In order to solve the problem of starting a surface mounted permanent synchronous magnet motor(SPMSM) with unknown initial position, this paper proposed a method for locating the initial position based on discrete Fourier transform feature extraction and polarity identification of forward and reverse pulse excitation. Based on the traditional high frequency voltage injection method of pulse vibration, this paper studied a new method of location feature extraction. The discrete Fourier transform is used to replace the filter to simplify the algorithm structure, and at the same time, this can also avoid amplitude attenuation and phase shift caused by filtering. In addition, for the polarity compensation of the estimated position, this paper designed a forward and reverse pulse excitation method with multi-position check and DC bias compensation, which overcomes the common problem that the current peak discrimination is easily affected by the bias voltage. The proposed method does not depend on the convex polarity of the motor structure, and effectively solves the problem that the surface mounted permanent synchronous magnet is difficult to locate the initial position. Finally, a prototype platform is built for experimental analysis. The results show that the proposed method can detect the initial position of the motor quickly and accurately in the static state, had has strong practicability and robustness.