This paper addresses the issue of classifying local gear faults that depends on vibration signal measurements. Most of the early gears fault detection and diagnosis methods result insufficient results when dealing either with time or frequency domain characteristic. In order to overcome these obstacles, a TimeFrequency based approach implemented on Field Programming Gate Array (FPGA) is proposed. The presented approach combines Bagged Trees Classifier (BTC) with Complex Analytic Wavelet Transform (CAWT) analysis. Moreover, the presented approach benefits from the superior realization nature of FPGA. Extensive simulations and experiments have been conducted in order to demonstrate the efficiency of the proposed approach. Experiments are performed on bevel gearbox with both normal and one missing tooth. The obtained results clarify that the presented approach has a superior classification accuracy rate over the other comparative approaches