Rolling Element Bearing is one of the most common mechanical components to be found in critical industrial rotating machinery. Since the failure of bearings will cause the machine to malfunction and may quickly lead to catastrophic failure of the machinery, it is very important to detect any bearing deterioration at an early stage. In this thesis, novel signal processing methods based on Acoustic Emission measurement are developed for bearing condition monitoring. The effectiveness of the proposed methods is experimentally demonstrated to detect and diagnose localised defects and incipient faults of rolling element bearings on a class of industrial rotating machinery – the iGX dry vacuum pump. Based on the cyclostationary signal model and probability law governing the interval distribution, the thesis proposes a simple signal processing method named LocMax-Interval on Acoustic Emission signals to detect a localised bearing defect. By examining whether the interval distribution is regular, a localised defect can be detected without a priori knowledge of shaft speed and bearing geometry. The Un-decimated Discrete Wavelet Transform denoising is then introduced as a pre-processing approach to improve the effective parameter range and the diagnostic capability of the LocMax-Interval method. The thesis also introduces Wavelet Packet quantifiers as a new tool for bearing fault detection and diagnosis. The quantifiers construct a quantitative time-frequency analysis of Acoustic Emission signals. The Bayesian method is studied to analyse and evaluate the performance of the quantifiers. This quantitative study method is also performed to investigate the relationships between the performance of the quantifiers and the factors which are important in implementation, including the wavelet order, length of signal segment and dimensionality of diagnostic scheme. The results of study provide useful directions for real-time implementation.