Implemented Model for Masking Threshold

To produce the masking threshold from the spread function and the masker, we need to know the tonality $t(z)$ and the spreading function $SF(z)$ of the masker. After an idea taken MPEG-1 layer I [2], the following model for masking threshold $M(z)$ is used:

$$M(z) = P(z_m) + SF(z-z_m) - k(t(z))z_m - l(t(z))\ {\rm (db)},$$ (14)

where $z_m$ is the frequency in barks for the masker, $z$ is the masked frequency in barks and $P(z_m)$ is the power of the masker in dB. The functions $k(t)$ and $l(t)$ are experimentally found as a linear combinations of constants for pure noise and pure tones:

$$k(t) = 0.3t+0.5(1-t),\ {\rm and}$$ (15)

$$l(t) = 34t+20(1-t)$$ (16)