The estimation of the trajectories of articulator movements from a given acoustic speech signal is known as the acoustic‐to‐articulatory inversion. In this work, it is shown that all acoustic frames are not equally important while performing the acoustic‐to‐articulatory inversion. A frame selection strategy is proposed, where the weight of each acoustic frame is determined for the articulatory inversion task. These weights are finally used in an optimization framework using a generalized smoothness criterion to estimate the articulator movement. We find that the accuracy of the estimates of articulator trajectories improves in terms of both average rms error and correlation with the measured reference trajectories using the proposed frame selection strategy.