Bistra Andreeva

Learn More
The hidden Markov modelling experiments presented in this paper show that consonant identification results can be improved substantially if a neural network is used to extract linguistically relevant information from the acoustic signal before applying hidden Markov modelling. The neural network – or in this case a combination of two Kohonen networks –(More)
This paper describes three cross-language ASR experiments which use hidden Markov modelling. The first one shows that consonant identification improves when vowel transitions are used. In particular, the consonants' place of articulation is identified better, because the vowel transitions contain formant trajectories which depend on the consonant's place of(More)
The reduction of the six Bulgarian vowels (i, , a, , , u) to a fouror (in some dialects) three-vowel subsystem (i, ( ), , u) in unstressed syllables is generally accepted. But a number of studies disagree on the exact nature of the reduction process. Claims differ as to whether or not /a/ merges phonetically with / /, and / / with /u/, or whether the(More)
The present paper is a first attempt to use the Fujisaki model to parameterize the F0 contours of utterances containing Accent 1 and Accent 2 tonal accents in Norwegian in different focus conditions. Differences in timing and amplitude of the accent commands are found, largely corresponding to descriptions in the literature. This shows that the model can be(More)
In a production study, Bulgarian, English and German verses with regular poetic metrical metres of different types and elicited prose utterances with varied accentual patterns are produced in textual and iterative (dada) form and measured at syllable level according to the pairwise variability index (PVI) principle. Systematic differences in PVI values show(More)
This article presents preliminary results indicating that speakers have a different pitch range when they speak a foreign language compared to the pitch variation that occurs when they speak their native language. To this end, a learner corpus with French and German speakers was analyzed. Results suggest that speakers indeed produce a smaller pitch range in(More)
1. Introduction It can be taken for granted that the first language (L1) influences the target language to be learned (L2) on all linguistic levels including the lexicon, morphosyntax, pragmatics and, most pertinent to the proposed corpus, sound structure and its phonetic implementation (e.g. Flege & Davidian 1984, Flege 1995). A typical example of L1-L2(More)
By mapping acoustic parameters onto phonetic features, it is possible to explicitly address the linguistic information in the signal. For the experiments presented in this paper, we mapped cepstral parameters onto two sets of phonetic features, one based on the IPA chart and the other on SPE. As a result, the phoneme identification rates in a hidden Markov(More)
This study presents the results of a large-scale comparison of various measures of pitch range and pitch variation in two Slavic (Bulgarian and Polish) and two Germanic (German and British English) languages. The productions of twenty-two speakers per language (eleven male and eleven female) in two different tasks (read passages and number sets) are(More)
We present the design of a corpus of native and non-native speech for the language pair French-German, with a special emphasis on phonetic and prosodic aspects. To our knowledge there is no suitable corpus, in terms of size and coverage, currently available for the target language pair. To select the target L1-L2 interference phenomena we prepare a small(More)