We propose a new technique with FFT spectrum minimum entropy deconvolution filtering to suppress the periodic narrow-band noise mixed in discharge signal. The principle is the narrow-band noise component will be treated as a new useful signal and the real pure discharge signal as a new noise in frequency domain. In frequency domain

the discharge signal component will be rejected and the narrow-band noise will be kept by minimum entropy deconvolution filtering. The remained noise will be suppressed in the FFT spectrum of the mixed discharge signal

and the pure discharge signal is obtained by IFFT process. The simulation and the experiment results indicate that the proposed method can accurately suppress the periodic narrow-band noise. Compared with the conventional energy ratio method based on FFT

the frequency window to compute the energy ratio is not necessary in this method

and the noise can be suppressed well just by classical threshold value criterion. The method offers a new effective technique for the periodic narrow-band noise suppression when detecting the discharge signals.