R. Paul McCarty wrote: > But I think the various GPL'd linux speech recognitions, Sphinx, and > ViaVoice are like apples and oranges. Correct me when I'm wrong here, > but for a wearable application a comparison might look like the > following: > > linux status system req. open/GPL corpus variable > freespeech y pre-alpha ? y ? > CSLU toolkit n beta ? y y > MSstate ISIP y alpha ? y y > Sphinx y alpha variable y y > ViaVoice y stable high n n? > > It seems to me the two critical criterion for incorporating speech > recognition into a wearble, are low system requirements and variable > corpa so you can fine tune the system to only respond to a small > vocablulary. This would make the system more compact, responsive, and > reduce processing time and memory. Okay, there's only one requirement. > :) I don't think ViaVoice will let you configure it with a smaller > vocablulary, since it's focus appears to be dictation. Although, I'm > probably wrong since there is info about configuring ViaVoice for > applications and there's probably a way to specify the vocabulary. Yes, you're wrong. ;-) ViaVoice lets you define your own vocabularies and grammars, which would be very useful if I could get it to install properly. Also, VV requires you train on the Windows side, which is icky. > Hm, so maybe these are really pretty close to one another. But ViaVoice > definitely has larger system requirements. Most of the GPL'd software is > less then 5MB including the dictionaries, and ViaVoice wants 100MB. On the plus side, the ViaVoice recognition is pretty darn good if you use a good microphone. > -Paul Alex Feinman -- Subcription/unsubscription/info requests: send e-mail with subject of "subscribe", "unsubscribe", or "info" toWear-Hard Mailing List Archive (searchable): http://wearables.blu.org
From Wear-Hard Mailing list Archive (WH)
Maintained by R. Paul McCarty
Archive created with babymail