Implementation of an on-Line System for Automatic Recognition of Spoken Words.
Final technical rept.,
FEDERAL SCIENTIFIC CORP NEW YORK
Pagination or Media Count:
The report describes the implementation of a system which automatically recognizes spoken words in near real time. The techniques on which the system is based were developed over a period of 10 years. During most of this time, the research activities were devoted to the problems of selecting, extracting, and utilizing essential information-bearing parameters of speech. A non-real-time version of the speech analysis and recognition techniques which were developed was tested and demonstrated in 1967. Subsequently, the objective was shifted to implementing an on-line speech recognition system. An initial version was completed in 1969. Since then, the final objective of this long-term effort has been to improve the effectiveness of the analysis and classification procedures, speed their operation, and provide the user with wide flexibility in the operation of the system. The system described was implemented in a DDP-116 computer and is capable of learning and recognizing a vocubulary of 100 words of up to three syllables each. Author
- Voice Communications