Development of Large Vocabulary Continuous Speech Recognition Using Phonetically Structured Speech Corpus

This paper presents the results of acoustic modeling used in a Large Vocabulary Continuous Speech Recognition (LVCSR) system designed with the use of a phonetically controlled large vocabulary corpus. Evaluation experiments showed that relatively good speech recognition results may be obtained with adequate training material, taking into account: a) the presence of lexical stress; b) speech styles (a variety of segmental and prosodic structures, various degree of spontaneity of speech, various pronunciation variants and dialects); c) the influence of the sound level and environment noise. Moreover, the article includes information about the speech corpus structure and also an outline of the design of the speech recognition system.