The key of hydroturbine fault diagnosis is to extract fault feature from monitoring signal. According to signal feature extraction, the combination method of empirical mode decomposition and index energy was used to extract the dynamic characteristic of draft tube. First, the signal was processed by EMD. Then, the index energy of IMF components were computed. Finally, based on the index energy, the multi-scale feature entropy as the characteristic vector of fault pattern recognition was established. The application of this theory was tested with an example of dynamic signal of prototype hydroturbine in non-optimal operating condition. The results showed that the developed method was more convenient and effective in extracting true characteristic parameters of hydroturbine. In addition, this method was appropriate for the analysis of complex and particular dynamic characteristics of hydroturbine.