Human ear has the masking property that certain audible sound becomes
inaudible in the presence of another sound. Masking is quantified by
the raised threshold from the absolute hearing threshold in quiet. It
is of scientific and practical importance to compute masking thresholds.
Empirical models on masking have applications in low bit rate digital
music compression. A first principle based two level model is developed
with partial differential equation (PDE) at the peripheral level and a
similarity transform to represent all higher levels of auditory perception.
Modeled masking thresholds of banded noise by tonal signals agree well
with hearing data.