Abstract: When using the Luenberger observer for sensorless control of surface permanent magnet synchronous motors (SPMSM), the dead time of inverter and rotor spatial structure caused high harmonics in the back electromotive force (EMF), affecting position estimation accuracy of the observer. A new high-order harmonic filter, the second-order complex low-pass filter (SOCLF), has been proposed. This filter has zero gain and no phase difference at the center frequency and effectively suppresses high-frequency signals. Firstly, the back EMF obtained from the Luenberger observer was preprocessed by the SOCLF, which can suppress the higher harmonics in the back electromotive force without losing the fundamental wave signal. Secondly, the phase-locked loop extracted position information from the preprocessed back EMF signal. Finally, an experimental platform has been built to analyze the harmonics of the back EMF. It is shown that the SOCLF preprocessing suppresses higher harmonics in the back EMF and position estimation error of the motor decreased, which proves the effectiveness of the method.