Skip to the main content

Original scientific paper

Croatian Large Vocabulary Automatic Speech Recognition

Sanda Martinčić-Ipšić orcid id orcid.org/0000-0002-1900-5333 ; University of Rijeka, Department of Informatics, Omladinska 14, 51000, Rijeka, Croatia
Miran Pobar orcid id orcid.org/0000-0001-5604-2128 ; University of Rijeka, Department of Informatics, Omladinska 14, 51000, Rijeka, Croatia
Ivo Ipšić ; University of Rijeka, Faculty of Engineering, Vukovarska 58, 51000, Rijeka, Croatia


Full text: english pdf 1.903 Kb

page 147-157

downloads: 1.057

cite


Abstract

This paper presents procedures used for development of a Croatian large vocabulary automatic speech recognition system (LVASR). The proposed acoustic model is based on context-dependent triphone hidden Markov models and Croatian phonetic rules. Different acoustic and language models, developed using a large collection of Croatian speech, are discussed and compared. The paper proposes the best feature vectors and acoustic modeling procedures using which lowest word error rates for Croatian speech are achieved. In addition, Croatian language modeling procedures are evaluated and adopted for speaker independent spontaneous speech recognition. Presented experiments and results show that the proposed approach for automatic speech recognition using context-dependent acoustic modeling based on Croatian phonetic rules and a parameter tying procedure can be used for efficient Croatian large vocabulary speech recognition with word error rates below 5%.

Keywords

Acoustic modeling; Automatic speech recognition; Context-dependent acoustic units; Language modeling

Hrčak ID:

71298

URI

https://hrcak.srce.hr/71298

Publication date:

22.7.2011.

Article data in other languages: croatian

Visits: 2.553 *