The separation of characteristic spectral signatures plays an important role in the laser induced fluorescence (LIF) system used for monitoring dissolved organic matter (DOM) in polluted water,which has advantages of high sensitivity,fast detection,remote measurement,etc. In this paper,a method of characteristic spectral signatures separation based on radial basis function networks (RBFN) will be presented. Using this method,we separated the spectral components of laser,Raman and fluorescence of DOM from the LIF spectral. The concentration of DOM in polluted water can be retrieved from the separated spectrums.