The speed fluctuation signal of electro hydraulic system contains a lot of information about the operating state of the system, which is usually the coupling effect of multiple speed fluctuation sources. Aiming at the feature extraction of speed fluctuation component in square wave signal produced by magnetoelectric speed sensor, a method of speed fluctuation component extraction based on Vold-Kalman timevarying filter was proposed. Firstly, the instantaneous speed signal of the system was obtained by sampling point counting speed measurement algorithm. Secondly, the instantaneous frequency estimation curve, which handled by the STFT and the binaryzation, was taken as the center frequency, and the structural equation and data equation based on the characteristic model of the speed fluctuation signal were constructed. Finally, the time domain waveform extraction of the speed fluctuation component was realized by least square filtering and PCG method. The experimental results show that this method can effectively extract the speed fluctuation caused by the the gearbox and the motor output shaft frequency of the variable speed pump controlled motor system, which can provide a new method of signal processing for online monitoring and fault diagnosis of electrohydraulic system.