The most fundamental problem in radar is the detection of an object or a physical phenomenon. This requires proper discrimination between signal and noise content at the receiver even after the echo containing target information is surrounded by clutter. Traditionally, a series of signal processing operations are carried out to perform this discrimination with varying levels of success. These series of signal processing operations can be supplemented by an Artificial Neural Network (ANN), which is a non-parametric prediction tool with the ability to retain the learning acquired from the surroundings. The Recurrent Neural Network (RNN) is a dynamic ANN which can track time variations in the input patterns. The RNN captures time-varying contextual information and use this knowledge subsequently to make discrimination between adjacent patterns. As viewed from the time domain, the target waveform can also be regarded as a time sequence such that it can be classified using RNN which is suitable for time sequence processing. This work describes the processing steps of signals from pulse radars so that these can be used to train a RNN for use to discriminate between target and false echoes. The experimental results show that the proposed system works effectively while dealing with target echoes surrounded by thermal noise and ground clutter at varying distances [1].

Radar target recognition recurrent neural network clutter processing