Some performance benchmarks for isolated work speech recognition systems

✍ Scribed by L.R. Rabiner; J.G. Wilpon

• Publisher: Elsevier Science
• Year: 1987
• Tongue: English
• Weight: 972 KB
• Volume: 2
• Category: Article
• ISSN: 0885-2308
• DOI: 10.1016/0885-2308(87)90016-7

✦ Synopsis

Tbc performance of isolated word speech recognition system has steadily improved over time as we learn more about how to represcnl the significant events in speech. and how to capture these cvcntz vi;1 appropriate analysis procedures and training algorithms. In particular algorithms based on both template matching (via dynamic time warping (DTW) procedures) and hidden Markov models (HMMs) have bcen developed which yield high accuracy on several standard vocabularies, including the IO digits (zero to nine) and the set of 35 Icttcrb of the English alphabet (A-Z). Results arc given showinp currently attainable performance of a laboratory system for both template-based (DTW) and HMM-based recognizers. operating in both speaker trained and speaker independent modes. on the digits and the alphabet vocabularies using telephone recordings. We shou that the average error rates of these systems. on standard vocabularies. are significantly lower than those reported several ycarz buck on the exact same databases. thereby reRecting the progrcsb which has been made in all aspects of the speech recognition proccsh.