How hard would it be to add voice command capability similar to a cell phone. That is, you record the command ahead of time (person's name in case of cell phone address book) and link it to a menu command. Not being a programmer, I assume you would need a voice recorder program as well as a program that could identify a recorded message and of course hijack to map it to a button press.

Perhaps, this isn't very helpful, but to give you a rough idea: hard.

The empeg team tried working on a more complete solution than that your described, but if failed completely when you introduced the background noise in the car. This would be cool, but it'd be quite hard to get working properly.

John

Ahhh. So one limitation is the background noise.

Would it be possible to record the background noise once and substract it from the input? Then the signal to noice ratio could be improved.

Background noise is, I'm sure, fairly random.

I meant the "driving sounds" from the motor and the tires. I know that this depends on speed, though.

empeg team tried working on a more complete solution

At the time, the software available just wasn't up to the job -- it couldn't cope adequately with the noise levels in a car, and it required too much CPU to do the job.

The situation was improving (and is probably up to the job now), but car player development got canned, so we had to stop work on the VR.

One problem that you'll likely face is that open-source (or free) VR packages just aren't up to the standard of the commercial stuff, and the commercial stuff is expensive to license.

I've just quoted you in the FAQ if that's OK...

Yeah, it sounds like the biggest problem is voice recognition even with one's own pre-recorded voice. I tried the voice commands on my phone. You really have to pronounce the command or person's name very carefully. If there is any noise in the background the phone is easily confused and asks you to repeat. My Onstar (when I was using the free 30 min of cell phone use) performed very well but as you say it 's probably very expensive to license if they'd let us use it at all.

Well, the empeg has a little more CPU power to work with and tons more space for storing higher bitrate voice samples than a cell phone. My biggest concern would be RAM. 16MB isn't enough to run the player, TTSd, empire, and emphatic, let alone adding VR and GPSapp to mix (not ot mention that I'd like to add ODBC monitoring in there too).

Too bad it couldn't just be a rather simple set of instructions. I'm sure recognizing an artist's name would be difficult, but a set of rather simple pre-known instructions would be fairly simple to deduce with background noise (as I would think. Instead of trying to guess what you were saying, it would obviously just have to determine a 'best match'). It would be pretty neat just to do something like 'Menu, Left, Left, Select, Play, Shuffle Random"

Now that Palantir is readily available, is there anyway that the voice recognition capabilities of a Palm Powered cell phone could be used to create playlists? I was just wondering. I've always liked the "jump" feature on the earlier versions of Winamp, and if anything could be made to duplicate that look-up, I would most definitely be in. (I'd have to upgrade to the Palm phone first...)

Hmm... Interesting... Which Palm cell phone has voice recognition?
Chances are that the voice recognition system is built into the phone/whatever app on the Palm and wouldn't be a generic API which anybody could call. Could be wrong though.

- Trevor

I'm pretty sure that the VR is built into the phone part of whatever phone you'd get (Samsung, Kyrocea, etc.) and is not available to the palm part of the device. There aren't any API calls for it anyway. You would probably need to have an ARM processor to work with it if it was available, also.