Acoustic model accuracy

[shamsudheen]

Hi All,

Let me clarify one thing,

I am creating acoustic model using the link http://cmusphinx.sourceforge.net/
wiki/tutorialadapt

first i have created a acoustic model with 20 wav files. it given 40-50%
accuracy(used first 20 words from CMU ARCTIC text-to-speech databases).
then i have created a acoustic model with 200 .wav files. it given 55 -60 %
accuracy
then i have created a acoustic model with 400.wav files. it given 70-75 %
accuracy.

will acoustic model accuracy is related to the number of files(.wav) we
trained?

if yes, what is the maximum .wav files count for acoustic model training?

Let me know...

Thanks..

[Nickolay V. Shmyrev]

if yes, what is the maximum .wav files count for acoustic model training?

There is no maximum. Good acoustic models use about 10 millions.