Bearing is an important unit in rotary machinery, performance degradation assessment is more effective than fault pattern recognition for proactive maintenance realizing near-zero downtime. Wavelet package decomposition (WPD) can depict signal finer, statistical learning theory (SLT) based support vector data description (SVDD) is a one-class classification method holding excellent computing ability. Here, we propose a bearing performance degradation assessment method based on them. This method uses nodes energy of WPD to compose feature vectors, and just needs normal data to build knowledge database using SVDD, then qualitative degradation for test data can be realized. Application results for bearing with different defect size and its whole life time of accelerated life test validated this method’s feasibility and validity.