The objective of this study is to investigate and develop an Artificial Neural Network approach based on vibration and AE signals for the detection, and characterization of wear, damage, and malfunction of an experimental gearbox. Five artificial defects were introduced to the gearbox and these are; (1) tooth face wear, (2) full tooth breakage (missing tooth), (3) clearance or backlash, (5) axial gear looseness, and (5) single internal bearing race wear. The signals, collected from extensive experimentation, were analyzed using time-frequency harmonic wavelet transform, Power Spectral Density (PSD), and four statistical measures of the time domain that captured the salient features of the vibration and AE signals. The results of the time and frequency domain analysis were used in developing a neural network-based estimator for on-line monitoring of gearbox operational condition. The results strongly suggest that vibration and acoustic emission (AE) signals have tremendous promise for machine health monitoring and diagnostics. The proposed technique can be adopted for on-line monitoring of power train and Engine systems, which could be a second phase of this project.