When a person has sensioneural impairment, they lose sensory dynamic range. The amount of loss varies with frequency, generally increasing degrees of impairment toward higher frequencies. Their loudness perception follows what is called a recruitment hearing curve.
The following graph is a schematic diagram comparing, at any one particular frequency, the way recruitment hearing works, compared to unimpaired hearing.
The figure shows increasing volume to the right, and increasing perception of volume in the upward direction.
For people without impairment, the straight line depicts the fact that they hear what is presented. But for persons with recruitment hearing, they don’t hear anything until the sound exceeds their elevated threshold, and only then as a barely perceptible sound. But as the input continues to rise above threshold levels, the perception is of a rapidly increasing loudness which will eventually slow as it reaches the loudest levels.
The diagram also illustrates the fact that at loud levels, impaired hearing is about the same as unimpaired hearing.
Now imagine a stack of these kinds of images, but with each additional image moving that threshold level to a different place. Each image has a similar kind of curve that approaches normalcy at the loudest levels, but where they intercept the threshold level moves toward the left or right, compared to this image.
That imaginary stack of images, arranged from lowest threshold level to highest threshold level, represents your hearing with respect to audible pitch, or frequency. The lowest frequencies in the bass region suffer the least impairment, and the highest frequencies in the treble region suffer the greatest impairment.
To overcome recruitment hearing, we need to take the entire dynamic range from whisper soft to blaring loud and compress it into the more restricted perceptual dynamic range of the listener. We do that with a nonlinear compression that acts to cancel the recruitment hearing curve.
This diagram shows two different recruitment curves in red, against the normal hearing response as the diagonal green line. Those could be two different pitch curves for one individual, or comparing two different individuals at any one pitch.
The light pink curves above the green line show the shape of recruitment canceling compression that must be applied to their mirror image recruitment curve, such that the result after compression would perceptually mimic the green curve. A person with recruitment hearing, listening through such compression, would hear the same thing that any person without impairment would hear.
Compression is just a varying amount of EQ boost which depends on the loudness of the input sound. Those blue lines illustrate how much boost is needed for the more severe recruitment curve to be able to hear properly when the input sound has a level of 45 dB, where the vertical blue line has been placed. The amount of gain needed is indicated by the length of the horizontal blue line.
You can see that the amount of boost that would be needed declines as the vertical blue line moves to the right, toward louder presentations. When things are loud, you don’t need much help. But you need increasing amounts of help to hear things that are progressively fainter.
Crescendo splits the presented sounds into 100 narrow frequency channels, then applies the required anti-recruitment compression to each channel – those pink compressor lines in the diagram.
Conventional approaches use straight line compression curves. And you can easily see how that would not work very well. What slope of compression should you choose? No single slope, or compression ratio, will be satisfactory over the entire dynamic range. A linear compressor might work acceptably well over one portion of dynamic range, but utterly fail everywhere else.
Crescendo uses nonlinear compression, exactly as shown by those pink compression curves, over the entire dynamic range of sounds.