Conclusion:

Our Wiener deconvolution process demonstrates that echo removal is indeed possible. Unfortunately the procedure generates a relatively noisy output signal and would only be useful in the presence of extreme environmental distortion (e.g. room echoes cause the signal to be unintelligible). With a few modifications however, deconvolution could remove echoes reliably as well as preserve the quality of our signal

Improvements:

Wiener deconvolution relies on previous statistical knowledge of both the additive noise and the original signal being restored. The statistical characteristics of the room could not be approximated; so normal deconvolution was used, resulting in a significant increase in noise. Possible alternatives to using deconvolution to find the room’s respone and address this issue are:

Also, advanced noise filtering techniques would improve signal quality at every step in the process. Since deconvolution tends to amplify noise, filtering would profoundly improve the quality of output signals in this multi-step procedure. For example, filtering to remove noise of the test recording before deconvolving it with the room response could help reduce the noise in the final recovered signal.