July 28, 2009

So we got our results last night. Jivko made us a KGram program that ran and got decent results. Using both sensory modalities and all five interactions we were able to get a 92% accuracy in recognition rate, which is really good considering how similar a lot of the objects are. Although what's not so great is that what we intended to show, that the use of two modalities is better than either one alone, didn't come out as clear cut as we'd hoped. The combined modalities do much better than proprioception alone, but only marginally better than audio alone. Ah well, they are at least results. We sent our poster in to be printed, so that's not too bad. Jivko also ran global alignment last night, and today we're going to finish up some results with that instead of KGram. The preliminary results Jivko showed us were really good, so we should get much better results. We already sent the poster off to print, so that will only show our KGram results, but our paper should have results from global alignment. Our paper is about half way done as far as first drafts go; Ugonna did a good job of getting that started while Taylor and I ran the results.