OnStar vs. empeg's Future Voice Recognition

I just saw an OnStar system in action, and i must say, i was impressed with the ability of the system to Recognise your voice, and do things like dial the phone. I was esp interested in the way that it interfaces with you. It will say things like, "pardon me?" and "ready" and other sayings that talk back to you as fast as a human. if you talk too fast, it will say, "please talk slower...".. i was impressed with how easy it was to use.

Can anyone in the development of the empeg's Recognition program compair the quality and/or features of emepg's future Recognition System and OnStar's system?



-CHiP

> Can anyone in the development of the empeg's Recognition program compair the quality and/or
> features of emepg's future Recognition System and OnStar's system?

No because right now we're testing five different voice recognition systems - plus I've only used OnStar for a few minutes.

We're aiming for something similar, though.

Rob

Any of them open source or affordable for the linux platform ? Looking for one now that I have gotten into X10 and fitting my recently acquired house with it.

I know about the IBM package and that's about it.

Cheers,
Hans


Mk2 - Blue & Red - 080000431

Actually, one of the packages we're evaluating is the same VR engine used in GM's Onstar system :)

Hugo

Will there be some way to use upcoming VR with MK1? Maybe its possible to add external microphone to serial port or smthng. Will there be a MK1 version of VR software to use with external mic? I would desperately love to have VR but i can't afford to upgrade to MK2 in near future.


Veiko
[email protected]

MK1 #317 - 20GB green

I doubt it very much. As well as the microphone problem, VR systems have a large-ish memory footprint which could be a problem with the Mk.1.

Rob

The one's we're looking at are all commercial (intended for OEM applications).

Rob

Looking for one now that I have gotten into X10 and fitting my recently acquired house with it.

X10 in Belgium? I thought there where only a handful of devices available designed for the european 220/230v, while the more interesting ones were usually for 115v only. Its a while back now, that i last checked though.

We use X10 in the office, to drive the Broken Build Beacon, which flashes when someone screws up.

Rob

the Broken Build Beacon, which flashes when someone screws up.

Bad, bad, naughty Zoot!

___________
Tony Fabris

http://www.x10europe.com/ ... nuff said :-).

Though I must admit not all of the gadgets are as readily available. Still ... the bulk of the usable modules are there and quite a few of the gimmicky ones too. With links to where to buy.

Got myself a couple of starter kits first and am gunna make an inventory of the rest soon.

Cheers,
Hans


Mk2 - Blue & Red - 080000431

With 12MB in the MKII, will there be enough memory for the voice reg? How much memory is used up currently with version 1.02? I would imagine that the voice reg would take up a lot of memory.

Also, is there a web site somewhere that goes into depth on how the voice reg progam works?

-CHiP

The deliberately put more RAM into the Mk2 so that they could do voice recognition at a future date. The Mk2 should have enough RAM to do the job.

There is no web site detailing how the voice recognition will work, because there isn't any voice recognition yet. They are still evaluating methods and have not settled on anything yet.

___________
Tony Fabris

The VR will probably involve spinning the disk up to load extra voice tables; ie, in normal operation it will recognise the key word to start recognition (eg "empeg") but before speaking further commands you may have to wait for an audio prompt.

The player takes up a couple of megs, the rest is linux and cache generally. VR footprints that we've been looking at are in the 1mb-4mb range.

Hugo

okay, thanks for the info.

-CHiP

That sounds like the Nuance engine. That engine is quite good, and works well over the phone, with static and background noise. It does stumble sometimes though. With the OnStar system there are two parts going on: You have the "personalities" piece built on top of the Nuance Engine. The Personalities module is from General Magic. The Nuance is purely the recognition piece. I'm more impressed with the personalities stuff than the nuance engine actually ;).

Calvin

I'm curious as to what the expectation is for Voice Recognition regarding features. I mean, will it virtually replace the remote for searching and tweaking playlists or will it be mostly for simple commands like volume up/down and next/prev track? Or is it still far too early days for such considerations?

___________________________

Steve - 18GB Mk II Sssmokey

I know what Doug wants it to do...

___________
Tony Fabris

I wonder how well that Voice recognition works when you've got your music at a "decent" volume level. (like is mostly the case with me). Won't the music interfere with the voice commands much?

Well they have both the music-speech (mic) and the pure music signals inside the empeg, so if they have problems here and want to be REALLY tricky they can subtract the music signal from the microphone input and end up with just the speech (+ engine noise etc.). Anything is possible

(list 6284, Mk1 S/N 00299 4GB blue [for sale]. Mk2 S/N 080000094 6GB blue)