Abstract: Acoustic emission(AE) technology has been applied in the field of coal rock fracture analysis, and many beneficial results have been achieved, which provides important guidance for coal rock dynamic hazard monitoring and early warning. However, there is still a need for further research and improvement in the area of coal rock fracture analysis using AE,e.g. novel AE analysis method. Therefore, this study conducted uniaxial compressive loading experiments on sandstone specimens with prefabricated cracks. Meanwhile, the stress-strain data, full-waveform AE data and high-resolution photographs of the specimens were collected simultaneously in the whole loading process. Mel-frequency Cepstral Coefficient(MFCC) was extracted from AE signal in the whole loading process of sandstones by utilizing sound wave analysis means. The advantages of applying MFCC of AE to analyzing sandstone fracture and its underlying mechanism were discussed. The response law of MFCC in the failure process of sandstone was investigated. On this basis, the fracture and failure evolution characteristics of sandstone specimens with prefabricated cracks were revealed, and the MFCC precursors of sandstone failure were obtained.Resultsshow that the evolution characteristics are similar, the variation is close, and the deviations of the variation are small(i.e. from 5% to 15%) even if the MFCCs were extracted from AE signals that were collected by different AE channels and from different specimens. In addition, the MFCCs of AE behave sensitively in the whole loading stage. This indicates that the MFCCs of AE have the advantage of stability and can be used as a characteristic parameter for characterizing the fracture of sandstone. It is found that the MFCCs are obtained without setting a threshold to AE waveform, and its value is determined by the overall shape of AE waveform over a certain period of time including the amplitude, density, and intervals among adjacent AE signals etc. In addition, the overall shape of AE waveform collected by different channels is quite similar. Thus, the MFCCs are capable of well characterizing AE signals, which is the reason why the MFCCs have the advantage of stability. In the stage of sandstone failure, the MFCCs behave with some rise and fall fluctuations in a periodic manner, which corresponds to the periodic fall and rise of the stress curve and the intermittent generation of high AE amplitude. Furthermore, there is a significant positive correlation among the changes of the MFCCs, stress and AE amplitude. These results indicate that the MFCCs are quite responsive to sandstone fracture, the fluctuation of MFCC can characterize the fracture process of sandstone, and the fluctuation amplitude of MFCC can reflect the fracture intensity of sandstone. By analyzing the MFCCs of AE,it is revealed that the evolution of sandstone fracture is in an intermittent manner, i.e. fracture initiates suddenly in a very short time, then fracture propagation pauses for a period of time, and then fracture continues to propagate. Shortly, fracture alternates between propagation and pause. This intermittent feature of fracture propagation becomes more intense as the sandstone approaches instability and failure. With this as a guide, the difference between the peaks and troughs of the periodic fluctuation of MFCC,i.e. fluctuation difference, is used to characterize the fluctuation amplitude of MFCC and describe the intermittent fracture intensity of sandstone. It is found that the fluctuation difference gradually increases as sandstone approaches its strength and jumps twice to its previous value upon failure, which can be recognized as the failure precursors of sandstone.