Hello,

I have tried to design a acoustic model for my application and would like to test its performance. So, I need your help. Please care to suggest and send your feedback so that i can improve my model.

https://www.dropbox.com/s/z2mxl088gsogu5z/Deutsch_Acousticmodel_check.7z

The problem i am finding and would like to get suggestions from you!

FAHRERKENNUNG 1 2 3 4 (FAHRERKENNUNG EINS ZWEI DREI VIER), it sometimes gives result with output FAHRERKENNUNG 1 2 3, How can we optimize it?

How to optimize for words like AM and FAM.. !! sometimes it swaps the expected result.

Should i take only commands as audio file and try to train the system?