Einar Meister

Perception of the short vs. long phonological category in Estonian by native and non-native listeners

Abstract This paper studies the perception of Estonian duration-based phonological oppositions by native Estonians and non-native speakers with Russian-language background. The short/long category boundary was examined by varying the duration of a vowel in three contexts involving isolated vowels (V vs. VV), one-syllable nonsense words (CVC vs. CVVC), and two-syllable real words (CVCV vs. CVVCV). Since vowel duration serves to distinguish lexical minimal pairs in Estonian but not in Russian, L1 and L2 subjects are expected to employ different perceptual strategies in a short/long categorization task. In particular, location and width of category boundaries as well as consistency of categorization are likely to vary between the groups. The results showed that L2 subjects were quite successful in distinguishing the Estonian short/long categories despite the non-categorical use of the duration cue in their native language. As a rule, the L2 subjects demonstrated (1) category boundaries at longer durations, (2) larger width of category boundaries, and (3) lower consistency of responses compared to those of the L1 group. The perceptual strategies of L2 subjects might be based on the continuous auditory perception of the salient duration cue, or on the variable duration patterns associated with word stress in their L1, or on a combination of both strategies.

Durational patterns in Northern Estonian and Northern Finnish

Abstract Estonian and Finnish are closely related languages in which quantity is extensively exploited for lexical and grammatical purposes (in both consonants and vowels, independent of each other), yet with several phonological differences between the quantity systems (e.g. a ternary opposition in Estonian, a binary one in Finnish). To date, segment durations in the two languages have not been systematically compared. This paper reports a necessarily explorative experiment with two primary goals: first, to compare the phonetic realisation of quantity in the two languages in selected word structures, and second, to relate the results on accentual lengthening to the predictions of a speech timing framework ( White, 2002 ) that has been developed on the basis of observations in an essentially non-quantity language, namely English. It was observed, besides cross-language differences in the durational realisation of the three-way and two-way quantity contrasts that durationally C1, outside the quantity system in both languages, behaves differently in the two languages. It was also observed that the patterns of accentual lengthening in the two languages are highly consistent with predictions of the timing framework.

Evaluation of Automatic Speech Recognition Prototype for Estonian Language in Radiology Domain: A Pilot Study

The aim of this study was to determine the dictation error rates in finalized radiology reports generated with a new automatic speech recognition (ASR) technology prototype for the Estonian language.

Real-time mimicking of estonian speaker's mouth movements on a 3D avatar using Kinect 2

In this paper an algorithm for the real-time imitation of lips movements of an Estonian speaker on a 3D avatar is proposed. This research is focused on obtaining the RGB and depth information from a Kinect 2 camera, processing it and moving the avatar accordingly. For the more stable extraction of information about lip shape and movement, cosine function is used. The depth info helps to determine the closest speaker and stabilize the movement of avatar and tracking of lips.

SpeechDat-Like Estonian Database

A speech database project has been launched last year, which aims the collection of telephone speech from a large number of speakers for speech and speaker recognition purposes. Up to 2000 speakers are expected to participate in recordings. To achieve this target different recruitment schemes have been implemented. SpeechDat databases, especially Finnish SpeechDat, have been chosen as a prototype for the Estonian database. The paper is a progress report of the project.

Retrospective Analysis of Clinical Performance of an Estonian Speech Recognition System for Radiology: Effects of Different Acoustic and Language Models

The aim of this study was to analyze retrospectively the influence of different acoustic and language models in order to determine the most important effects to the clinical performance of an Estonian language-based non-commercial radiology-oriented automatic speech recognition (ASR) system. An ASR system was developed for Estonian language in radiology domain by utilizing open-source software components (Kaldi toolkit, Thrax). The ASR system was trained with the real radiology text reports and dictations collected during development phases. The final version of the ASR system was tested by 11 radiologists who dictated 219 reports in total, in spontaneous manner in a real clinical environment. The audio files collected in the final phase were used to measure the performance of different versions of the ASR system retrospectively. ASR system versions were evaluated by word error rate (WER) for each speaker and modality and by WER difference for the first and the last version of the ASR system. Total average WER for the final version throughout all material was improved from 18.4% of the first version (v1) to 5.8% of the last (v8) version which corresponds to relative improvement of 68.5%. WER improvement was strongly related to modality and radiologist. In summary, the performance of the final ASR system version was close to optimal, delivering similar results to all modalities and being independent on user, the complexity of the radiology reports, user experience, and speech characteristics.

A Preliminary Comparison of Estonian and Finnish Plosives

It has been impressionistically suggested that Estonian short plosives are somehow ”weaker” than Finnish short plosives. This paper reports the results of an acoustic comparison of short plosives in the two languages in segmentally highly similar words. It was observed that, both in word-initial and word-medial position, closure durations were systematically shorter in Estonian than in Finnish. In word-medial position, the Estonian /k/ had shorter burst duration than Finnish /k/. Moreover, a large proportion of the Estonian /k/ tokens were burstless, i.e. completely voiced. The short burst of the velar plosive seems to be a specific characteristic of Estonian, which sets the language apart from many other languages.

Vowel quantity and quality perception in Finnish and Estonian speakers.

Vowel quality perception in quantity languages might be expected to be unrelated to duration of the vowels since duration is used to realize phonemic duration oppositions. Our studies, both earlier reported ones and our latest project, indicate, though, that vowel‐intrinsic duration changes have a consistent effect on vowel identification in Estonian and Finnish speakers. Our latest experimental setup for investigating perceptional effects of microduration consists of two blocks, first a set of tests to identify the subject’s formant‐based categorial boundaries between vowel pairs on the close‐open axis, then the accordingly selected listening tests presenting formant‐wise ambiguous vowel exemplars produced with varying durations. Results from ten Estonian and ten Finnish speakers will be presented. In addition, the phenomenon observed will be compared with formant structure and duration perception in languages where these parameters co‐vary (e.g., English) and thus an influence of sub‐phonemic durational...