This paper proposes imitated phoneme models that represent non-verbal sounds, especially cough sounds here, as phoneme sequences. The purpose of this research is to detect the cough sounds from user utterances accurately for the home health care task because coughing is one of the most important barometers to check the health condition. To deal with the variety of the cough sounds, the imitated phoneme models are constructed by clustering of phoneme sequences obtained in phoneme recognition. The experimental results show that this approach can improve the correct rates and the accuracies for words and coughs compared with the approach using HMM constructed from cough waveforms.
Cite as: Takahashi, S.-y., Morimoto, T., Maeda, S., Tsuruta, N. (2005) Detection of coughs from user utterances using imitated phoneme model. Proc. Interspeech 2005, 1357-1360, doi: 10.21437/Interspeech.2005-491
@inproceedings{takahashi05_interspeech,
author={Shin-ya Takahashi and Tsuyoshi Morimoto and Sakashi Maeda and Naoyuki Tsuruta},
title={{Detection of coughs from user utterances using imitated phoneme model}},
year=2005,
booktitle={Proc. Interspeech 2005},
pages={1357--1360},
doi={10.21437/Interspeech.2005-491}
}