Conventional partial discharge(PD) fault pattern recognition using ultra-high frequency(UHF) method has strict requirements of sample rate, transmission bandwidth, and processing speed. To solve this problem, we proposed a new method based on signal energy spectrum for PD pattern recognition. In this method, by extracting energy of PD's UHF electromagnetic signals in different frequency bands, we built the energy spectrum sequence as a criterion for pattern recognition which could greatly reduce the recognition's demands for sample rate, bandwidth, and processing speed. Through PD experiments of five typical insulation defects, we obtained some basic recognition judgment that could prove the proposed method to be effective. Plus, using the method, we established a corresponding signal processing system based on radio frequency(RF) theory. Then through PD experiments, the recognition judgment of this system was also proposed. Further experiments indicated that the system has correct rate of 90% in recognition.