I spent the summer of 1995 working at MIT. The group that I was working for is the Hearing Aid Research Group, which is a subgroup within the Aids for Sensory Communications Group of the Research Laboratory of Electronics at MIT . One of the projects that my group is working on is a speech recognition system.
It takes only one word to describe the work that I was doing: sentences. Why sentences? Simple, when one is testing a speech recognition system, one needs to have a lot test sentences to train a computer with. I cannot possibly count the number of sentences that I had to deal with, but I should mention that everyone else in my group probably had had more exposure to test sentences than I had, so I didn't complain. While I was working, I recorded, transcribed, and edited test sentences many, many times. It has got to the point that I had some of the sentences memorized.
In addition to working on the speech recognition system, I also did some work for Jeanie Krause, a graduate student who completed her master's thesis that summer. For her thesis, she studied the effects of factors such as speaking rate, volume, and separation of words on the intelligibility of speech in noisy situations. Once again, the work that I did involved sentences. However, the sentences that she used were sentences that could have been produced by only a computer. The sentences were randomly generated by a computer. They were grammatically correct, but made absolutely no sense. Some of them were quite amusing. Even more interesting were what some of the subjects wrote when they tried to dictate what they heard, or, more accurately, what they thought they heard. I'll put some of the test sentences and interesting responses here when I get the chance to.
R.B. Vilim, H.E. Garcia, and F.W. Chen. "An Identification Scheme Combining First Principle Knowledge, Neural Networks, and Likelihood Ratio Test," IEEE Transactions on Control Systems Technology, Volume 9, Issue 1, Jan. 2001, pp. 186-199.
R.B. Vilim, H.E. Garcia, and F.W. Chen. The United States of America as represented by the United States Department of Energy. Statistically qualified neuro-analytic failure detection method and system. 2002. U.S. Patent 6,353,815.
D.H. Staelin and F.W. Chen. "Precipitation Observations Near 54 and 183 GHz Using the NOAA-15 Satellite." IEEE Transactions on Geoscience and Remote Sensing, Volume 38, Issue 5, Part 1, Sept. 2000, pp. 2322-2332.
W.J. Blackwell, J.W. Barrett, F.W. Chen, R.V. Leslie, P.W. Rosenkranz, M.J. Schwartz, D.H. Staelin. "NPOESS Aircraft Sounder Testbed-Microwave (NAST-M): instrument description and initial flight results." IEEE Transactions on Geoscience and Remote Sensing, Volume 39, Issue 11, Nov. 2001, pp. 2444-2453.
