Abstract: Feature analysis and extraction is a key issue of target classification, in this paper, at first, the nonlinear analysis method is used for studying the limit cycle phenomenon of ship noise,and the result indicates that the period-doubling or chaos would happened with limit cycle of vibration noise in phase space,which is the main component of shop noise. Second,the fractal dimension and distribution density ration are used for calculating the strangeness and shape of limit cycle,and a new algorithm of fractal dimension calculation is proposed,At last,the fractal dimension and distribution density ration are chosen as feataure parameters of ship noise for calssifying sea ship and underasea ship by using neural network classifier,The results of experiment on our ship noises show that the nonlinear features based on ship noise limit cycle is effective in classification of two classes of ship targets,The research of this paper gives a novel effective method for feature extraction of ship noise signals.