In this study, an accelerometer was installed on a computer numerical control (CNC) grinding machine, and the vibration signal of the accelerometer was processed by short-time Fourier transform (STFT) to observe the correlation between the wear behavior of the grinding wheel and the characteristic frequency during the grinding process. Two sets of accelerometers were adopted to collect the vibration signal, which was transformed to a time–frequency diagram by STFT. From the time–frequency diagram, it was deduced that the frequency bands of the grinding wheel and the workpiece were 500–700 Hz and 1200–1280 Hz, respectively. According to the comparison between the frequency band and the time–frequency diagram of the sensing signal at the workpiece with surface roughnesses Ra and RZ, when the spectral energy increased, the surface roughness of the workpiece increased significantly, the self-sharpening was completed at the same time, and the increased surface roughness of the workpiece was maintained.