I had the same problem, microsoft speech SDK need to be installed and the the wav file you are working on MUST be encoded to 44khz 16 bit stereo (use sound recorder and just do a " save as") when it wasn't my version of face poser just crashed.
As far as I know this is a common error. As long as your files are
PCM format
Stereo
then that is correct. The error seems to be just a coding fault. the results you get with most extractions are pretty poor anyway so you're best off building them yourself.