Experiments and Results

I performed some experiments to see how my filter works. I grabbed a speech signal from the internet of voice signal. I stored it in one of my files called homer. I used the facilities provided by Matlab to load an audio file to my program. His speech signal is as shown in the following diagram.

I used the default unix sampling frequency of 8192 Hz for Matlab to do my simulations. The time of the speech signal was approximately 6 seconds. I used the adaptive filter's length to be 20 so that I may not fall in any trouble. Well later after doing a lot of work on it I realized it didnot much on the filter length because the co-effecients of the filter would distribute among themselves accordingly. The time period was the inverse of the sampling frequency. The following took me some time but it was worth the pain. For my experiments I added the noise to the known filter and used a form of the noise to be the reference input in order to reduce the noise. For the first part I kept the noise frequency constant and for the second part varied the noise frequency.