Representation and Compression: audio

• Audio can be music, often recorded on sheets which gives a very compact representation, or performed by humans (adding in their own unique interpretation and skill) or computers (often, deterministically). The ISO standard for music is called SMDL and is based on the HyTime standard.
• Most audio, however, is represented as a recording of sound.
• Analog sound is often stored on magnetic tapes.
• Digital audio can be stored directly using Pulse Code Modulation, where many bits (e.g., 16) are used to represent amplitude for each of many (e.g., 44 thousand/second) samples --- this is used for audio CDs.
• To save space, digital audio can be compressed.
• ADPCM is Adaptive Delta (Differential) PCM. It uses fewer bits per sample (e.g., 4) by allowing changes in scale (e.g., 2, 4, 8 times the scale used in PCM) and by recording differentials from one sample to the next, instead of the specific value for each sample.
• ADPCM is used in CD-I and CD-ROM/XA. There are 3 important levels of quality, that vary in terms of the number of samples and frequency range covered. Type A is good for FM quality, B for AM quality, and C for voice or telephone quality.
• MPEG is the most sophisticated compression scheme for audio and can often reduce space to 1/6th with no apparent loss in quality. But this takes extensive processing, and requires special hardware for encoding and decoding.