High voltage circuit breaker is the key of control and protection equipment in power system, in allusion to the deficiency of the fault diagnosis methods, this paper boosts the vibration and acoustic data level fusion and feature level fusion method used in high voltage circuit breaker fault diagnosis. The vibration and acoustic feature level fusion diagnosis method firstly makes acoustic signals collected achieve blind source separation processing through fast kernel independent component analysis (Fast KICA), and extracts the vibration signal and acoustic signal feature vector by the improved ensemble empirical mode decomposition (EEMD). The vibration and acoustic data level fusion diagnosis method firstly builds vibration acoustic joint image, and extracts feature vector by the improved BEEMD. Finally, the feature vector extracted by the two methods are input into support vector machine (SVM) for fault diagnosis. The experiment shows that the proposed method is effective to diagnose the faults of high voltage circuit breakers.