Abstract: It is very important to classify sound speed profile for analyzing and understanding its variation.In order to eliminate the artificial effect in the classification,self-organization manage feature(SOMF) neuron network is introduced into the work.For the efficient appli-cation of SOMF in the classification,integrated with character of sound speed profile,the expression ways of sound speed profile,the determination of neuron cell number,and the definition of adjacent area neuron cell and the construction of the network in the network are studied.By trainings and experiments of the network,the effect of training times for the frame and performance of the network are analyzed and discussed.Finally,a SOMF network is built which is used for classifying 3 teams sound speed profiles.The results are compared with the results obtained by custom method and proved to be right and credible.