The ADPCM (Adaptive Differential Pulse Code Modulation) audio compression system, proposed by M. Nishiguchi, K. Akagiri, T. Suzuki in 1986 for CD-I audio [Sy], has been adopted (with some minor changes [GB]) by Sony and Philips for the audio of CD-I and by Sony, Philips and Microsoft for adding audio to the CD-ROM in the CD-ROM XA standard.
Result of a development process that has been lasting for more than 30 years, since the Philips researchers proposed the delta modulation (DM) for compressing data in telecommunications, ADPCM incorporates the concepts of switched linear prediction and adaptive quantization, that allow to treat in a different way segments of signals with different characteristics. The information theory developed by Shannon in the late forties have in fact shown that signals do contain information which have a certain degree of predictability, furthermore the amount of data required for representing the signal is a function of the predictability of the information it contains.
These findings have deeply influenced the study of speech coding, a field that was intensively studied given the explicit relation with the problems of information transmission. In this respect several digital compression techniques (among which ADPCM) have been devised some of them providing good quality results. The concept of quality, on turn, has been revised, by introducing measurements like the intelligibility coefficient, that takes into account of the main task of speech: communication .
Notwithstanding that many results have been obtained with speech, given the very large musical phenomenology compared with that of a spoken language, there is still a fundamental lack of knowledge concerning the informational content of musical signals. The current digital standard for professional audio (PCM), based on 16 bit linear quantization, does not exploit the signal characteristics at all, and the fact that all segments of music are treated in the same way bears witness that a certain degree of redundancy has been allowed in the PCM coder design to cope with the most demanding situations.
The adoption of ADPCM for high-medium quality music reproduction requires therefore a new approach from the audio engineers side, to be aware for example that quality tests can no more be based on the usual stationary correlated signals, and that the distortion can no more be harmonic, as the encoder does not act on the harmonics of the signal.
Furthermore, during the beta test period of the CD-I some problems have been reported, particularly when correlated signals (like the usual test sinewaves) are processed. This has led many peop]e to ask whether there is some way to improve the coder performance, by adopting different (CD-I specs compatible) coding techniques, like a different prediction strategy.
[ Index | Next Paragraph ]