Here, Hilbert-Huang transformation (HHT) was used to analyze nuclear explosion and lightning electromagnetic pulses (NEMP and LEMP), these signals had non-stationary and non-linear characteristics. It was found that the energy in a Hilbert spectrum is focused on certain time and frequency regions. Based on the above analysis and discovery, the regional energy ratio of Hilbert spectrum was presented, with it the discrimination of NEMP and LEMP was studied. The test results indicated that using the nearest neighbor pattern classification, the average discrimination rate of NEMP and LEMP is over 90% based on regional energy ratio of Hilbert spectrum.