A novel method of multisensor information fusion was presented for e a rly fault recognition of the pinion bearing in the main reducer of vehicle diver axle. The pinion bearing is vulnerable to deteriorate in actual operation of th e vehicle, which brings disadvantageous influence on the transmission system. Th e early fault recognition of pinion bearing was researched on a test rig in orde r to evaluate its servicelife. In the experiment, various operating conditions of the vehicle were simulated and the information fusion was tested using the fu zzy theory and DempsterShafer (DS) evidential theory. The signals were gathe re d by the acceleration sensors, the noise meters and the infrared temperature sen sor. Experimental results demonstrate that bearing conditions can be identified within a single symptom from redundant information as compared with the methods using the single information source, which makes up for shortage of priori proba bility and meets the demands of recognition of fault symptoms. The study shows t hat the synthetic information fusion method based on fuzzy theory and DS evide ntial theory has excellent reliability to early recognition of bearing failure.