One of the most challenging scenarios in bearing diagnosis is the extraction of fault signatures from within other strong components which mask the vibration signal. Usually, the bearing vibration signals are dominated by those of other components such as gears and shafts. A good example of this scenario is the wind turbine gearbox which presents one of the most difficult bearing detection tasks. The non stationary signal analysis is considered one of the main topics in the field of machinery fault diagnosis. In this paper, a set of signal processing techniques has been studied to investigate their feasibility for bearing fault detection in wind turbine gearbox. These techniques include statistical condition indicators, spectral kurtosis, and envelope analysis. The results of vibration analysis showed the possibility of bearing fault detection in wind turbine high-speed shafts using multiple signal processing techniques. However, among these signal processing techniques, spectral kurtosis followed by envelope analysis provides early fault detection compared to the other techniques employed. In addition, outer race bearing fault indicator provides clear indication of the crack severity and progress. (C) 2017 Elsevier Ltd. All rights reserved.