Early fault detection leads to faster and more effective maintenance, avoiding serious damage in machines and increasing the reliability, security and fault tolerance in industrial scenarios. It is observed that a vibration signal of an onset fault in different parts of the bearing (inner race, outer race and ball) describes certain stable profile among their respective cycles. A methodology to estimate the stability of two dynamic handwriting signatures has been proposed using the Direct Matching Points (DMP) of the elastic align function Dynamic Time Warping (DTW). This methodology is evaluated in this paper and extrapolated to machinery vibration signals affected by an early bearing fault. Results confirm the feasibility to study the stability in both free-fault and fault bearing signals to detect automatically early faults in the bearings by using recorded vibration signals.