An efficient algorithm for lossless compression of IEEE float audio

Abstract

Audio data in the IEEE float format is used by the professional quality audio editors as the format of choice during the editing process and also for storing the intermediate results. The proposed algorithm transforms the floating-point values into integer values, in a completely portable way across architectures and compilers. The transform maintains all the important properties of the original values, such as magnitude relations, linear temporal relations and, most importantly, compressibility. It produces a sequence of integers and an additional binary stream used for lossless reconstruction of the original values. Furthermore, the proposed algorithm can be successfully applied to other areas such as medical, science and space data compression. Our algorithm obtains on average 7% better compression than the existing multimedia-aware compression algorithms, while being also faster.