is speech-to-noise ratio of 15 really acceptable? Options

[Pat]

Gist of my post: I'm wondering if I should train further, given that 16 is the best speech-to-noise ratio I've gotten, or if I should focus instead on getting a better speech-to-noise ratio, perhaps convincing my employer to buy a different microphone or a Soundblaster card (plus get Tech. Support to disable my integrated sound card and install, etc. the Soundblaster). I don't know if 15 (current benchmark) is good enough to get decent accuracy.

Tried making a wav file recording off the microphone: I can see why DNS6 Pro says that there's a fair amount of noise -- I hear it! When played back at top volume, wav has my voice at a very reasonable sound level (my voice is quiet), but there is a fair amount of noise, a bit of crackling, but mostly white noise. It seems somewhat rhythmic -- I believe it is probably coming from the many fans that this computer has. When I play it back at the normal volume I have set for my computer speakers, I have to strain to catch what I'm saying, but the noise is much less obvious. I played around with microphone boom position, and whether or not I was facing the monitor, close to walls, etc. Nothing made the noise diminish. (John Wickett, thanks so much for posting the information that let me know to try these tests! -- and for your help in March). I even tried putting the microphone near the fan outlets. That increased the noise a bit.

Does it sound like my microphone is bad? Or is the computer too noisy (fan-wise or electric wise)? Could a standalone sound card solve this problem? Or should I just proceed and see what kind of accuracy I can get?

More details:
I'm getting initial speech to noise ratios of 12 to 16, using DNS6 Pro with the Emkay microphone it came with, on a desktop (an IBM NetVista M41 (#6793) with a Pentium IV 1.8G processor, just over 1 G of RAM). At ScanSoft's suggestion, just installed the latest sound card drivers (but I think I already had them). Sound card is integrated. I do notice now that the computer itself is somewhat noisy -- humming and the sound of various fans.

I gather that acceptable speech to noise ratios are 15 - 30, usually. I'm wondering if I should even bother trying to train further, given that 16 is the best I've gotten, or if I should try to get a better speech to noise ratio immediately, trying perhaps a different microphone or a Soundblaster card. I can manage to get an acceptable speech-to-noise ratio, but if I try again or start a new user, I usually don't. Range is 12 to 16 so far.

I've read through relevant posts since I last visited in March. I have played around with microphone position, computer position, etc. I'm even wondering if bending the microphone boom back on itself was a bad idea (my ear and the corner of my mouth are closer to one another than is apparently standard!). I tried straightening it -- didn't improve the wav file noise, but maybe I already busted it.

I have a low, soft voice for a woman; I speak slowly, too. I need voice recognition because of RSI injuries to both arms that haven't healed.

[This message has been edited by Pat (edited 06-14-2002).]

[Pat]

How different in quality are the microphones that tend to be recommended here (Audigy 700 ANC, e.g.) from the microphone shipping with DNS 6 Pro (Emkay)?? Emkay is noise-cancelling, but I don't know if it's ANC.

[This message has been edited by Pat (edited 06-14-2002).]

[Pat]

Did wav testing -- edited first post accordingly.

[Judy Evans]

Pat

the signal to noise ratio is normally less important than the recognition rate. But it looks to me as though your sound system is noisy and you should get rid of the integrated card and have a SoundBlaster Live fitted. (The noise is likely to be the card, not the fans; noisy fans have never affected me.) John though may suggest trying a hotter mike before changing the sound card or opting for a USB mike; see my comments on mikes below.

>>>>>>>>
I have a low, soft voice for a woman; I speak slowly, too. I need voice recognition because of RSI injuries to both arms that haven't healed
>>>>>

Same here -- my hands though are better than they were -- I suggest you get a new mike as well as a new soundcard. I also suggest you try the Vxi TalkPro (John, is that the right name?) as it will be "hotter" than the ANC 700 but is (I hear) as noise-cancelling. I use a hot mike that is less noise- cancelling than either, but I dictate in a quiet room.

>>>>>>>>>>>>
How different in quality are the microphones that tend to be recommended here (Audigy 700 ANC, e.g.) from the microphone shipping with DNS 6 Pro (Emkay)?? Emkay is noise-cancelling, but I don't know if it's ANC.
>>>>>>

I don't know that particular Emkay. I doubt it's noise-cancelling. Only Andrea mikes (Audigy is a sound card) are ANC; but I have as you'll have seen recommended a different mike for you given your voice, which as I said will be very like mine

Judy

[John Wickett]

Pat,

The mics that are shipped with DNS are really very cheap and do not give good recognition except under the most optimal conditions.

As Judy says, with what you are hearing from the .wav file you definitely need either a standalone SoundBlaster with an amplified mic(VXI TalkPro is good, but the Andrea ANC 700 can also be amplified and is also very good) or a USB mic. I typically use either the TalkPro or the ANC 700 and can't decide which is best at recognition. I haven't found a USB that beats either of them. I am, however, still attached to a 5-6 year old Andrea that came with DragonDictate - the most comfortable mic I have ever used.

The "white noise" is very typical of an integrated soundcard.

John

[Pat]

Dear John & Judy:

Thank you!

I seem to have told you everything except that I'm running NT 4, sp 6. So a USB mike is out for now, unless I ask permission to get Windows 2000 early. The latest is that the corporation may switch to Windows 2000 this fall or so. They tried a Windows XP pilot and had too many problems with it (I didn't participate in the pilot because I use many older versions of software programs).

What does "hot" mean about a microphone?

Unfortunately, I'm having more problems with DNS6 Pro -- C++ runtime error about 5 - 10 minutes into general training, causing abnormal program termination, which I'll post under a different thread.

[This message has been edited by Pat (edited 06-17-2002).]

[John Wickett]

Pat,

A "hot" microphone is one where the signal is slightly amplified before it reaches the sound card. This is normally accomplished with two AA batteries but some, generally those that are used for both dictation and telephone, plug-in to normal electrical outlets. This allows the audio input signal to overcome a certain amount of electrical interference or noise.

John
Dictated with Dragon Naturally Speaking
John Wickett
PEI Canada
902-368-0141
902-651-2152
jwickett@auracom.com
jcwicket@vac-acc.gc.ca

[Pat]

Wow, thanks! Since the topic has evolved to microphones, I'll post my next question under a new thread.

Just sent my boss an e-mail breaking the news about the microphone, soundcard, and possible need to fdisk (with mucho help from tech. support).

[This message has been edited by Pat (edited 06-18-2002).]