Aiming at the shortcomings of the traditional model reference adaptive observer in the lowspeed range of interior permanent magnet synchronous motor, including the sensitivity to changes in motor parameters, low accuracy of position detection and poor robustness, an improved sensorless control strategy was presented. Instead of using PI regulators as adaptive mechanisms in traditional model reference adaptive observers, a construction of the sliding surface by using the difference between the reference model of the motor body and the corresponding adjustable model was proposed, and the fuzzy controller was used to adjust the sliding gain adaptively to suppress the chattering of the sliding motion. The simulation model was built in the MATLAB/Simulink environment. The simulation results showed that both the estimated speed and the estimated rotor position could track the actual speed and rotor position when the external disturbance and the motor parameters changed, indicating the validity of the proposed method.