Silicon Investor (SI) -- The First Internet Community

STOCKTALK

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor. We ask that you disable ad blocking while on Silicon Investor in the best interests of our community. If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.

Strategies & Market Trends : Gorilla and King Portfolio Candidates -- Ignore unavailable to you. Want to Upgrade?

To: Apollo who wrote (20339)	3/16/2000 4:05:00 PM
From: Thomas Mercer-Hursh	Respond to of 54805

I have been interested in speech recognition for several years. I have tried Dragon systems/IBM ViaVoice and been disappointed with the inordinate amount of time necessary to train the computer I haven't kept up with speech recognition for a while, but let me throw out a little historical tidbit that might be of interest. Many years back, a small lab in MA got some patents for speech recognition based on phase-shift patterns instead of frequency patterns. They argument for this I find quite good since the information in speech is conveyed by the unique patterns resulting from the shape of the mouth, tongue, etc. A bass and a soprano may be using virtually exclusive ranges of frequencies, but the shape of their mouth, throat, tongue etc. will be the same to articulate the same word or phoneme. But, the funny part was that they developed this stuff based on an interest in human-dolphin communication, not speech to text. The last stuff I saw from them was about 1965 when they were able to have sensors that responded to particular phase-shift patterns, regardless of the speaker, with very high reliability levels, from the perception of the dolphin. After that, everything they did was classified top secret since they were working for the Navy and I have never heard from them again, although I did get one "leak" later on that indicated they were continuing to be very successful, still with dolphins. Ever since I can't help but feel that all of the people out there doing frequency pattern analysis, i.e., everyone in the field as far as I can tell, isn't barking up the wrong tree, succeeding as well as they do mostly because computer power has advanced enough to sortof solve the problem by brute force, but still requiring a lot of training to get the kind of accuracy one would really like. Sounds like an opportunity for a discontinuous innovation!