In spectral analysis such as X-ray photoelectron spectroscopy, there is a problem that it is difficult to decide the number of peaks due to the overlapping peaks or large noise. Bayesian spectral deconvolution is a method of automatically estimating the number of peaks and parameters such as the position, height, and width of each peak by applying model selection framework of Bayesian inference to spectral data. We can obtain unambiguous and more reliable results independent of analysts using this method. In this study, we improved the design method of the model used for Bayesian spectral deconvolution and showed that the more reliable result than the conventional model was obtained.
