The information of elements to be measured in the energy dispersion X-ray fluorescence(EDXRF) spectrum is included in the characteristic peak position and the characteristic peak net peak area. Accurate detection of characteristic peaks is the key to energy dispersive X-ray fluorescence spectroscopy. The energy difference between the characteristic X-rays of many low sequence elements is very small, there are many kinds of interference in the process of fluorescence spectrum generation, resulting in serious overlapping peaks of measured X-ray fluorescence data, in this paper, overlapping peaks are taken as the research object, this paper presents a method to deal with overlapping peaks by combining the fourth derivative with the three-spline wavelet transform. The effectiveness of the method was tested by simulating overlapping peaks. The data of X-ray fluorescence spectrum and measured data are verified and analyzed. Firstly, the principle of the derivative method and three-spline wavelet method to decompose overlap is introduced in detail. The higher the derivative order, the more distorted the signal, but it can effectively improve the separation degree of the overlapping peak. The three-spline wavelet transform is weak for the to deal with peak with low separation degree, but it can effectively maintain the peak shape. By simulating the data. Among the three overlapping peaks, the separation degree of peak 1 and peak 2 is R=0.33. The separation degree of peak two and peak three R=0.67, after the fourth derivative there is some overlap in the signal, but the fourth derivative not only retains the peak position of the signal, and the degree of separation increases. Combined with the characteristics of the three-spline wavelet transform, by adjusting the value of the decomposition hierarchy, and reconstructed by scaling up the high frequency signal by a factor greater than 1, the simulated overlapping peaks are decomposed. The number of decomposition layers of the three-spline wavelet is four, and the amplification factor of high frequency is six times. Then, the overlapping spectrum of element K is simulated. The decomposition of overlapping peaks is realized. The simulation results show that the new method can accurately identify the peak position, and the error is within 1%. The applicability of the new method to X -ray fluorescence spectrum overlap peak decomposition is proved. It is verified that this method is feasible to decompose overlapping peaks. The last, is the Ca element X-ray fluorescence spectrum data and Mixed light element X-ray fluorescence spectrum data detected by the CIT-3000SY X-ray fluorescence element logging instrument were processed. Now the decomposition of the overlapping peaks and the peak position error after decomposition are controlled within 1%, with high accuracy. The experimental results show that: The fourth derivative combined with three-spline wavelet transform can effectively separate overlapping peaks. And it is practical to deal with the overlapping peak decomposition of X-ray fluorescence spectrum.