I have trained a CMUSPHINX custom acoustic model. It was trained without any errors and successfully decoded with WER of about 35%.
Now using the trained model, I wis