Phonemic Category Experiments.
Final technical rept. Jan 69-Jan 70,
PHILCO-FORD CORP WILLOW GROVE PA
Pagination or Media Count:
This report describes the efforts undertaken to improve the experimental model Voice Sound Recognizer originally built under Contract AF 30602-67-C-0300. This equipment utilized the techniques of Single Equivalent Formant parameter extraction, phonemic category recognition, and category-sequence word recognition. Extensive hardware and software modifications to the basic recognizer system were made during the program which include the use of semiautomatic speaker adaptation by means of distance functions defined by sets of phonemic category strings and nearest neighbor word recognition decisions. The final recognizer configuration displayed a reduced speaker sensitivity and an average recognition rate for four speakers of 85 when using a 25-word vocabulary. Author
- Computer Programming and Software
- Voice Communications