Drew Vassallo wrote: > I'm not sure what the original question was supposed to ask, I guess. > To perform TRUE voice recognition, I seriously doubt it could be done > within 8K worth of instructions, as in a PIC. I think that it could be done in much less space if a suitable preprocessor would be available. Something like a chip that normally drives real time audio spectrum displays with suitable narrow bandwidths would be very nice imho. Any ideas along this line ? Someone (Samsung ?) makes such chips cheaply, there remains to find a way to interface them without using too many resources. In theory, if the speech sample can be analyzed in the frequency-power distribution domain in real time and a normalized set of characteristics generated, then the job of recognition would be much simplified, perhaps just a walk in a stored tree in the simplest case. I'd be happy if a 70% recognition rate could be achieved for 2-10 words with some room left for a simple application (door opener, timer etc). With time and effort it should improve. Peter -- http://www.piclist.com hint: The list server can filter out subtopics (like ads or off topics) for you. See http://www.piclist.com/#topics