Active magnetic bearings (AMB) provide contactless levitation of the rotor but are unstable and require continuous closed-loop control of the actuator currents using air gap measurement. Using the gap sensor signal of the rotor under integral feedback control, the response of the rotor to unbalance forces acting on the bearing can be estimated. It is also possible to compensate for the unbalance forces if the operating speed is within the bandwidth of the control system. This paper presents a methodology to estimate the phase and amplitude of the unbalance response in real time in a low-memory embedded platform from the gap sensor signal. The real-time implementation is integrated into an AMB test rig with closed-loop control, and the suppression of the unbalance response from estimated response parameters is demonstrated.