In using the Speech Recognition service, we have found that if a user uses a device to enroll a Microsoft speech profile and then uses a different device to try to verify on that same profile, that the scores are much lower than when verifying on the same device. Is this a known limitation of the service? For example, in recent testing we found scores from enrolling on a laptop and verifying with that same laptop to be scoring in the 80s. While trying to verify with a pair of headphones on the same profile with the same user we see scores in the 50s and as low as 42. We understand that 50 is a passing score in the service's eyes but from testing with other users trying to "spoof" the system we found scores as high as 57 so in our eyes anything below 60 isn't an acceptable score.
