View Single Post
Posts: 81 | Thanked: 45 times | Joined on Dec 2009
#53
Originally Posted by Gadgety View Post
Yes. I'd love to see that as well. It seems to take a lot of computing power, though. I know there are programs for Windows OS (Dragon Naturally Speaking, Windows Vista speech to text) but they require powerful computers or else it takes too long to execute. The OQO was able to do it with one giga of RAM, just. I believe there are some cloud based utilities that provide this, for a substantial fee, of course, and tied to certain phone providers only. Furthermore, I understand (well my understanding is based on the past few days acquaintance with maemo, so it is very limited) there currently are no voice recognition apps for / in maemo.
Have a look at Wikipedia and you will see that Naturally speaking has been around since the early ninetees. The computers in those days were a couple of order of magnitudes weaker than the N900. If those computers could manage voice processing so can the N900. But also remember that to do voice dialing it is not a question of the much more difficult free speech recognition, but only of mapping a short sound stream to one out of e.g. 200 distinct names in your phone book. But I don't even want the phone to do the speech to text mapping. I would settle for making my own recording for each phone entry. The software would then only need to search for the entry with the best correlation of the phone entries to the search entry, which is a much simpler problem. That's the way voice dialing worked in my old Siemens phone, btw.
 

The Following User Says Thank You to dov For This Useful Post: