A Comparison of Different Approaches to Automatic Speech Segmentation.pdf

A Comparison of Different Approaches to Automatic Speech Segmentation Kris Demuynck and Tom Laureys? K.U.Leuven ESAT/PSI Kasteelpark Arenberg 10 B-3001 Leuven, Belgium {kris.demuynck,tom.laureys}@esat.kuleuven.ac.be http://www.esat.kuleuven.ac.be/～spch Abstract. We compare different methods for obtaining accurate speech segmentations starting from the corresponding orthography. The com- plete segmentation process can be decomposed into two basic steps. First, a phonetic transcription is automatically produced with the help of large vocabulary continuous speech recognition (LVCSR). Then, the phonetic information and the speech signal serve as input to a speech segmentation tool. We compare two automatic approaches to segmentation, based on the Viterbi and the Forward-Backward algorithm respectively. Further, we develop different techniques to cope with biases between automatic and manual segmentations. Experiments were performed to evaluate the generation of phonetic transcriptions as well as the different speech seg- mentation methods. 1 Introduction In this paper we investigate the development of an accurate speech segmenta- tion system for the Spoken Dutch Corpus project. Speech segmentations, on phoneme (e.g. TIMIT) or word level (e.g. Switchboard, CGN), have become a standard annotation in speech corpora. Corpus users can benefit from the fact that the segmentation couples the speech signal to the other annotation layers (orthography, phonetics) by means of time stamps, thus providing easy access to audio fragments in the corpus. For the speech technologist segmentations are indispensable for the initial training of acoustic ASR models, the development of TTS systems and speech research in general. Some speech corpora only provide automatic segmentations, obviously requir- ing an accurate segmentation algorithm. In other corpora speech segmentations are checked manually. The latter case requires a high-quality automatic segmen- tation system as well, since a bet