Our group brainstormed through many different project ideas.
After much deliberation, one of our group members told us of one
of his performances with the Rice Choral Group. It was a
lovely performance and it was well received. However, during
the recording of the live performance, someone began to cough.
This was a tragic story, and we decided that being able to take
out unwanted sounds from audio files would be a good idea for
a project.
We began to hash out ideas, and were first led to time domain
analysis. Using pattern matching, we thought that we could find
where the unwanted noise was in time, and remove it. However,
how to remove the signal delicately enough to preserve the music
was an item that required more analysis. Specifically, we then
discussed Fourier analysis. We then thought we could analyze harmonics
of the music, and look for anomalies throughout the spectrum.
This, we believed, would provide limited functionality because,
we would only be taking out noise maybe once in a sound file,
and the rest of the time we would be removing musical instruments
such as drums or cymbals.
Realizing that our noise would be a fairly unique event in the
music, we began to discuss more sophisticated methods of analysis.
Seeing how our problem was localized in time, we turned to short-time
Fourier analysis. Research was then conducted into means of accomplishing
short-time analysis.