Active acoustic emission (AE) signal estimation is crucial for realizing high-precision bearing fault diagnosis. However, the identification of the bearing fault in the low-speed motor is still a challenging issue. In this article, observerbased low-speed bearing fault identification is investigated, and an observer with adaptive fuzzy switching gain is proposed for improving the accuracy and stability of anomaly identification. First, a normal signal modeling (NSM) is established, based on the Gaussian autoregressive approach integrated with the Laguerre method. Second, a fault observer (FOB) is proposed in the bearing, based on the tracking differentiator technique in different conditions. Third, a fuzzy with an adaptive law is designed to increase the fault estimate accuracy of the FOB. The proposed method instantly increases the signal differentiation when the bearing is working in abnormal conditions. The proposed scheme is robust against suddenly changing the motor speed. Moreover, the fuzzy with adaptive law decay the difference between two crack sizes in the same condition of signal. The fuzzy with adaptive law is designed to guarantees the convergence (robustness) of the proposed FOB. Furthermore, the support vector machine (SVM) is used for residual signal classification. This approach is not only suitable for the bearing fault diagnosis using AE signals but also extendable to the bearing anomaly identification using vibration signals. The proposed algorithm was evaluated experimentally; the results demonstrated that it increases the accuracy of fault identification in the bearing using AE signals.