The artificial signal and rolling bearing signal were analyzed by Db4 lifting wavelet combined with T-K energy operator, the best detail signal was obtained based on principle of minimum wavelet coefficient, then its T-K energy spectrum was worked out and feature information was extracted. The approximate information of T-K energy operator was obtained to classify the bearing faults. The result shows:the detail signal of artificial and actual signal can extracted by lifting wavelet, the T-K energy spectrum can be used to extract fault features, and T-K operator approximate entropy can be used to classify experiment bearing signals basically.