MP3Car.com - View Single Post - SD and voice recognition

parksgm · 01-31-2008, 12:37 AM

Quote: Originally Posted by S4Per

Take a look at my post over here. I've been using SD voice recognition for the past few weeks, and it's been 99.5% accurate, even at highway speeds.

I bet some basic noise cancellation will help you tremendously. I use the NC that's included with my Realtek Audio HD on board sound card.

Quote: Originally Posted by S4Per

I don't think there's one perfect tool for safe input, but rather a collection. Here's what I use:

1. Touchscreen
2. ccStick (i.e. rotary encoder)
3. VOICE!

I use Streetdeck for my front end which has very thorough voice commands. I use the rotary encoder to trigger the voice input (never have to look down at it, it's in the same spot as my car's OEM nav rotary control. My hand naturally rests there if using the arm rest). So, I use voice to go to different modules, pause, skip tracks, create a route, route me home, tell it to tell me the next nav direction, go to the equalizer, call home, and about 30 other things.

If I'm looking for artists to play, I use the ccStick to scroll the list (requires periodic glance, so not ultra safe, but better than holding down a button on the touchscreen). I can click the ccStick to select/ hit OK, and also use it to 'go back'. Between voice and ccStick, i'm barely even using the touchscreen at this point while driving.

Also, btw - Dragon Naturally speaking is great for normal PC control. With one command, I can open my OBD-II software of choice, select my template of 32 variables to log, and start logging.....or do a zillion other things.

EDIT: FYI - I'm using a cheap $15 mic that I picked up a few years ago, mounted by my sunroof controls where an Onstar mic would go. Key to good voice recognition is training the mic, and also having a finite set of commands available. For example: Dragon Naturally Speaking is the best voice recognition software available today, I could use it to navigate in Streetdeck. I don't, though, because Dragon is listening for anything like a command ("Call Home") to dictating text...it's constantly evaluating which one of the million things it can do, it should do. There is an option to put it into 'Command Mode' or 'Spell Mode' etc. but it's a bit of a hassle to move in and out of those modes. Streetdeck voice recognition doesn't have to work as hard: it just listens for commands, and there's only about 40 of them (that I've configured). So as long as the commands don't sound too much like each other, it's easy for it to figure out what I want. anyway...

Jst my 2 cents,
Per

Yeah, if you're using a $15 mic and getting consistent results, the noise cancellation must be working wonders. We have essentially the same setup...I just have a Powermate mounted on a very shortened joystick base in place of your ccStick, I have the Powermate set up to trigger voice command on a long press (click), and the microphone is mounted essentially in the same place as yours.

Right now I'm using the onboard sound from my Dell Gx240. My two PCI slots are taken up by my ATI video card and a USB 2.0 card, both of which I need. If I could find a USB sound card with noise cancellation features built in, then that might be a possible solution.

From the little bit I've experimented with it, voice control seems to be far superior to touchscreen control...I'd be willing to invest in a decent sound card if it worked.

Thanks for the advice...hopefully this will help others, as there just isn't that much in the forums on SD and voice recognition.

01-31-2008, 12:37 AM	#3
parksgm FLAC is for flaccid Join Date: Dec 2005 Location: Dallas, Texas Posts: 1,117	Quote: Originally Posted by S4Per Take a look at my post over here. I've been using SD voice recognition for the past few weeks, and it's been 99.5% accurate, even at highway speeds. I bet some basic noise cancellation will help you tremendously. I use the NC that's included with my Realtek Audio HD on board sound card. Quote: Originally Posted by S4Per I don't think there's one perfect tool for safe input, but rather a collection. Here's what I use: 1. Touchscreen 2. ccStick (i.e. rotary encoder) 3. VOICE! I use Streetdeck for my front end which has very thorough voice commands. I use the rotary encoder to trigger the voice input (never have to look down at it, it's in the same spot as my car's OEM nav rotary control. My hand naturally rests there if using the arm rest). So, I use voice to go to different modules, pause, skip tracks, create a route, route me home, tell it to tell me the next nav direction, go to the equalizer, call home, and about 30 other things. If I'm looking for artists to play, I use the ccStick to scroll the list (requires periodic glance, so not ultra safe, but better than holding down a button on the touchscreen). I can click the ccStick to select/ hit OK, and also use it to 'go back'. Between voice and ccStick, i'm barely even using the touchscreen at this point while driving. Also, btw - Dragon Naturally speaking is great for normal PC control. With one command, I can open my OBD-II software of choice, select my template of 32 variables to log, and start logging.....or do a zillion other things. EDIT: FYI - I'm using a cheap $15 mic that I picked up a few years ago, mounted by my sunroof controls where an Onstar mic would go. Key to good voice recognition is training the mic, and also having a finite set of commands available. For example: Dragon Naturally Speaking is the best voice recognition software available today, I could use it to navigate in Streetdeck. I don't, though, because Dragon is listening for anything like a command ("Call Home") to dictating text...it's constantly evaluating which one of the million things it can do, it should do. There is an option to put it into 'Command Mode' or 'Spell Mode' etc. but it's a bit of a hassle to move in and out of those modes. Streetdeck voice recognition doesn't have to work as hard: it just listens for commands, and there's only about 40 of them (that I've configured). So as long as the commands don't sound too much like each other, it's easy for it to figure out what I want. anyway... Jst my 2 cents, Per Yeah, if you're using a $15 mic and getting consistent results, the noise cancellation must be working wonders. We have essentially the same setup...I just have a Powermate mounted on a very shortened joystick base in place of your ccStick, I have the Powermate set up to trigger voice command on a long press (click), and the microphone is mounted essentially in the same place as yours. Right now I'm using the onboard sound from my Dell Gx240. My two PCI slots are taken up by my ATI video card and a USB 2.0 card, both of which I need. If I could find a USB sound card with noise cancellation features built in, then that might be a possible solution. From the little bit I've experimented with it, voice control seems to be far superior to touchscreen control...I'd be willing to invest in a decent sound card if it worked. Thanks for the advice...hopefully this will help others, as there just isn't that much in the forums on SD and voice recognition. __________________ Monthly North Texas Meet thread My worklog gDrive Install pics How to add your CarPC specs to your username An amateur built the Ark. The Titanic was built by professionals.