There is a direct inverse relationship between the size and slope of the signal and it’s signal spectrum. The purpose of signal quality measurements is to maximize Signal to Noise ratio chosen to match the signal spectrum with the least distortion.
This methodology improves recognition and reduces error rate (Shannon’s Law) or as much as possible outside that spectrum. If the signal response has sharp leading edge, this represent to upper frequency limit and a slow tail that corresponds to the lower frequency limit.
The approximate correlation is the time Tr, it would take to rise from 10 to 90% of peak such that the half power edge of the bandwidth f(BW) [Hz]= 0.35/Tr [s] in seconds/Hz or milliseconds per kilohertz. This should match the impulse response rise and fall times in the article to 5000 Hz and 700Hz ( or not ). Then to ignore the pulse polarity , a full wave rectifier is used and amplify it above the noise level of your measurement device with gain of 400 seems a reasonable approach. More stages of gain may be required if one wanted to reach 5V to optimize digital resolution. The 150 Hz low pass filter after full wave rectification further reduces the noise yet also some of the signal of less interest (fast rising edge)