The problem of different length of the same word
Submitted by Hassanat on Mon, 06/16/2008 - 14:12.
I am trying to test/train my own speech recognition system, but I stuck with the problem of the different length of the signal that represent the same word, for instance some people say the word six in 500 milliseconds, others need 1000 milliseconds, and so on, I am using KNN and/or Support Vector Machines, but according to my knowledge both need the feature vector to be of the same length.
any ideas?
