Abstract: This paper discusses methods of recognizing different kinds: of ship objects by the shape of averaged power spectrum of radiated noise. This paper considers the pre-processing of data first. After applying general correlation method, distance method and weighted distance method, a new method is presented. The basis of this new method i.e. M-m₂ plane partition method is the shape of averaged power spectrum of ship radiated noise may be approximated by Ecs noise model. Such noise model which may be characterized by a set of three parameters f₀, f_m and K or ω₀, a and K is discribed in detail. Controlling effects of the three parameters on shapes of curves of autocorrelation function R（τ） and power spectrum density function G（f） are analyzed. The three parameters f₀, f_m and K describe the following basic attributes of power spectrum density curve of Ecs noise: f₀ approximately is the frequency of its maximum;f_m reflects the degree of sharpness of this curve; K represents certain obliquity of this curve. Methods for extracting these features are also discussed. In M-m₂ plane partition method we use the location of the maximum of averaged power spectrum M and the second order central moment of normalized averaged power spectrum level m² as recognition features, apply piece-linear classifier, get 92% of correct recognition rate for two different types of ship objects.Compared with the another three methods, the performance of M-m₂ plane partition method is excellent.