Speech like all other forms of signals gets corrupted by noise. It is desired that these signals are as noise-free as possible to obtain the ‘true’ signals. We looked at two methods for signal denoising: Wiener filtering and Spectral Subtraction.

To do this, a variety of signals corrupted by different levels of noises were denoised using the two approaches and the results were compared to see if one method fared better than the other.

I:   Spectral Subtraction

II:  Wiener Filtering

III: Signal Processing

IV: Sample Plots

V:  Results

VI: Conclusions

Addendum (MATLAB code)