|
Abstract
|
The paper addresses the narrowband direction of arrival estimation problem in the presence of multiplicative noise, namely, the local scatterers affect the propagation channel characteristics. Deep learning-based neural network, composed of autoencoders and softmax layer, has been trained to spatially filter the wave with desired reception angle. When the array receives at the desired angle, the filter warns, its output rises to one; otherwise it falls to zero. The filter’s response in different scenarios has been investigated by performing various simulations. It has been observed that for the vast majority of scenarios, the filter can detect the direction, regardless of the exception for channels with severe fading
|