Correlation provides a measure of similarity between two signals. This video explains process of correlating discrete signals and highlights when normalised correlation is required.
Great vid! Where can I learn about cross correlation? I'm writing a matlab code that can understand basic voice commands by comparing two audio signals. I'm using the resample function to make both signals the same length. However, if one signal is slightly shifted it kinda messes up the whole correlation calculation. What method would you recommend for this situation? THANKS!
Cross correlation would seem to be the way to go alright. Matlab has a built-in function xcorr that would be of help. xcorr returns an array of correlation values - if you take the max of these values this is (most likely) what you would get if shifted the signals to line up and then correlated. I'm going to put a video together on this sometime soon (hopefullly over the next few weeks). Best of luck.
You have to make practice with this definition for fourier transform. Fourier transform correlates sin(wt) and cos(wt) functions with any function f(t)
WHY would we want to know about the similarity between two signals? How does that advantage us in, say, signal processing? I admit I'm only 3 mins into the video and so don't know if this issue is talked about in the video.
+thetruereality There is a subtle difference mathematically and a massive difference conceptually (at least initially). Check out my reply to Mehrdad Ahadi.