To improve the measurement accuracy for the Raman spectrum of micro-substance is one of the difficulties in Raman analysis technique, which is prominent when measuring the micro-substance in organism under the background of high fluorescence. According to the burst and dispersed characteristics to Raman spectrum peak, the fluorescence background and noise fitting function was given in this paper respectively. The location of Raman spectrum peak and interference signal filtering had been realized through monitoring A-type uncertain function in the overall fitting deviation. The spectrum section had been divided according to the spectrum peak’s location further, and the spectrum peak signal had been enhanced through weakening the non-spectrum peak signal within monotone interval. Compared to other fluorescent processing methods, this method could split the overlapped characteristic peak accurately, however, the height of the characteristic peak will not be lowered and more sensitive semi-spectrum peak index could be offered. The experiment indicates that while processing the skin Raman spectrum, such method could obtain the affiliation Raman spectra of the amide I-belt, ceramide and CO ester in helical conformation. What’s more, the measurement model with the content of water-soluble sugar (rice leaf) had been established with the data after being processed by such method, and it had been found that the precision is superior to that of the model established by using the processing data of wavelet decomposition, fitting of a polynomial and nonlinear least square method.