Próximos experimentos

!!!Treinar com Male e Testar com Children [COM VTLN] 
[VTLN Using Analytically Determined Linear-Transformation on Conventional MFCC - D. R. Sanand and S. Umesh - IEEE 2012]
http://ieeexplore.ieee.org

Treinar com 8/11/16KHz e Testar com 8/11/16 KHz
[Speech Recognition at Multiple Sampling Rates - Hirsch,Dobler -Ericsson Eurolab Deutschland - Eurospeech 2001]
http://dnt.kr.hsnr.de/papers/eurospeech01.pdf

Treinar com Male/Woman/Children e Testar com Male/Woman/Children
[Acoustic-Phonetic Speech Parameter for Speaker-Independent Speech Recognition -Deshmukh,Wilson,Juneja - University of Maryland, Department of Electrical ]
http://terpconnect.umd.edu/~juneja/icassp_digits.pdf

Treinar com 1/2/4/8/16/32 mix gaussian
[Approximate Test Risk Bound Minimization Through Soft Margin Estimation - Li,Yuan,Lee IEEE, 2007]
http://www2.isye.gatech.edu/~myuan/papers/sme.final.pdf

Treinar com diferentes conjuntos de locutores Male/Woman/Children [COM VTLN]
[Model space size scaling for speaker adaptation - Mats Blomberg - Speech, Music and Hearing, KTH/CSC, Stockholm- 2011 ]
http://www.speech.kth.se/prod/publications/files/3597.pdf