K-weighting
The frequency filter audio passes through before LUFS is measured: a rough model of how the head and ear colour what we hear.
ITU-R BS.1770, published in 2006, defines it in two stages. First a high shelf that lifts everything above roughly 1.5 kHz by about 4 dB, approximating the acoustic effect of a head on sound reaching the ear. Then a high-pass around 40 Hz, called RLB, so deep bass counts for less. After that the signal is squared, averaged and weighted per channel, with surround channels counted slightly louder.
The filter is deliberately crude; the equal-loudness contours are more complicated and level-dependent. But it is cheap to compute and works surprisingly well, and a 2011 revision added gating so silence doesn't drag the number down. The result is the figure that EBU R 128 and the streaming services normalise against. Worth knowing when a sub-heavy master measures quieter than it feels: K-weighting simply cares less about bass.