Abstract: To a non-cooperative receiver, the blind carrier recovery is a key step of the demodulation process for communication signals. In order to increase the convergence speed, expand the scope of frequency acquisition and effectively improve the property of tracking time varying frequency, the variation characteristic of the filter loop bandwidth is gained from theoretical derivation and a carrier synchronization algorithm based on adaptive decision directed extended Kalman filter is proposed for Orthogonal Amplitude Modulation (QAM) signals. The algorithm which based on existing Kalman filter algorithms, adopts adaptive multi-states switching mechanism and regulates the parameter of the Kalman filter adaptively, which are chosen from the coordination principle obtained through the analysis of the experimental test data. It is indicated from the analysis and simulation results that compared with the existing Kalman filter based algorithm and phase lock loops, adaptive decision directed extended Kalman filter algorithm not only enlarges the catching scope, improves the convergence speed and works well in the situation with time-varying doppler distortion, but also has advantages on the tracking performance under larger frequency offset and robustness.