speaker-independent system for notes Options

[speaker-independ...]

Hi all,

I am interested in applying DNS to a system for speaker-independent VR for rough notes, similar to a project of a group at UC Berkeley; see their project description at this website:
http://www-inst.eecs.berkeley.edu/~assist/hearing.html

I don't need near-perfect accuracy but just tolerable accuracy to have rough notes (with some errors) of what has been said-- these notes never become a document but serve as a transient reference during consecutive interpretation (verbal translation) work.

Does anyone have any ideas on how to get this working? Specifically, does anyone know of a unified pitch device or software that can unify pitches before the sound is entered into Dragon?

Thanks!!!!!

[John Wickett]

Just do a search on Google for "voice processors" and you'll see a number of devices that are originally designed for music but certainly work with voice. Either that, or talk to a sound engineer at a radio station or TV station and they can easily demonstrate how you can, for example, make everyone sound like Donald Duck.

John
Dictated with Dragon Naturally Speaking Version 8 Professional
John Wickett
PEI Canada
902-368-0141
902-651-2152
jwickett@auracom.com
john.wickett@vac-acc.gc.ca

[speaker-independ...]

Thanks, John. As I am interested in a minimalist setting that doesn't look so big during, say, a conference (when I do interpretation), I don't wish to use a big machine with all those fancy features that the music industry uses. Would you or anyone know about software or a very small device that can unify pitches? The Berkeley project report says it might be possible to use a DVP (digital voice processing) "chip" by itself, but I haven't a clue as to how to get that chip or how to set it up. I've written to the professor but haven't got a response yet.

Originally posted by John Wickett:
Just do a search on Google for \"voice processors\" and you'll see a number of devices that are originally designed for music but certainly work with voice.  Either that, or talk to a sound engineer at a radio station or TV station and they can easily demonstrate how you can, for example, make everyone sound like Donald Duck.

John
Dictated with Dragon Naturally Speaking Version 8 Professional
John Wickett
PEI Canada
902-368-0141
902-651-2152
jwickett@auracom.com
john.wickett@vac-acc.gc.ca

[John Wickett]

Speaker,

In my experience the DVP chip does not work as it does not apply the same voice modelling to each user. I think that your choice here is to use a voice processor (they aren't that big) or spend over $10,000.00 to get custom made soundcard with the inherently possible software conflicts.

John

[speaker-independ...]

John,

Thanks. Sorry, I meant "DSP" chip, not "DVP" chip. What do you think about a DSP chip?

In my experience the DVP chip does not work as it does not apply the same voice modelling to each user. John[/B][/QUOTE]

[speaker-independ...]

John, I did a search on Google. What do you think about the ZSP voice processor? from the picture it looks like a small square that I guess can be installed inside a computer, but I'm not sure if it can unify pitches. The big blue machines with all the dials are too big and visually intrusive for what I need to do.

Is there any kind of software or algorithm I can install on Windows XP to automatically change all incoming voice into a voice with the same pitch? Even if there is still variation in other properties of the voices, it's ok, as I am not looking for great accuracy and would be willing to experiment with only pitch at this time? And then route that unified-pitch voice to Dragon? Any ideas about such software or who could customize it?

Thanks again for all your patient indulgence.

Originally posted by John Wickett:
Speaker,

In my experience the DVP chip does not work as it does not apply the same voice modelling to each user. I think that your choice here is to use a voice processor (they aren't that big) or spend over $10,000.00 to get custom made soundcard with the inherently possible software conflicts.

John

[John Wickett]

>>What do you think about the ZSP voice processor

I haven't tried it, but after reading about it I see it is designed for VoIP and telephony - totally different applications than voice recognition.

I tend to go with the Symetrix voice processors and I don't find them to be very intrusive.

John
Dictated with Dragon Naturally Speaking Version 8 Professional
John Wickett
PEI Canada
902-368-0141
902-651-2152
jwickett@auracom.com
john.wickett@vac-acc.gc.ca

[speaker-independ...]

John,

That's great to hear that you have experience with this! What do you use the Symetrix voice processors for? For a system approximating speaker-independent VR? If so, I'd love to hear more of the details of the system you have, to learn from your experience.

I am really tech illiterate; what do you think about the DSP (digital signal processor) chip? is it basically the same thing as a DVP chip?

Thanks a million again!

Originally posted by John Wickett:
>>What do you think about the ZSP voice processor

I haven't tried it, but after reading about it I see it is designed for VoIP and telephony - totally different applications than voice recognition.

I tend to go with the Symetrix voice processors and I don't find them to be very intrusive.

John
Dictated with Dragon Naturally Speaking Version 8 Professional
John Wickett
PEI Canada
902-368-0141
902-651-2152
jwickett@auracom.com
john.wickett@vac-acc.gc.ca  

[speaker-independ...]

I just went to the Symetrix website; about how big (length and width) are the 528E and the 628? They look kind-of big; I think it might look weird if I put something like that on a conference table.

I discovered recently that my Toshiba Portege 3505 PC has a great built-in internal microphone or mic array inside the monitor; when I use the "microphone boost" feature, it can can pick up a voice clearly and crisply from a couple yards away, even if the voice is speaking to the back side of the monitor. And this is ideal, so I wouldn't have to have a visually unpleasing external microphone stuck to the top of my monitor during a conference and I wouldn't need to bother someone to wear a lapel mic. So now what I am directing all my attention to is finding some kind of software that I can install directly on my operating system, or some kind of very small hardware that I can install inside my computer. Any ideas here? Thanks again!

