I was wondering if anyone has tried using different acoustic models from the default ones that come with OpenEars (which are trained on 8kHz sampled speech)?
If so, could you please describe which models you’ve used and let me know if you noticed any difference in the ASR performance. (this will obviously depend on the task/grammar, but even at the high-level I’m curious in other people experience with different acoustic models).
Two that come to mind are Voxforge (actually they seem to have couple of different acoustic models) and CMU Spinhx 16kHz HUB4 models from http://sourceforge.net/projects/cmusphinx/files/Acoustic%20and%20Language%20Models/US%20English%20HUB4%20Acoustic%20Model/