Dragon Naturally Speaking - Which Headset Do You Use?
- Thread starter stevesdata
- Start date
I've been managing DNS deploys across a wide corporate domain, and happened to find this forum as I was searching for headsets.
From the experience of a LOT of our users (about 3000), here is what I've learned:
1> The quality of digital data presented to DNS is critical for extremely high recognition rates (and most folks are dissatisfied at less than 99.999% - no kidding!).
2> Avoid "plug" microphones; these require your computer to do the analog to digital translation. The better ($80+) USB headsets have a small powered module which does this work. Thus the requirement that you "tell" DNS which one you are using, so it know if it should do the A-D conversion.
3> Low quality headsets must be "trained" in each environment you will work. (For me, that's 3 user files). The highest end ($300) do not.
4> Just about any headset that passes the internal DNS test will work in a quiet environment. DNS degrades quickly (and rather catastrophically) in any environment where noise or background speech is present.
5> For noise, use any headset with "noise reduction" (for wind, have a wind sock on the mike). A very good model for many of our folks is "The Boom", around $80. I use Logitech as noted above, with great results.
6> For voice noise and changing loudness of noise (planes flying overhead is a typical challenge for us), only the high-end $200-300 USB microphones give acceptable results. Because of how they maximize the signal-to-noise ratio, you can work with one user-file in many environments, simplifying use. (Big benefit dealing with non-tech types.)
7> Yes, diction counts, but less than you might expect. Dick Tracy's "Mumbles" would have terrible problems no matter what he used. Speech speed is related more to noise (rapid speech sounds like noise to the computer) than conversion speed. Thus, rapid speakers will find a marginal benefit to USB-headsets (and probably should have a lot of abbreviated-text-to-paragraph items set up).
So, if you are speaking clearly, slowly, and not getting great results, first try going to an $80 USB headset (save your receipt), and then to the high-end $200-300 headsets. Make sure the $80 headset has the module that converts to digital (virtually all do if USB). In any case, only use the cheap plug-ins if you are an occasional user in a quiet environment, and close background software (besides the target of the transcription) on slower desktops. (Task Manager's CPU load graphic can help determine this.)
Hope this helps...
Oh... this was done in DNS with no edits.
From the experience of a LOT of our users (about 3000), here is what I've learned:
1> The quality of digital data presented to DNS is critical for extremely high recognition rates (and most folks are dissatisfied at less than 99.999% - no kidding!).
2> Avoid "plug" microphones; these require your computer to do the analog to digital translation. The better ($80+) USB headsets have a small powered module which does this work. Thus the requirement that you "tell" DNS which one you are using, so it know if it should do the A-D conversion.
3> Low quality headsets must be "trained" in each environment you will work. (For me, that's 3 user files). The highest end ($300) do not.
4> Just about any headset that passes the internal DNS test will work in a quiet environment. DNS degrades quickly (and rather catastrophically) in any environment where noise or background speech is present.
5> For noise, use any headset with "noise reduction" (for wind, have a wind sock on the mike). A very good model for many of our folks is "The Boom", around $80. I use Logitech as noted above, with great results.
6> For voice noise and changing loudness of noise (planes flying overhead is a typical challenge for us), only the high-end $200-300 USB microphones give acceptable results. Because of how they maximize the signal-to-noise ratio, you can work with one user-file in many environments, simplifying use. (Big benefit dealing with non-tech types.)
7> Yes, diction counts, but less than you might expect. Dick Tracy's "Mumbles" would have terrible problems no matter what he used. Speech speed is related more to noise (rapid speech sounds like noise to the computer) than conversion speed. Thus, rapid speakers will find a marginal benefit to USB-headsets (and probably should have a lot of abbreviated-text-to-paragraph items set up).
So, if you are speaking clearly, slowly, and not getting great results, first try going to an $80 USB headset (save your receipt), and then to the high-end $200-300 headsets. Make sure the $80 headset has the module that converts to digital (virtually all do if USB). In any case, only use the cheap plug-ins if you are an occasional user in a quiet environment, and close background software (besides the target of the transcription) on slower desktops. (Task Manager's CPU load graphic can help determine this.)
Hope this helps...
Oh... this was done in DNS with no edits.
