Bearings remaining useful life (RUL) estimation is crucial for condition-based maintenance of some rotating mechanics. This research aims to extract an efficient indicator of degradation to establish an RUL estimation model for bearings. For that purpose, an approach is presented based on segmented projection error (SPE) and a fuzzy inference system (FIS). The approach utilizes the time domain features of bearings vibration signals to describe the bearing degradation roughly. Thus, SPE's novel indicator, based on principal component analysis (PCA) and segmentation methods, is obtained by processing the time domain features. The SPE indicator, which effectively tracks bearings degradation, is used to develop a fuzzy inference system (FIS) by the clustering method. For illustration, a benchmark data set is used in this paper. The results show that the novel indicator SPE presented in this paper promises to predict the bearings RUL for maintenance decisions by an early warning.