This paper proposes a new diagnostic approach for identifying bearing faults in induction motors, which are common issues that can affect the performance and durability of these motors. Although various methods have been developed to diagnose these faults, we propose a high-resolution technique based on stator current analysis, enabling more effective detection and identification of these anomalies. This new approach leverages a variant of the Root-Prony method, designed to overcome the limitations of conventional periodogram methods, such as the inability to achieve high-frequency resolution with very short acquisition times. These limitations make it challenging to analyze the frequency components associated with bearing faults. In contrast, high-resolution spectral analysis using the Root-Prony method offers enhanced frequency resolution and the ability to detect faults even at low amplitudes. However, the method's high computation time remains a limiting factor. To address this, a zoomed Root-Prony method is proposed, allowing for faster and more precise diagnosis of bearing faults. Simulation and experimental tests are presented in this paper to validate the effectiveness of the proposed method.