Voice Recognition?

[jkmonroe]

speaking from some personal experience, I can say that theres a reason that radio stations and the like use 10's of thousands of dollars worth of voice equipment. The same would hold true for VR, so I highly doubt this would be a project that would require anything less than a coupla-few grand on the inside buying new equipment. :-D

[IVB]

Well, I just played the jkmonroe game and got an AP800 off eBay for $74 shipped (as-is, but the pretty buttons are known to work). Hopefully a few of the mic inputs actually work.

Trying to figure out what other pieces/parts I need to do a test. The parts closet was looking a little too empty anyhow...

[Acropolis]

I have worked with VR professionally for many years (in vehicles and on telephone systems). Over a phone without background noise it works fantastic. Even natural language (full sentences) work very well. Even a speakerphone with no background noise works fairly well. Once you start to talk about multi-directional microphones (so that you can pick up sound from all areas of a room) with the potential for lots of different background noise you are talking about a completely different game. You could probably get a directed dialog to work OK (system asks very directed questions and you provide a very specific response), but expecting the system to always be listening for any variety of commands is not likely to happen without lots of programming and tuning.

Also keep in mind that low end VR systems must be trained and typically only work well with the trained voice. Once you start talking about ASR (Advanced Speech Recognition) that will work with many voices and no training, the cost shoots up fast. Just the software licenses are going to cost you around $1500 per listening port. One port would be the equivalent of one zone that you want to listen to. If you want to listen to 5 rooms/zones, you would need to buy licenses for 5 ports.

Expensive microphones and mixers are completely unnecessary. You can get 95-98% recognition rates over a telephone with no training. The microphone in your average phone probably costs less than $1.00. Also keep in mind that the frequency response over a phone line is very narrow (do a search on mu-law if you are interested in the specs). I personally wouldn't invest in expensive equipment thinking it is going to make a big difference.

I think the best compromise you can hope for is to wait for Vista support. You will still have some challenges, but you will probably get some limited usage out of it. You could either create a "command" or "talk" button on a touch screen so that the system will pause music and listen. Or you may have some success in a quieter room. Also keep in mind that most VR systems are only designed to listen for short periods of time (5 seconds or less).

Acropolis