Geek Answers Handbook

How to use mfcc function to train svm classifier for voice recognition?

I am currently participating in the voice recognition discussion phase, I am using the MFCC function extraction, but the MFCC function returned by the function is the matrix, e, g. a (20.38) for each voice file (wav). But how to pass this function to the SVM classifier. For SVM (and another classifier), each pattern is represented by a vector, right? but the MFCC function for each sample is a matrix. Suppose that Xi is an MFCC function for sample i, then the function for pass i for SVM is: 1) vector 20 * 38, e, g. Xi (:) in a matte shape. 2) means (Xi). 3) one of the columns or rows in Xi. which direction? any useful code document for this?

thanks! Shine

+4
source share
1 answer

For the task of tagging a sequence, such as speech recognition, you need to use a combination of SVM and HMM, not just SVM

  • Align feature matrix with states with GMM-HMM, get a function corresponding to each HMM state
  • Configure SVM for features belonging to each state
  • Implement SVM-HMM instead of GMM-HMM

More details ...

http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.27.442

To do this quickly, use existing toolkits, for example:

http://www.cs.cornell.edu/people/tj/svm_light/svm_hmm.html

+5
source

More articles:


All Articles