Frequency masking is modelled as convolution with a spreading function

Evidence

The next processing step emulates how the excitation evoked by a stimulus spreads to other frequency bands. The simplest emulation approach comprises a convolution with a spreading function, as in Equation (10.6), and such an approach is also exploited in the code above. The function approximates the shape of the simultaneous frequency masking curve at 60 dB SPL. In reality, the shape of such a curve is level dependent, which can be accounted for by selecting the spreading function in a level-dependent manner. This is neglected here, since it would make the structure of the auditory model far more complex. The excitation patterns following the convolution can then be scaled into short-term specific loudness spectra following Equation (10.7). However, the presented model omits this scaling and consequently the resulting auditory spectrum describes the level of the excitation pattern in dB in different frequency bands. In other words, the output provides an estimate of the sound spectrum that is available to hearing mechanisms. — Ville Pulkki, Matti Karjalainen, Communication Acoustics- An…, 1588

Categorization

• DEFINE: Clarifies that excitation-pattern spreading due to masking is computationally implemented as a convolution with a spreading function approximating the masking curve.
• XREF: connects to: Excitation spread across frequency bands is modelled as convolution with a level-dependent spreading function. — Ville Pulkki, Matti Karjalainen, Communication Acoustics- An…, 1588