Output details

Data reduction of audio by exploiting musical repetition

<15> This paper presents an entirely novel approach to compressing music audio files. The method works by coding single instances of often-repeated musical blocks (bars, sequences, riffs, choruses, etc.) and representing subsequent instances as variants rather than full versions. The algorithm searches for the most effective pattern-exploiting structure for each piece of music. Although fairly resource-hungry, a single run would code each track ‘for ever’, which is ultimately quite efficient. For some music genres, and some pieces in particular, the compression ratios are very impressive and for practically every music file, the process is ultimately worthwhile.