An expert target recognition system using a genetic wavelet neural network

In this paper, a target recognition system is presented for target recognition using target echo signals of High Range Resolution (HRR) radars. This paper especially deals with a combination of an adaptive feature extraction and classification using optimum wavelet entropy parameter values. The features are obtained from measured target echo signals using a X-band pulse radar. A genetic wavelet neural network model is developed for target recognition. This model consists of three layers. These layers are genetic algorithm, wavelet analysis and multi-layer perceptron respectively. The genetic algorithm layer is used for selecting the feature extraction method and obtaining the optimum wavelet entropy parameter values. Here, the optimal one of four different feature extraction methods is selected by using a genetic algorithm. The proposed four feature extraction methods are: (i) standard wavelet decomposition, (ii) wavelet decomposition—short-time Fourier transform, (iii) wavelet decomposition—Born-Jordan time-frequency representation, (iv) wavelet decomposition—Choi-Williams time-frequency representation. The wavelet layer is used for optimum feature extraction in the time-frequency domain. It is composed of wavelet decomposition and wavelet entropies. The multi layer perception is used for evaluating the fitness function of the genetic algorithm and for classifying radar targets. The performance of the developed system is evaluated by using noisy radar target echo signals. The test results show that this system is effective in rating real radar target echo signals. The correct classification rate is about 90% for target subjects.