine the sampling grids of the search space in depth. We also compared the neural network performance using input vectors of mode amplitude distribution with those using input vectors of mode energy distribution.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call