In neutron reaction cross-section measurements, the prompt gamma ray method is a method of obtaining cross-section data by measuring the characteristic gamma rays emitted by a nuclear reaction, thereby avoiding the interference generated by competing reaction channels. However, the prompt gamma ray method is an on-line experiment with abundant background sources, high background counts of the obtained experimental spectra, and numerous interferences such as weak peaks, overlapping peaks, Compton scattering peaks, and neutron effect peaks of Ge in HPGe, which cause the difficulty in analysing the on-line experimental spectra and the high uncertainty in the results. In this work, we study and summarise the spectrum analysis techniques of the prompt gamma ray method that can be used for measuring the neutron cross-section, and comprehensively consider the physical processes of the formation of different characteristic peaks of the prompt gamma ray method, so as to reduce the uncertainty of calculating the net area of the effect peaks in the process of on-line experimental spectrum processing. The Compton edge, weak peaks, overlapping peaks, and the neutron response peaks of the HPGe detector on-line experiment are discussed and analysed, and the net area of the effect peaks is accurately extracted by combining several reasonable functions to fit the total energy peak, the background, and the interferences. For the net area of weak peaks, this method can reduce the peak area selection caused fluctuation from 30% to less than 1%, and the difference between the fitted value of the net area and the theoretical value is comparable to the statistical uncertainty; for the overlapping peaks’ decomposition, the difference between the results obtained by this method and the theoretical value is significantly lower than 1%. The reliability of the spectral analysis method is simultaneously verified by efficiency curve analysis and goodness-of-fit calculation.