How does audio compression work?

Audio compression

Audio compression is based on psychoacoustics and uses compression methods that are based on human hearing behavior. For this purpose, the human hearing characteristics are taken into account with the hearing threshold, as is the sluggishness of the ear, which is expressed in the technique of masking.

Audio compression is also about reducing irrelevance. This means that tones of the audio signal that the ear cannot perceive are not processed either. This technique is implemented in most audio compression schemes.

Lossy audio compression

The audio compression can be lossy compression or lossless compression. In general, with lossy audio compression, the signals are transformed into the frequency level using Fourier transformation, DCT transformation or FFT transformation and divided into individual frequency bands using subband coding. With this technique, each frequency band can have a variable quantization and sample depth. Frequency bands with high audio density and dynamics can be digitized with high resolution, while others can be digitized with lower resolution. Well-known lossy compression methods are MPEG, MP3, Audio Code Number 3 (AC-3), Adaptive Transform Acoustic Coding (ATRAC), TwinVQ, Ogg Vorbis, Advanced Audio Coding (AAC), aacPlus and mp3pro.

The lossless audio compression

The lossless compression aims to reproduce the original quality and to reduce the digital audio data in order to save storage space. With lossless audio compression, the dependencies in the audio signals are eliminated in the coder by means of decorrelation. This is done with transformation coding and linear prediction mechanisms such as those used in Adaptive Delta Pulse Code Modulation (ADPCM), Lossless Predictive Audio Compression (LPAC) or Linear Predictive Coding (LPC). The prediction technique, predictive coding, is used to reduce the data of audio signals. Only the difference of one sample to the previous one is saved.

A typical method is adaptive, differential pulse code modulation, in which the differential audio signals are adaptively quantized. The second method used in the coder is the entropy coding of the independent symbols. Other compression methods are comparable to the ZIP technology for data. These methods include OptimFrog and Monkeys Audio.

In addition to the ADPCM standardized by the International Telecommunication Union (ITU), there are other methods for dynamic compression standardized by the ITU, such as the A-Law method and the ยต-Law method.