The masking threshold is the sound pressure level (SPL) of a sound you need to make hearing another in presence of a masker signal. This threshold depends on the frequency and the kind of the masker and maskee. This effect normally appears between two sounds close in frequency. Not hearing implies some advantages when you speak in transmission terms. In audio codifying, p.ex, implies the possibility to exclude this tone and get a better compression. In other words, codifying with less bits and reduce the size of the final file.
It is not common to work with only one tone, normally you work with some of them simultaneously. When this happens we’ll have a lot of possible maskers at the same frequency. In this situation it’s necessary to compute the global masking threshold. It uses a high ressolution FFT via 512 or 1024 points to know the different tones there are in the sound. Due to the fact that there are bands that humans are not able to hear it is necessary to know the signal level, masker type and the frequency band before computing the individual thresholds. To avoid having the masking threshold under the threshold in quite we add the last one to the compute of partial thresholds. Finally you can compute the SMR (Signal to Mask Ratio). The last operation is typical in the audio codifying.
The next image shows the spectrum of a 1kHz tone. Any sound will be unheard if is under the threshold in quite. This limit changes around the masker frequency, 1kHz in this case, doing more difficult to hear a tone nearby. The slope of the masking threshold is steeper toward lower frequencies than higher frequencies. It means is easier to mask with high frequency tones.
The Psychoacoustic Model
There’s an application of the masking threshold. We find it in the audio encoding process in MPEG. In this scheme there is a block called ‘Psychoacoustic model’. This is communicated with the band filter and the quantify block. The psychoacoustic model has to analyze the samples the filter band sends it. This computes the masking threshold in each frequency band. Doing this process needs a FFT to know the differents bands present in the signal. Depending on the MPEG Layer, we can use more or fewer points. Using these thresholds we’ll know the SMR. It is sent to the quantifier. The quantifier has to assign more or less bits in each block knowing the SMR. The block which has the maximum SMR will codify with the maximum number of bits and the block which has the minimum with the minimum number of bits. If it is necessary we could skip and do not assign bits. Using this procedure we need less bits and in consequence we reduce the length of the file reaching a better compression.