In the finishing stage of thin-walled parts, due to the tool overhang, the workpiece stiffness is low, so it is easy to deform during machining and cause flutter. Therefore, a reliable standard is needed to monitor the machining state and determine whether the machining parameters are reasonable. In this study, the sound pressure data including chatter phenomenon are collected and analyzed. The time domain effective value and power spectrum of frequency domain are analyzed. The characteristics of different states are compared, and these characteristics are used as the basis for monitoring. When the flutter occurs, the energy concentration shifts at the frequency band. After the wavelet packet decomposition, the characteristic quantities reflecting the feature are constructed. The wavelet transform time-frequency map is used as the state judgment basis, and the correlation threshold is set up by off-line analysis. After setting multiple standards to meet the requirements of the time domain effective value and the frequency domain energy ratio threshold, the eigenvalues are calculated and the processing status is judged. The flutter phenomenon can be identified accurately, and the acoustic pressure signal can reflect the flutter characteristics. After the threshold setting, the invention can provide a judgment standard for the on-line monitoring of the subsequent processing and avoid the damage to the workpiece or the machine tool due to the unreasonable selection of the processing.