Hierarchical and Lossless Coding of audio objects in Dolby TrueHD

摘要

Dolby TrueHD is a lossless and hierarchical audio coding format that not only enables compact bit-exact representation of the source multichannel audio signal, but also facilitates low complexity reconstruction of downmixes thereof. The dual objective is achieved by linear transformation of input channels into internal channels coded in the bitstream, via primitive matrices that are exactly invertible with finite precision, such that a subset of the internal channels spans the subpsace of the required downmix. A decoder need only extract and linearly combine this subset of internal channels to reconstruct the downmix presentation. Until recently, TrueHD was primarily employed for lossless carriage of speaker feeds (typically, 7.1ch), in which case downmixes of interest (5.1ch or stereo) could be obtained via time-invariant linear transformations. Concurrent with Dolby Laboratories' efforts towards a next generation surround sound experience, however, the format has been extended to support lossless transmission of audio objects such that renderings (downmixes) of these moving objects to standard speaker layouts are still accessible to legacy decoders. The mandated representation of a continuously varying downmix matrix trajectory is facilitated by the novel paradigm of interpolated primitive matrices and the associated matrix decomposition strategy presented here.