Hi Bob,
Forgot to tell you one more thing. I'll try to sum up what John Oberteuffer told me today on the phone about the difference of an hmm embedded tech. and a fonix embedded tech.
To start with he said that any tech could be scaled to a chip, the differences though are that to embed hmm you HAVE to make an accuracy/speed/memory trade-off. The fonix technology being neural net based is much easier to scale down to a chip without making above mentioned trade-offs. So if it has been said until now, including myself, that HMM can not be scaled to a chip, in absolute terms is wrong. Realistically though one cannot expect a robust, accurate, big vocab, user independent continuous speech chip embedded HMM based VR product.
Bob, I'm just quoting, or at least trying to , an expert in this field. One that is able to judge best the two techs in question, since he profoundly knows LHSPF and Fonx. Now, you will certainly shoot with an aweful lot of technological questions into my direction, but I'm afraid that most probably I'll not be able to answer them. I hope some techy here lurking might jump in and help me out of this forthcoming avalange of difficult questions.<bg>
This time definitely good night,
Frank |
