Reputation:
I am using cross-correlation to find where an audio signal occurs within a recording. When doing this, the point of highest correlation is always found somewhere within the signal in the recording, but I'm looking for a way to find the point where that signal BEGINS in the recording. Does anybody know of a way to go about doing this, or if cross-correlation will even do the job? Thanks in advance.
Upvotes: 1
Views: 1261
Reputation: 70703
If your signal is stationary, then instead of looking for a maxima using a single cross-correlation window, try looking for a maximum difference between 2 adjacent signal-sized cross correlation windows. If the prior window shows a very low correlation, and the current window shows a very high correlation, then the likelihood that you are right at a transition edge is good.
Upvotes: 2