In order to reduce the high frequency chattering of sliding mode observer (SMO) of sensorless permanent magnet synchronous motor (PMSM) and improve the observer characteristics, on the basis of conventional SMO, Sigmoid function is used as the switching function to decrease system chattering. Meanwhile, the combination of the low-pass variable cutoff frequency filter and Kalman filter is used to filter high frequency signals and decrease measurement noise and errors. At last, the speed and position of PMSM are acquired by using phase locked loop (PLL) technique. The effectiveness and feasibility of SMO are verified by MATLAB/Simulink simulation. The results show that the improved SMO has higher quality observation signals than conventional SMO, and the dynamic and static characteristics of the control system are also better.