PDA

View Full Version : Dictation + Voice Recognition on NX vs Tungsten C


Nickster
04-24-2003, 09:59 AM
Anyone with experience using their NX for dictation and/or voice recognition transcription with Dragon Naturally Speaking?

I am trying to decide if this vs the Tungsten C is a suitable device for professional dictation and voice transcription purposes?

I know it won't be as good or accurate as my Plantronics digital signal processing USB microphone I use with my computer, but the convience of handheld dictation for later transcription via Dragon Naturally Speaking voice recognition is appealing.

Sony's personal recorders do work with Dragon Naturally Speaking, but how much nicer would it be if a full-functioned PDA could eliminate the extra device...

The technical requirements for the files is that they be 16 bit mono wave files at 11.025 or 22.050 kHz. Can the NXs do this.

The new Tungsten C looks like stiff competition for the Clie NX, but apparently it does not offer a DSP chip, which, I guess, is needed for dictation (true?).

Thanks,
Newbie Nick

CDR
04-24-2003, 11:31 AM
Nickster,

I actually just tried this yesterday, and it did not work. I checked the file I recorded using the NX, and it says that it is 4 bit mono at 22 kHz.

Unfortunately, it doesn't look like it will do it. I was hopeful, though.

Michael Quach
04-24-2003, 11:55 AM
How are you getting the voice file read? Do you download it onto the computer first? If so, I didn't even know you could do this....

I've had lots of success with the Dragon. I record the voice files, and play them into the microphone input, using a standard male/male wire from Radio Shack. Works beautifully! The sound comes through clear as a whistle. You may need to train a new voice file, using the NX70 for data, though.

The extra little perk, is that if you record on the NX70 while the wire is still connected, it plays the sound into the computer. Voila! instant digital microphone!

Lemme know if it still gives you problems.

hansschmucker
04-24-2003, 11:55 AM
You can use tools like like RADutil ( www.smacker.com ) or BeSweet to automatically turn the 4-bit ADPCM audio into 16-bit PCM files. A simple batch file should do that. Of course, the quality will not increase, but 4-bit is fine for voice (You can even understand 2-bit audio pretty well)

dmale7
04-24-2003, 12:17 PM
What would it take to have the NX recognize voice commands internally...is it even possible. I'm a very low level geek, so please be patient with me.

CDR
04-24-2003, 12:25 PM
Michael,

I am using NS Preferred 7. I just came out. Among the tools in this version is one of transcribing audio files, including .wav files. You just have to point it towards the audio file and, i assume, it takes over from there. But that is just a guess, since yesterday is the first time i have tried.

As for the digitial microphone, I already have a Philips SpeechMike Pro (USB) that works great.

ruubster
04-24-2003, 12:25 PM
You can convert the wav files from the NX to the format required by Dragon just by using the open/save as function in the standard Windows Sound Recorder.

Unfortunately I was hoping to use this to transcribe my Professors' lectures from class - but, as far as I can tell, voice recognition technology is still way behind the curve when it comes to untrained voices.

Does anyone know of software that can do this well?

ruby

Winca
04-24-2003, 01:08 PM
Originally posted by ruubster


Does anyone know of software that can do this well?

ruby

Well, you could pay a little kid five dollars to transcribe it for you...

I know as a kid, five bucks was a lot; don't know about today...

sralmas
04-24-2003, 01:21 PM
If anyone has had any success transcribing voice recording, please post your story here.

finger123
04-24-2003, 05:36 PM
A long time ago before the infamous cliesource crash I posted that I had found a way to dictate to my NX60. It is sort of old- school but it works. I recorded an audio file with the recorder on the Clie then plug the earphone output into the mic jack of my computer and turned on my Via Voice then pressed play. It works well if you adjust the volume a little.

alexasha
07-01-2003, 06:25 PM
For some reason, wav file from MS does not recognized by neither windows program. I have recorded on LP. Any suggestions. I am tested Dragon NS 6. to see if I can record voice memo on NX70, place MS in memory stick reader and transcribe this memo on computer.

ruubster
07-01-2003, 06:39 PM
Alexasha, you should be able to directly access the WAV file with the Windows Sound Recorder. To make the file readable by Dragon you'll need to first convert it to 11KHz, 16bit Mono (I think that's the right configuration - I don't have Dragon installed anymore 'cause it doesn't transcribe untrained voices - i.e. lectures or meeting discussions - well). This whole conversion process is described in the Dragon software Help.

In any case - it's relatively easy to convert the file - just open the WAV file with the Windows Sound Recorder, then do a Save As, select the "Change" button for the format, keep it PCM format, and from the attributes drop down window select the right frequency and bit width format. I'd suggest looking for this in the Dragon help menu as I'm writing this from memory...

hope this helps.

ruby

alexasha
07-01-2003, 08:21 PM
Thanks, I will try it and will keep you posted

alexasha
07-01-2003, 08:50 PM
For some reason I keep getting following error:
Sound recorder was unable to complete the conversion.
I have tried PCM 11.025kHz, 16bit, Mono or " " " stereo

jack22182
07-01-2003, 08:58 PM
I posted the steps I followed to do this a while back. The Clie WAV file has to be converted to a format that Naturally Speaking can import. You also have to train NS by reading the training passages into the Clie and then training the NS software against the WAV files. It acutally works fairly well and the more you train the better it gets.

http://www.cliesource.com/forums/showthread.php?s=&threadid=4842

Have fun!

Jack

alexasha
07-01-2003, 09:03 PM
It looks like I was recording in LP (8kHz) mode and you recomend SP mode (22kHz). That might be a source of error. I will try to rerecord it. Thanks.