Title :
VOICED-a small dialog system
Author :
Urlep, I. ; Horvat, B. ; Kacic, Z.
Author_Institution :
Fac. of Tech. Sci., Maribor Univ., Yugoslavia
Abstract :
A small dialog system called VOICED (VOICE eDitor), which allows voice activation of editing commands, is described. The system has two main modules: the text processor and the speech recognizer. The vocabulary consists of 18 Slovene words. Each word is modeled with a five-state hidden Markov model. In the learning phase, the Baum-Welch learning algorithm is used. A recognition accuracy of 94.5% has been achieved for the speech database used
Keywords :
Markov processes; speech recognition; text editing; Baum-Welch learning algorithm; HMM; Slovene words; VOICED; dialog system; editing commands; five-state hidden Markov model; recognition accuracy; speech database; speech recognition; speech recognizer; text editor; text processor; voice activation; voice editor; Automata; Filters; Hidden Markov models; Quantization; Signal analysis; Speech analysis; Speech recognition; Text recognition; Vectors; Vocabulary;
Conference_Titel :
Electrotechnical Conference, 1991. Proceedings., 6th Mediterranean
Conference_Location :
LJubljana
Print_ISBN :
0-87942-655-1
DOI :
10.1109/MELCON.1991.162053
