Framing in LPC method, speech recognition

[adm89]

Hello,
I wrote a program for recognition a few isolated words. It works, but with little effectiveness. The algorithm based on single FFT spectrum features of whole word and k-Nearest Neighbors classifier.
Now I want to build a program based on Linear Prediction Coefficient. At this moment I split word signal on frames. My problem is that the number of frames depends on signal length. How can I normalize it? What can I do to obtain a matrix with coefficient independent of the signal length?
I tried to use a resample() function on word signal, therefore every analyzed word had the same length, therefore the same number of frames. But I'm not convinced it's correct procedure.


PS: Sorry for my English, but I'm still learning .

 

[Rasmina Karim]

Hello.!
I had a projectt on speech recognition and i was wondering if you could help me up on it. I used a MFCC approach for fwature extraction. As the next step, i 'd like to use HMM for recognition.. I have trouble on the HMM code. Can you can help me out with the code please?