Abstract: Analyzes the frequency and power spectra of acoustic signal due to the vibration of coal that is pulverizing in a ball mill. The results show that the low-frequency signal contains the information on coal level in ball mill, while the high-frequency signal is caused by the noises of working high-speed motor, exhaust fan and rotating drum of ball mill. Both are in a modulating relation. The acoustic signal is analyzed by Hilbert transform, from which the low-frequency signal is picked out to show the corresponding coal level in rotating drum by an envelope of analyzed signal. The relationship model between coal level in ball mill and relevant envelope of acoustic signal due to vibration is therefore developed to recognize automatically the coal level. The comparison of calculated values from the model with measured values indicates that the recognition accuracy is within ±1.5%.