Due to critical role of bearings in global vibration levels of rotating machines, vibration-based fault detection for rolling bearings is considered as one of the most common and reliable approaches in machine condition monitoring. In line with the purpose of fault detection, many diagnosis methods aim to identify the fault repetition period in the vibration signal measured from a system with suspected faulty bearing. Captured vibration signals, however, may suffer from heavy background noise due to environmental conditions. In fact, a signal enhancing/de-noising remains a crucial step in a proper rolling bearing fault diagnosis process. In this paper, a novel rolling bearing fault diagnosis method based on Recursive Autocorrelation (RAC) analysis and autoregressive (AR) signal modelling is proposed and validated for fault detection in rolling bearings. The results presented here show superior diagnosis results compared to traditional envelope analysis.