In this paper a new method for automatic segmentation of continuous speech into phone-like units is addressed. Our method is based on a very fast presegmentation algorithm which uses a new statistical modeling of speech and searching in a multilevel structure, called Dendrogram, for decreasing insertion rate. Performance of algorithms have been tested over a large set of TIMIT sentences. According to these tests, our final segmentation algorithm is capable of detecting nearly 97% of segments with an average boundary position error of less than 7 msec and average insertion rate of less than 12.7%. In addition to acceptable precision, our overall segmentation scheme has very low computation cost and it can be implemented in real time on an average Pentium PC. The major advantage of presented algorithms is that no training or threshold estimation is needed in realizing them. Details of proposed algorithms and their performance results are included in the paper.
Cite as: Gholampour, I., Nayebi, K. (1998) A new fast algorithm for automatic segmentation of continuous speech. Proc. 5th International Conference on Spoken Language Processing (ICSLP 1998), paper 0182, doi: 10.21437/ICSLP.1998-591
@inproceedings{gholampour98_icslp,
author={Iman Gholampour and Kambiz Nayebi},
title={{A new fast algorithm for automatic segmentation of continuous speech}},
year=1998,
booktitle={Proc. 5th International Conference on Spoken Language Processing (ICSLP 1998)},
pages={paper 0182},
doi={10.21437/ICSLP.1998-591}
}