Lossless Audio Coding by Predictor Blending

In the paper potential bit rate savings obtained when implementing lossless audio coding using predictor blending technique are investigated. Three algorithms are analyzed: the simplest Blend-11, Blend-12, and Blend-13 containing 11, 12, and 13 subpredictors, respectively, and tested on 16 widely used benchmark recordings. Savings linked with inter-channel audio dependencies are also considered, appropriate predictor forms the part of Blend-13 method. Results for the last algorithms are also obtained when modeling stage is followed by an advanced arithmetical coder. It is shown that indeed, predictor blending approach to lossless audio compression improves data compaction properties of coding algorithms.