The EM algorithm for a Gaussian mixture model is applied to the separation of peaks that overlap one another on a chromatogram. However, the number of overlapping components is unknown. Thus, a suitable number of models is set, and the fitting of model parameters is performed while an actually measured signal is appropriately divided for each model by the EM algorithm. Then, when a solution converges, a determination is made as to whether a peak-like waveform is present in a residue signal that is not divided. When the peak-like waveform is present, a peak model is added. The EM algorithm is executed again. In the M step, optimization is performed using, not only a simple Gaussian function, but also a modified Gaussian function assuming a tailing. In the M step, the estimation of a spectrum assuming a chromatogram and the estimation of a chromatogram assuming a spectrum are repeatedly performed.