In General, keyword spotting systems employ the connected word recognition network, which consists of both keyword models and filler models as the recognition strategy. That is why those systems cannot construct the language models of word appearance effectively for detecting keywords in large vocabulary continuous speech system that has large text data. In this paper, we propose a keyword spotting system using pseudo N-gram language model for detecting keywords. We then investigated the performance of the proposed system according to the changes of the frequencies of appearances (uni-gram) of both keyword and filler models. As the results, when the uni-gram probability of keyword and filler models set to 0.2 and 0.8 respectively, the experimental results showed 8.6% of equal error rate. This means that the proposed has 14% of higher performance than conventional methods in terms of error reduction rate.
Cite as: Kim, J.-G., Jung, H.-Y., Chung, H.-Y. (2004) A keyword spotting approach based on pseudo N-gram language model. Proc. 9th Conference on Speech and Computer (SPECOM 2004), 256-259
@inproceedings{kim04_specom,
author={Joo-Gon Kim and Ho-Youl Jung and Hyun-Yeol Chung},
title={{A keyword spotting approach based on pseudo N-gram language model}},
year=2004,
booktitle={Proc. 9th Conference on Speech and Computer (SPECOM 2004)},
pages={256--259}
}