Content Based Packet Loss Recovery for Classical Music Transmissions over the Internet

摘要

Various techniques have been proposed to deal with the problems created by packets loss. The error concealment is one of the possible techniques. Most of the researchers focused on error concealment techniques in general or specifically for human speech. These techniques are not efficient for music audio error concealment because the structure of the music signal is different from the audio speech. This paper proposes a novel error concealment scheme for the classical music. The scheme can handle the loss of two consecutive packets containing more than one musical note onsets. The sender report containing positions of note onsets and the cluster information of the signal sections is sent to the receiver. The cluster information is used for the error concealment when the lost packet has more than one onset or for the two consecutive packet losses. Listening tests are performed to evaluate the efficiency of the proposed scheme.