In the current fault diagnosis technology of asynchronous motors, frequency resolution for short time data was low and hardware overhead was high. For the sake of solving these shortcomings, a novel approach for rotor fault based on rotating filtermatrix pencil of short time data was proposed. Matrix pencil algorithm was used to accurately estimate the fundamental frequency component in stator current, and the fundamental frequency component was removed by inverse and forward rotation transform. Accurate estimation of frequency and amplitude in stator current was realized with the short time data by using matrix pencil algorithm, which broke through the limited resolution of traditional method based on fast Fourier transform (FFT) analysis. The simulation and experimental verification showed that rotating filtermatrix pencil method was valid for broken rotor bar faults with short time data.