| 摘要 |
A two-layer hierarchical audio bit stream can have a frame-based structure for the base layer bit stream and can be decoded independently from a higher layer and the decoding can start following every sync header. In the extension layer bit stream the frame structure may not be reflected on bit stream level, i.e. while still a fixed number of samples are represented by a certain portion of the bit stream, there is no way to find in the raw bit stream the boundaries between adjacent frames by analysing the stream of bits. To facilitate seek operations with such highly compressed extension-layer data, the header of the extension layer bit stream comprises an FAT table with seek target positions. Because there are fewer entry points in the enhancement layer than sync headers in the base layer, a re-synchronisation and some base layer frames are required to start decoding of the enhancement layer and to generate the full audio quality. Three seeking ways of seeking are described, of which each one offers a different compromise between seeking accuracy, re-synchronisation latency and audio quality. |
