Coal pore systems can be commonly classified as diffusion pores, permeation pores and cleats. The classification accuracy influences the coalbed methane (CBM) migration processes from diffusion to permeation and then to outflow, and finally affects the predicted CBM recoverability. To classify coal pore systems precisely, measurements of nuclear magnetic resonance (NMR), mercury intrusion porosimetry (MIP), and nitrogen adsorption isotherm (NAI) are conducted in this paper, and then a comprehensive classification method is proposed. The following cognitions are achieved. NMR spectra can be divided into three categories of three-peak, single narrow peak, and non-three/non-single-narrow peak spectra. The former two categories can be directly used to identify coal pore systems as one peak representing one pore system, and pore systems of the last category can be distinguished by using cumulative amplitudes at the fully water-saturated and centrifuged conditions. Fractal theory suggests that the dividing radii of diffusion–permeation pores obtained by MIP and NAI are quite close, which indicates that the two methods are both effective and accurate. Comparisons between mercury intrusive and cumulative amplitudes indicate that the classification results obtained by measurements of MIP and NMR are similar, which can be a base for transforming transverse relaxation time to pore radius. As a result, the dividing radius of diffusion–permeation pores is about 65 nm, and that of permeation–cleat pores is approximately 600–700 nm.