The original signal played in the test room is shown below:
 |
When this signal is played in the staircase of Sid Rich College the following signal was recorded back:
 |
Using normal deconvolution of this signal with the room’s response the following singal is recovered:
 |
In contrast, when the recorded signal is Wiener deconvolved with the room’s response the following signal is recovered:
 |
By comparing the audio files of the original signal, the recording of the signal in the staircase, and the two deconvolved signals it’s clear that both methods remove the echo at the price of amplifying the background white noise in the signal. It’s also clear by both listening to the audio and comparing Figures 3 and 4 that Wiener deconvolution removes the echo with significantly less noise amplification. Comparing the two deconvolution results with the original test signal by means square error showed that Wiener deconvolution had about a 25% of the error of normal deconvolution.
