During touchdowns of active magnetic bearings (AMB), the violent collision between rotors and touchdown bearings (TDB) can cause damages to both parts. Orbit response recognition provides a way for the AMB controller to automatically switch the control algorithm to actively suppress the rotor–TDB vibration and promptly relevitate the rotor during touchdowns. A novel method based on Hilbert transform (HT) is proposed to recognize the orbit responses (pendulum vibration, combined rub and bouncing, and full rub) in touchdowns. In this method, the rotor suspension status is monitored by the AMB controller in real-time. When touchdown is detected, the rotor displacement signal during the sampling period is intercepted, and the instantaneous frequency (IF) is calculated by HT. Then, the local variance of IF during the sampling period is calculated, and it is compared with the threshold value. Combined rub and bouncing can be identified for it has the largest local variance. Finally, the mean value of IF during the sampling period is calculated and is compared with the other threshold value. Pendulum vibration can be identified for it has a lower and fixed mean value, while full rub has a larger value. The principle of the recognition method is demonstrated by the simulated results of a thermo-dynamic model. The results reveal that the method is feasible in recognizing the orbit responses and can be implemented in the AMB controller to help switch the control algorithms automatically in case of touchdowns.