The automated approach to sound compression

by David Bradley

Could cellular automata be the hi-fi enthusiast's new best friend? Is it time to say MP3 is out? A new lossless compression system for digital audio files is on the horizon thanks to work by Japanese physicists working with the rules of cellular automata.
  If you are a music fan with a computer then chances are you have come across the MP3 format for compressing digital sounds using a Fourier transform. A five-minute song, which would take up 55 megabytes of space on a compact disk is squashed down to just five megabytes or so in the MP3 format. The format uses a computer algorithm to strip out data from the original sound file that would not normally be heard on all but the highest of hi-fi music systems.
  However, the MP3 approach is not perfect. Because it strips out some of the audio data from the sound file, it is said to be a 'lossy' system because there is a loss of information and so aural quality. Hi-fi enthusiasts and scientists who work with sound files in their research would much prefer a 'lossless' compression system.
  Writing in Phys Lett A, recently Masato Wada and Jousuke Kuroiwa of Hiroshima University, and Shigetoshi Nara of Okayama University, Japan describe how they have devised a simple rules system that allows an audio file, whether spoken word or music, to be compressed using the dynamics rules of cellular automata. The process developed by the Japanese team can be used to reproduce completely the sound using only two rules in one-dimensional cellular automata with no loss of information. According to Kuroiwa and colleagues chaotic phenomena have become increasingly well understood. But, complex systems with large but finite degrees of freedom - the sound of a symphony, a heated debate or a searing guitar solo, for instance, are less easy to define. However, rule dynamics, the team believes could be used to generate a large complex system using. So, just as fractal visual patterns that look like the repeating fronds of a fern or the valleys and peaks of a mountain-scape can be generated with a simple set of rules repeatedly applied so too might a set of rules be used to define the sequence of frequencies and amplitudes that make up a sound.
  The team has now defined a set of rules based on the rule dynamics of one-dimensional cellular automata. The cellular automata have two states and three neighbors so are described as "1-2-3 CA".
  Possible scientific applications of the reproducible compression approach might be in spotting characteristic features of a digital sound or evaluating complexity of given data in the sense of Kolmogorov complexity. However, the compression of sound could ultimately have a more popular appeal for the wide dissemination of audio for entertainment purposes too. Listen carefully.