Motor current signal analysis (MCSA) based on fast Fourier transform (FFT) has its inherent drawback such as low resolution in frequency domain, thus, the detection performance becomes inaccurate for incipient broken rotor bar in squirrelcage induction motor. In this paper, a highresolution spectral analysis method is proposed to solve this issue. Firstly, data preprocessing for single phase stator current is achieved by Hilbert transform and discrete wavelet transform (DWT). Then, extended Prony algorithm is utilized for preprocessed signal qualitative and quantitative analysis. The steadystate stator current of 3kw squirrelcage induction motor can be analyzed under different fault degrees and different operating conditions. The comparison is conducted with FFT. The experimental results shows the effectiveness and superiority on incipient broken rotor bar fault diagnosis even for shorttime data sequence.