I am a newbie in speech processing and experimenting to get a feel. I have extracted some speech segments using a window function and I want to find distance between a pair of segments. First, I took DFT of a segment and downsampled from the resulting frequencies (like 16 out of 256) and used them as features.

Then, I saw cepstrum (cceps in matlab) and linear predictive coding (lpc in matlab) coefficients can be extracted from a speech segment. My question is; what is a good way to use these coefficients to find distance. Should I use the first k coefficients or downsample like in DFT? Any hints about their usage would be appreciated.

  • $\begingroup$ What do you mean by "distance"? $\endgroup$ – Jim Clay Mar 18 '13 at 3:25
  • $\begingroup$ I have so far experimented with Euclidean norm. My first aim is to discriminate wovels looking at these distances. $\endgroup$ – Zoran Mar 18 '13 at 9:53

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.