A neural network approach to pulse radar detection

Abstract

A .-approach using a multilayer feedforward neural network to pulse co .... resslon Is presented. The 13-element Barker code and the maximum-length sequences (m-sequeuees) with length'llS, 31, a nd 63 bits were used as the signal codes, and four networks were Implemented, respectively. In each of these networks, the nwmer or luput units was the same as the signal length wblle tbe nuder or hidden units was three and t he nuder of output unit was one. In tralning each of these networks, the backpropagatlon learnlng was used and tbe nunmer of training epochs was 500 . Using this approach, a more than 40 dB output peak slgnal-ta-sidelobe ratio can be achieved. Moreover, these rault·tolerant neural networks can provide a robust means for pulse radar detection.

Improving performance in pulse radar detection using Bayesian regularization for neural network training

“…A novel approach of using a multilayer feedforward network was proposed by Kwan and Lee in [4] which resulted in robust performance compared to the previously mentioned techniques. The method used for training the network was Backpropagation (BP) algorithm and the signal code used was Barker code which has the sequence [1, 1, 1, 1, 1, −1, −1, 1, 1, −1, 1, −1, 1].…”

A comparison between functional networks and artificial neural networks for the prediction of fishing catches

“…If the network does not reach a solution, more hidden nodes may be needed; if it does, we can try to eliminate a node. On the other hand, a large number of neurons may imply that the network is simply memorizing training sets, which means that there is very little generalisation and increased computing time [14]. -Training data: We could use all available patterns to train the red, but this is neither necessary nor recommendable.…”

A Novel Application of Support Vector Machines to Detect Targets

“…This property was first exploited in [5], and subsequently by several researchers [6]- [10]. The objective in these papers is to make the ANN approximate an ideal autocorrelation sequence; i.e., peak when the autocorrelation lag is zero, and zero elsewhere.…”