MP3 Basics: Psychoacoustics

Ten hours of music on one CD. And that’s without any audible loss of quality. MP3 makes it possible. but how does it work?
The core of MP3 is a compression process that filters out unnecessary information. With MPEG audio, filtering out superfluous information means reducing data that the human ear cannot or barely perceives. The basis for this is psychoacoustics. This science is about how the human ear perceives sound and is the key to MP3 technology.

Imagine you are at the disco. Music resounds from huge boxes. This is hard work on the ear, as sound levels of 110 dB and more are achieved. Due to the extreme volume, it is almost impossible to speak unless you are yelling at yourself. In acoustics, this is called masking. To eliminate masking, the sound level of speech must be raised so high that the interfering signal (in this case loud music) no longer covers it.

Psychoacoustics is just one part of MP3 encoding. The audio signal goes through many more stations. In figure 2 you can see the basic structure of an MP3 encoder.

An audio signal passes through a filter bank that divides the signal into individual areas (subbands). At the same time, the audio signal goes through the psychoacoustic model. Here, the masking threshold is determined for each component with the help of the discrete Fourier transform (DFT). The psychoacoustic model specifies, among other things, the maximum allowable quantization error with which encoding can still be performed without the human ear being perceived. To do this, you specify the number of encoding bits that are required to reduce the quantization noise to such an extent that it becomes (almost) inaudible. In the last step, the data, the previously divided subbands, is processed (formatted) in such a way that a stream of bits is obtained that a decoder can decipher.