Because of the characteristic frequency ingredient’s peak-to-peak value is small, and close to the fundamental component, it is easily submerged by fundamental component. To solve the problem, this paper addressed a new approach based on digital filter, zoom-FFT, and rotor slot harmonics based slip estimation to detect rotor bar faults online, calculated the zoom-FFT of stator current to improve the resolution of frequency ingredient, then designed a digital filter to filter the fundamental component, introduced the theory of rotor slot harmonics based slip estimation detailed. The method found the characteristic component exactly and correctly distinguished rotor bars breaking and loan oscillation. Result of numerical simulation and test shows that the approach can find the fault signal accurately, solve the interference caused by fundamertal component leaking, and improve the reliability and accuraly of fault detection.