Given up cause getting recognition in real world driving conditions sucks balls.
Hi everyone. I'm starting a thread for Mac speech recognition because I'm planning on installing a Mac Mini in my car and I want to be able to control *almost* everything by voice.
It's been awhile since I've used a Mac but the Mini, IMO, is the most attractive vessel for a carpc out there. As such, I've been reading up on Mac OS X's Speech. While it's great that you can do alot of stuff with the Applescript + Speech combination, you cannot, to the best of my knowledge obtained from what little I have gelamed from these forums and elsewhere on the web, for instance, in iTunes, say "search 'Greenday', play" and iTunes would search for all songs with "Greenday" in the tags and play them. And spoken commands such as "find corner of Shaw Ave. and Willow" in Route 66 are impossible for Applescript to carry out.
I did come across a program called iListen, by MacSpeech, which claims to be able to, in addition to almost everything Mac OS X's Speech can do, convey spoken words into text in any field of any program. Well if this is true then I can use this in the search field of iTunes. Great!, if it's true. I have have not used Route 66 however I assume there's a similar field in which you type addresses.
Anyway, I just wanted to know about what other people have done in this matter. What solutions have you guys come up with? Is anyone out there a veteran of iListen? How usefuld would iListen be for what I'm proposing? Am I wrong about Applescript and what it's capable of? I hope I'm not the only one with these concerns. Good luck to everyone with their projects!
Lol I think I'd have to agree with this. I have enough trouble with my bluetooth headset in my noisy truck, nevermind a computer.Originally Posted by god_of_cpu
Mac: - Researching and Awaiting Funds
Changed to planning for a Mac Mini Trukputer...w3rd.
I'm planning on using Griffin's iMic w/ a good mic. Any suggestions on a mic?
While I haven't had much time to research the issues surrounding speech recognition in a driving environment, I have read through the apple documentation and it looks promising that you could do anything you want with it if you are a developer. However you are probably not, so iListen seems like the better solution for you. iListen supports apple script, although I don't see any where on the site that you can fill in the blanks in the apple script. ie "search 'BLANK'.
That is something you might want to ask the company before you spend the money. They also seem to have an iTunes ScriptPak, http://www.macspeech.com/store/produ...roducts_id=188. That might be something to take a look at.
If I would contact the company and ask if their product supports that use before going out and purchasing the software.
I'm just now learning how to develop in Applescript and my sense is that yes, you can do the search function you are describing in applescript. I know that you can easily access the speech recognition server via applescript and once you have what it recognizes as input, can search iTunes playlists or song names, etc. There ought to be no reason you can't and that's a good thing because I intend to control my Mac via speech as well. I've had relatively decent results doing it with my PC, even a car environment.
What will probably limit the usefulness of this approach is that what the system recognizes is often not 100% of what you said. So "search 'Greenday', play" may come across as "erch, greendai, ply". A good speech recognition engine, which the Mac supposedly has, should be able to turn that into "search, greendai, play".
The approach I'm considering is something a little less intuitive but breaks the problem into chunks the computer can solve. Something like this:
You: "Computer, music search"
Comp: "Artist, album, or title?"
You: "Artist Greenday"
Comp: "Found Greenday"
If it didn't find it, I'd pop up a text box and the computer would ask you to spell the artist. Each letter that goes in would winnow down the list until you either select artist 1, 2, 3, or if there was only one left, I'd just play it.
Since you know the playlists, artists, song names, and albums, all of these can be used to form a "speakable" items list that the speech server can compare against when trying to decide what you are saying. That helps increase voice recognition accuracy as well.
I don't think this will be easy, but I do think it can be done. How well it works in a car environment remains to be seen, but that's what this hobby is all about, isn't it?