Title :
Recognition confidence scoring using recognition results from perturbed input feature vectors
Author :
Jeong, Yongwon ; Kim, Hyung Soon
Author_Institution :
Dept. of Electron. Eng., Pusan Nat. Univ., South Korea
fDate :
8/30/2001 12:00:00 AM
Abstract :
New information to score confidence for out-of-vocabulary (OOV) rejection is introduced. Using recognition results from perturbed input feature vectors, additional information is attained and then combined with a conventional method. Simulation results show that the information can be effectively used for improving OOV rejection performance
Keywords :
error statistics; perturbation techniques; speech recognition; vectors; OOV rejection performance; automatic speech recognition; equal error rate; out-of-vocabulary rejection; perturbed input feature vectors; simulation results; speech recognition confidence scoring;
Journal_Title :
Electronics Letters
DOI :
10.1049/el:20010776
