Based on the characteristics of random, gray and fuzzy information in the process of rolling bearing fault pattern recognition, a new fuse frame is presented by the view of information fusion. The characteristic vectors are calculated from wavelet, amplitude and frequency domain information, and the evidence theory is used. The diagnose result is gained by Hidden Markov Mode. Then, the recognition results from the view of gray and fuzzy are obtained. Results are fused so that rolling bearing fault pattern can be identified from multi-angle of view information. The fuse frame is applied to rolling bearing real test vibration data. The result shows that the class validity probability can be remarkably more advanced from multi-angle of random, gray and fuzzy information than single information.