This paper introduces a multi-component time-frequency analysis method

which conducts matching demodulation transform with a Kalman filter. This method builds the model on the local frequency properties of the signal

with model parameters seen as the state variables and the TF ridges treated as the observations. With the state and observation equation established

Kalman filter can be used to estimate the frequency of the signal. Finally

demodulation transform is used to generate a higher energy concentrated time-frequency distribution. The simulation results show our method has a better performance than existing methods under various SNR circumstances.