vern@zebra.UUCP (Vernon C. Hoxie) (11/12/88)
>>Lenny Tropiano writes: >>Voice Recognition, how? That's the question... Since my >>involvement with the Voice Power product on the UNIX pc, I've learned >>a lot. Learning bits and pieces about CODEC's, PCM (pulse code modulation), >Charles Brunow writes: > Ultimately the process probably consists of determining the > coefficients for the filter nodes and looking for the best > match with the set of known words and updating the coefficients > either completely or with a damping factor for learning. Vern Hoxie (me) writes: I am also interested in trying to unlock the mysteries of speech recognition by computer. The method I have considered using was to build a set of files of the various phonemes. These would be extracted from recorded phrases by a variety of voices. The extraction procedure would be using an editor similar to the 'voice editor' supplied with the voice software. After a significant supply of samples, I would try a number of analytical tools such as a Fast Fourier Transform, envelope curve matching etc. There would be a lot of just looking at wave patterns and filter outputs to try to recognize something distinctive about each phoneme. The next step would be to mold these phonemes into English spelling. This would require a very impressive dictionary. Because of this requirement, I thought of the comp.cog-eng group on USENET and decided to cross-post this discussion to them with the hope that they may review my short outline above and offer suggestions and point us in a direction where we will have the best possibility of success. I really don't expect our unix-pc's to be able to do real time conversions. Hardware of the nature of Job's NeXT machine or something more specific will be required for that. However, getting a reasonable text entry from voice input in a day or two would be an encouraging first step. -- Vernon C. Hoxie {ncar,nbires,boulder,isis}!scicom!zebra!vern 3975 W. 29th Ave. voice: 303-477-1780 Denver, Colo., 80212 uucp: 303-455-2670