A study on several feature selection methods in target classification and recognition

In the paper, based on the analysis to several feature selection methods, such as principle component analysis (PCA), maximal gradient selection and exploratory pursuit are presented. First merits and demerits of several methods are compared. Then to true and false underwater target echo signal, Wigner and Burg features are extracted and selected by those methods. Finally, the selected features are trained and recognized by Fuzzy Adaption Resonance Theory (FART) network to compare the effect of several methods to the two kinds of echo signal. The number of training samples to the number of testing samples ratio is 1 to 4. The results show the two kinds of method, maximal gradient selection and exploratory pursuit are not only less computation but also low dimension. The higher recognition can be achieved by the two methods.