Blog Contents

Digifon Logo link to main page

12. Audio Coding Algorithms Part 2: DEEP SCIENCE!    6/21/12 by Dave Immer

For those of you who are curious about what goes on under the hood, the following is a brief description. There will be a quiz.

THE ALGORITHM
An audio coding algorithm is a set of rules and instructions that govern the behavior of certain types of digital audio processing. Digital audio bit-reduction schemes such as MPEG Layer II, III and AAC, and waveform prediction schemes such as APT-X, are commonly referred to as a codec (COder-DECoder). The hardware that hosts these software programs is also called a codec.

The coding algorithms are what make ISDN a practical audio transport medium. They also enable smooth audio streaming and quick downloads over IP. So-called “linear audio” as embodied on a standard CD requires 1.411Mbs to deliver “full frequency” 20Hz-20kHz  stereo audio. The problem is, an ISDN circuit is only able to carry 128kbs – about 1/12th the bits needed for linear stereo (1/6th for linear mono). The public internet is also subject to bit-rate pressures. This problem is solved by the coding algorithms.

HOW IT WORKS
In the case of MPEG codecs, most of the bits representing the original waveform get discarded resulting in noise that gets hidden, or masked, by the remaining waveform. It’s a sleight-of-hand procedure based on psychoacoustic models of the human hearing process. Advantage: scalable and low-cost decoders.

In the case of APT-X codecs, a predictive analysis results in the encoder transmitting only the difference between the prediction and the actual sample and then adds the redundant information back at the decoder. Advantage: very low delay and near-lossless coding.

Let me know if you have questions or comments about coding algorithms. Thanks,

-Dave