The on-load tap changer is the core device for the power transformer to complete the voltage regulation operation. It has the characteristics of current-carrying sliding electrical contact. The tapping process is prone to wear, sparks and ablation, and it is easy to cause the performance of the contacts to decrease. Thus, this paper combines the actual structure characteristics of the on-load tap changer contacts to build an on-load tap changer contact simulation experiment platform. Based on the experimental platform, a high frequency current transducer is used to collect the high frequency current signals of the normal state, loose state, wear state and fault state of the contacts. Then, the wavelet packet transform is used to denoise the high-frequency current signals in different states of contacts. Finally, the ensemble empirical mode decomposition and Hilbert transform are used to decompose, feature extraction and analysis of the denoised high frequency current signals in sequence. By analyzing the characteristic change process of the high frequency current signal of the on-load tap changer contacts in different states, the derived characteristics of the on-load tap changer contacts from the normal state to the fault state can be obtained. The research results can be combined with the vibration signal fault analysis of on-load tap changer to improve the condition monitoring and fault diagnosis accuracy of on-load tap changer.