Perceptual coding

Quantization of MDCT coefficients (the main source of compression) is driven by a psychoacoustic model which decides on how much of the quantization distortion may be introduced to various frequency ranges of the spectrum without causing significant audible artifacts. Psychoacoustic model determines the current shape of the masking profile. By allocating certain number of bits to the quantized coefficients in various parts of the spectrum, the quantization noise is appropriately formed so that a constant safety margin is preserved between the level of quantization noise and the masking threshold [14, 15].

Modification of the hearing threshold due to the presence of strong components masking other components