Author(s)

E.A. Yfantis and J.D. Elison

Abstract

Vector interpolation for time alignment in speech recognition E. A. Yfantis and J. D. Elison Department of Computer Science, University of Nevada, Las Vegas, 4505 S Mary/aWPM,, &w yzgaj, W 89734-407 9, (/&4 Email: yfantis@cs.unlv.edu and elisonj@cs.unlv.edu Abstract In this paper, a new algorithm for time alignment in speech recognition is presented. This new algorithm, Vector Interpolation, has a classification rate comparable to that of Dynamic Time Warping. While Vector Interpo- lation is able to match Dynamic Time Warping for recognition accuracy, it requires significantly less computation, making it much faster than DTW based algorithms. Both algorithms are presented and a comparison of the two is made. 1 Introduction Automatic speech recognition has applications in many areas. From contin- uous recognition for dictation systems to isolated word spotting systems for information retrieval and command and control, speech recognition is being used

Keywords