Firstly, the nuclear magnetic resonance technology (NMR) and the conventional test method were used to characterize the pore-fracture system. The effectiveness of the NMR technique was verified. Secondly, mercury intrusion porosimetry (MIP), nuclear magnetic resonance under different confine pressure (DP-NMR) and dynamic permeability tests (DP-P) were conducted on the four groups of middle and high rank coal samples, so the compressibility of matrix, pore and fracture of those samples were studied. The matrix compressibility coefficient (C-m) was calculated by calibrated MIP data, and compressibility coefficient of pore (C-p) and fracture (C-f) were derived based on the permeability and T-2 spectra measured under different confine pressure. Finally, the applicability of the test results is compared and analyzed. The conclusions are as follows. 1) Those samples can be divided into pore developed samples and fracture developed samples based on NMR technology. The variation of C-m in values (0.74-0.81x10(-4) MPa-1) among the samples is much smaller. Meanwhile, the values gradually increase with the increase of moisture and vitrinite contents during the same grade. 2) The C-f based on DP-P tests is much larger than the C-p, which is mainly controlled by the degree of fracture development. The stress sensitivity of fracture is significantly larger than seepage pore and adsorption pore based on DP-NMR tests. Compared with the pore-developed samples, the compressibility of adsorption pore in fracture developed samples is not significantly changed with stress. 3) Different test media and calculation methods result in the compressibility obtained by the permeability is larger than the compressibility obtained by the T-2 spectrum in the same pressure range. The porosity sensitivity coefficient indicates that the two methods were consistent when calculating the compressibility of the pore developed sample.