Gaussian Mixture Model for speech recognition

Hi all! I'm implementing a tool for speech recognition (command based).

My training data are 21 commands (7 different commands with 3 utterances for each). I did:

• the pre-processing phase (silence removal and end-point detection)
• the features extraction phase (with MFCC calculation).

So, for every utterance in my training set, i have a MFCC matrix with 12 columns (12=number of MFCC) and as much rows as the number of frames i divided the signal.

For the recognition phase, i was wondering to use the gmdistribution tool.

I read this article:

http://www.mathworks.it/company/newsletters/digest/2010/jan/word-recognition-system-matlab.html but i didn't understand this code line:

    % model = gmdistribution.fit(MFCCtraindata,M);

What is the MFCCtraindata parameter?

Is it the MFCC matrix associated with every utterance?

For each command i have 3 utterances, so i have 3 different MFCC matrixes.

How can i do to create a unique gmm if, for every command, i will got 3 different gmm?

Any kind of help will be appreciated.

Thank you!!

Answer 1

[Castalia]

Nobody could give me any advice, please?

Answer 2

[Rania Ziedan]

i really need help in the same issue if you handled it could you help me thanks in advance

Answer 3

[MUZITIANXINJIE]

Yes,I want,but no one help me! I really need to use the deep learning tu classfy the voice recognition . thanks for your help.

Answer 4

[yasir riaz]

please help

Answer 5

[hanieh rafiee]

Hi Is the answer to your question receipts? Will you help me please?