The research objective is to investigate the feasibility of utilizing the hardware-implemented order tracking technique and angle-order analysis approach to diagnose the flaws of roller bearings, such as inner race defect, rolling defect, and outer race defect, in case of variable rotating speeds. The envelopes of the identical-angular signals that are measured through the hardware are first decomposed by means of the empirical mode decomposition (EMD) technique. The features of bearing flaw are then observed and analyzed on the angle-order distribution as well as the marginal order spectra. The energy levels and the root-mean-square (RMS) values at the corresponding characteristic orders are extracted from the marginal order spectra to form the feature vectors. The support vector machine (SVM) is employed to identify the extracted feature vectors. Through the experimental studies and vibration data analysis, the classification results of the SVM demonstrate the effectiveness and accuracy of diagnosing the different types and levels of bearing defects.