The health indicator (HI) extracted by the traditional phase space warping (PSW) method cannot track the degradation trend of rolling bearings under the variable working conditions. To solve this issue, the improved phase space warping (IPSW) is proposed in this study. From the perspective of ensuring the independence of PSW components and maximizing their information, the highdimensional reconstruction of the phase space phase trajectory distortion and entanglement can be avoided by the integrated average mutual information local minimization and mutual information entropy local minimization between the high-dimensional reconstruction components. The slow-changing damage trend item and the working condition change interference item are decoupled to realize the HI extraction that is independent of the working condition change and reflects the fault evolution trend. Simulation and experimental signal verification results show that the extracted HI with the IPSW method can effectively avoid the influence of speed change and track the bearing damage degradation trend effectively.