Evaluation method for a voice recognition system modeled with discrete Markov chain

Author

Osada, Hiroyasu

Author_Institution

Railway Tech. Res. Inst., Tokyo, Japan

Volume

2

fYear

1997

fDate

20-22 Aug 1997

Firstpage

600

Abstract

In this paper, we suggest an evaluation method for a voice recognition system modeled with discrete Markov chain. Our evaluations are mainly dependent on two characteristics of a system: 1) correctness of recognition for input data and 2) number of utterances. The main frame of the voice recognition system is composed of: 1) a voice input unit, 2) a voice recognition unit and 3) an output unit. Generally, a voice recognition system has its own list of words to recognize a word correctly. It means that the words which the system can recognize are restricted. A simple system can be expressed as a state transition diagram. Then we can make a transition probability matrix and calculate the state probability for each state. This means that we can evaluate the system numerically. Moreover we extend the simple model to the more general one and get numerical results to evaluate the two characteristics for the system

Keywords

Markov processes; matrix algebra; probability; speech recognition; discrete Markov chain; evaluation method; numerical result; output unit; state probability; state transition diagram; transition probability matrix; voice input unit; voice recognition system; voice recognition unit; Character recognition; Equations; Humans; Probability; Rail transportation; Speech recognition; State-space methods;

fLanguage

English

Publisher

ieee

Conference_Titel

Communications, Computers and Signal Processing, 1997. 10 Years PACRIM 1987-1997 - Networking the Pacific Rim. 1997 IEEE Pacific Rim Conference on

Conference_Location

Victoria, BC

Print_ISBN

0-7803-3905-3

Type

conf

DOI

10.1109/PACRIM.1997.620334

Filename

620334