EUROSPEECH 2001 Scandinavia
7th European Conference on Speech Communication and Technology

Aalborg, Denmark
September 3-7, 2001


The Effect of Time Stress on Automatic Speech recognition Accuracy When Using Second Language

Fang Chen (1), Jonas Sääv (2)

(1) Swedish Center for Human Factors, Sweden
(2) Linköping Univ., Sweden

The purpose of the present study is to compare the ASR performance when Swedish people speaking Swedish and English under time-stress and due-task performance. Fifteen university students (20 to 40 years of age, native Swedish language speaking) participated in the experiment. Three factors were studied: time-stress, which was manipulated by PWSP program. Two models of presenting the commands, one is by displaying the text on the screen and another is by headphone voice. Swedish and English languages were tested on Philips FreeSpeech 2000 speech recognition system. There is no individual voice file training and pre-designed grammar file for the speech recognition system. The results show that there are no interactions between any of the factors. The individual differences are large. There is a significant decrease of recognition accuracy (p<0.05) for both languages during stress. The recognition accuracy on Swedish language is significant higher (p<0.01) than English Language due to the Swedish accents.

Full Paper

Bibliographic reference.  Chen, Fang / Sääv, Jonas (2001): "The effect of time stress on automatic speech recognition accuracy when using second language", In EUROSPEECH-2001, 2737-2740.