This paper proposes a speech recognition framework using a language-independent phonetic code system that locates in intermediate between the acoustic-phonetic layer and the language-dependent phone layer. Its original framework is directed toward an alternative to the conventional sample-based statistical methods comprised of phonetic environment-dependent techniques. In the first part, we describe the basic framework of the proposed processing, and the next, specific procedures for extracting acoustic features using non-linear spectral enhancement and for deriving the language-independent Intermediate Phonetic Code (IPC) set. A recognition system is implemented as phrase spotting in symbolic domain on the basis of this framework. In the last, preliminary results of recognition experiments are shown, in which isolated word recognition using the language-independent IPC set is examined compared with that of a language-dependent phone set.
Cite as: Tanaka, K., Kojima, H. (2000) A speech recognition method with a language-independent intermediate phonetic code. Proc. 6th International Conference on Spoken Language Processing (ICSLP 2000), vol. 4, 191-194, doi: 10.21437/ICSLP.2000-783
@inproceedings{tanaka00_icslp,
author={Kazuyo Tanaka and Hiroaki Kojima},
title={{A speech recognition method with a language-independent intermediate phonetic code}},
year=2000,
booktitle={Proc. 6th International Conference on Spoken Language Processing (ICSLP 2000)},
pages={vol. 4, 191-194},
doi={10.21437/ICSLP.2000-783}
}