Since H.264/AVC achieves a high compression ratio by reducing spatio-temporal redundancy in video sequences, the payload of a single packet can often contain a whole frame encoded by H.264/AVC. Therefore, the loss of a single packet does not only cause the loss of a whole frame, but also produce error propagation into succeeding frames. To deal with this problem, in this paper, we propose a novel frame error concealment method for H.264/AVC. First, the proposed method extrapolates motion vectors from available neighboring frames onto the lost frame. Then, inter-modes of all macroblocks in the lost frame are adoptively estimated by using the extrapolated motion vectors and features of H.264/AVC. Experimental results exhibit that the proposed method outperforms the conventional methods in terms of both objective and subjective video quality.