A piezoelectric selfsensing actuator is realized based on a neural network for separating the strain rate signal from the control voltage, since the conventional strain rate bridge circuit is difficult to be adjusted to the balance. A vibration velocity estimation system using a neural network is developed. In the new system, a piezoelectric element embedded in a cantilever composite beam is used as SSA, an adaptive feedback controller is constructed by using the Filtered-X LMS algorithm for the active vibration control. Experimental results show that the estimated vibration velocity of cantilever beam agrees well with the vibration velocity directly measured by a laser displacement sensor, and the new system exhibits good damping performance in the vibration control of the cantilever composite beam.