In the process of sensorless control of permanent magnet synchronous linear motor using sliding mode observer, the chattering problem exists in the sliding mode observer itself, and the detent force of the motor will also cause error harmonics in the observed back electromotive force, which will cause the error of estimated speed and position. In order to solve the above problems, the traditional method is optimized in this paper. Firstly, replace the sign function with the sigmoid function to suppress the chattering problem, and then a cross-cancellation second-order generalized integrator is introduced to replace the low-pass filter to suppress the harmonics and reduce the estimated error of velocity and position. Verified by simulation, the optimized algorithm in this paper can effectively suppress the error harmonics, and greatly optimize estimation performance of velocity and position.