I've been having this weird thought lately about how to increase the 'definition' of audio using convolutions. The idea is as follows

1. Convolve a rectangular function of a certain width against the audio

2. The local maxima's of the convolved function show where there is the highest average amplitude, so increase the amplitude of those regions a bit

3. Repeat with smaller rectangular functions to capture finer details

Does this already exist, and if it doesn't, does this idea even work