[John Wickett]

speaker,

>>what do you think about the DSP (digital signal processor) chip? is it basically the same thing as a DVP chip?

DSP is basically for telephony while DVP is intended for video.

>>What do you use the Symetrix voice processors for?

Many of my clients are disabled and have dysarthric or aphasic speech, so it is necessary to alter the audio input to the soundcard so their voices are recognizable by DNS.

The Symetrix 528e is about two thirds the size of a keyboard and is one of my favorites. However, I am not using this for multiple voice input and, as far as I'm aware, there is no product on the market that will deliver what you are looking for. At this point in time you will basically have to settle for a piece of hardware that will shrink all voices so that they are as similar as possible and then deal with the high error rate. The recognition of multiple voices is not supported at this time, but my three sons can use my voice file without any problems. If my daughter or grandchildren try to use my voice file the results can be almost incomprehensible.

We are well aware that a lot of people are looking for this technology and hopefully it will be available in the near future. In the meantime, all you can do is try to alter voices prior to input to the computer.

John
Dictated with Dragon Naturally Speaking Version 8 Professional
John Wickett
PEI Canada
902-368-0141
902-651-2152
jwickett@auracom.com
john.wickett@vac-acc.gc.ca

[speaker-independ...]

John,

>>> but my three sons can use my voice file without any problems. If my daughter or grandchildren try to use my voice file the results can be almost incomprehensible.


This is good to hear! This confirms the results of a similar experiment with men and women's voice files described on the ASSIST webpage at UC Berkeley. Just to confirm-- are you referring to your voice file after it has been processed by the Symetrix? Or without the Symetrix?

I am thinking of making different sample voice files, definitely at least one for adult men and another for adult women; I won't be working with children.

Thanks again, I am learning a lot.

[speaker-independ...]

>>> Just to confirm-- are you referring to your voice file after it has been processed by the Symetrix? Or without the Symetrix?


Sorry, I meant to say "...after the SOUND of your voice has been processed..."

[speaker-independ...]

John,

>>>> a piece of hardware that will shrink all voices so that they are as similar as possible


Is there way to set up a piece of hardware, for example the 528e, so that it automatically skrinks all incoming voices to the SAME PREDETERMINED level of "shrinkness" ? So that I wouldn't have to adjust the "shrinkness" every time a different speaker speaks? By the way, what does it mean to shrink a voice?

And can the 528e automatically change the pitch of a voice to a predetermined pitch value?

[speaker-independ...]

John,

Do you think the "Voice Changer Software" (see webpage link below), which can change voices "in real time", could work in conjunction with Dragon? Would be there be a way to take the digital output of such a program and feed it in as the input sound of Dragon? The webpage for the software is:
http://www.softwarevault.com/Download/Voic...ftware-Download

Thanks again for your experienced views.

[John Wickett]

speaker,

>>are you referring to your voice file after it has been processed by the Symetrix?

When my sons are using my voice file (I don't save it after they use it) there is no need for the Symetrix. Of course, having been raised by me, their vocabulary, phraseology and pronunciation are very similar to mine.

>>so that it automatically skrinks all incoming voices to the SAME PREDETERMINED level of "shrinkness" ? So that I wouldn't have to adjust the "shrinkness" every time a different speaker speaks? By the way, what does it mean to shrink a voice?

Once the Symetrix is adjusted, the output to the soundcard is very similar for all voices. "Shrinking" refers to modulating all audio input so that the output to the soundcard is almost a monotone; no highs or lows as you adjust once for the decibel and frequency range that is desired.

>>Do you think the "Voice Changer Software" (see webpage link below), which can change voices "in real time", could work in conjunction with Dragon?

I have never heard about it, but it sounds like it alters the output rather than the input.

I have to tell you that this is an imperfect solution to what you want, but it is probably the best you are going to get right now. You have said that you are willing to accept a lot of errors and that is what you are going to get - the hardware will just minimize them as much as possible.

John

[speaker-independ...]

John,

>>> I have never heard about it, but it sounds like it alters the output rather than the input.


That is the concept I'm trying to get my head around; how can I make the voice output of a software package to be the voice input for DNS? Can ScanSoft customize DNS for me to do that? Or can I have something inside my computer rewired?

[John Wickett]

speaker,

>>how can I make the voice output of a software package to be the voice input for DNS?

I am not sure what you mean by "software package", but in my experience you have to alter the voices before they get to your soundcard and that is why I am suggesting an external hardware device.

John
Dictated with Dragon Naturally Speaking Version 8 Professional
John Wickett
PEI Canada
902-368-0141
902-651-2152
jwickett@auracom.com
john.wickett@vac-acc.gc.ca

[speaker-independ...]

>>>I am not sure what you mean by "software package", but in my experience you have to alter the voices before they get to your soundcard and that is why I am suggesting an external hardware device.


Hi John,

I experimented with a couple voice-changing software products, and good news! the AV Voice Changer 3.1 Gold software can change the microphone voice input in real time BEFORE the sound goes into DNS. On playback DNS remembers the voice that is outputted by the Voice Changer software. I guess the better Diamond version of their software can also do it, but I haven't tried yet.