Daytona IEEE Section Calander of Events

Machine recognition of human speech has been under development since the early 60's. Everybody is familiar with the tedious telephone menus that we are forced to muddle through to communicate with our suppliers. The problems arise from two sources: poor recognition of phonemes and neglecting to relate the phoneme identification process to the context of the dialog. The basic approach is to isolate phonemes from the sound pattern. If the phonemes can be identified then there is a one to one correspondence between a sequence of phonemes and a word. Unfortunately, phonemes are hard to isolate and so a statistical prediction algorithm, called a Hidden Markov Model (HMM) is used. Another method is called Dynamic Time Warping (DTW). Vocoder patterns (frequency vs. time) are presented and it is shown that they are basically pictures of a sequence of phonemes blended together. DTW can adjust for speed and emphasis in speech. The HMM can help identify phonemes by their relationship to adjacent phonemes. This presentation includes a demonstration of voice recognition and vocoder operation.

Professor Wichmann graduated from MIT with a degree in Physics and did his graduate studies at Ohio State and UCLA in Electrical Engineering. His career specialty was Electronic Warfare and he is a charter member of the AOC (Electronic Defense Association). He joined the IEEE in 1952. He worked for the Air Force as an Electronic Scientist and for Litton, Teledyne and Hughes Aircraft Company in various engineering and program management capacities . He did his Speech Recognition work originally at Litton in the 1960s. Mr. Wichmann has published over 50 papers in various scientific journals and given dozens of technical presentations at technical symposia and conventions over the years.