To accurately extract the fault information of traveling wave head and improve the traveling wave feature extraction efficiency in power system, a new information extraction method for transient traveling is presented. This method uses the principle of intrinsic time-scale decomposition to analyze fault traveling wave signal, decomposes a signal into several intrinsic rotation components and a trend component,and extract the high frequency component for fault analysis.It is also used to calculate the instantaneous phase and frequency of each component, and the arrival time of traveling wave can be detected accurately. Comparison analysis and simulation results show that this method is better in the local feature information retention and decomposition rate compared with empirical mode decomposition and local mean decomposition, and the precision and efficiency of fault location are in range of practical acceptance.