Principles of Phonetic Segmentation

Skarnitzl, Radek; Machač, Pavel

doi:10.1159/000331902

Cited by 26 publications

(16 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…The audio files were manually annotated in Praat ( 2021) by trained research assistants, based on the segmentation criteria in Machač & Skarnitzl (2009), and checked by the authors. Disfluent responses (due to pauses, errors, false starts, throat clearing, etc.)…”

Section: Data Processingmentioning

confidence: 99%

Acoustic correlates of word stress and focus marking in Udmurt

Borise,

Georgieva

2024

Journal of the International Phonetic Association

View full text Add to dashboard Cite

In this paper, we investigate the prosodic realization of stress and focus in Udmurt (Uralic, Permic). According to the literature, Udmurt has fixed final stress, but also has several sets of morphosyntactic exceptions with initial stress. We report the results of two production studies. The first one targets nominals with final stress, and the second one investigates the stress properties of minimal pairs consisting of (i) indicative verbs (prs.3sg; final stress) and (ii) imperative verbs (imp.2sg/pl; initial stress). To control for the information-structural contexts, the test words are studied in contexts that elicit narrow focus either on the test word (‘F’ condition) or on another constituent (‘non-F’ condition). The results show that all four acoustic parameters surveyed in the paper – duration, intensity, fundamental frequency (f0), and first formant (F1) values – participate in stress marking in Udmurt. The results for focus marking vary by study and demonstrate that all cues except for intensity may be involved in focus marking. At the same time, we find wide interspeaker variation with respect to the acoustic cues marking stress and/or focus. Finally, we outline a preliminary Autosegmental-Metrical interpretation of our f0 results; a full account of Udmurt intonation awaits further research.

show abstract

Section: Data Processingmentioning

confidence: 99%

Acoustic correlates of word stress and focus marking in Udmurt

Borise,

Georgieva

2024

Journal of the International Phonetic Association

View full text Add to dashboard Cite

show abstract

“…Its output is a TextGrid containing, among other things, a tier with the phonetic boundaries, which was used for further manual boundary corrections, followed by the extraction of the fricatives with Praat (Boersma and Weenink, 2021). To define the onset and offset of the full consonant, the broadband spectrogram was considered as more important than the start of an aperiodic waveform with rising zero crossing rates, and in intervocalic fricatives, the presence of formant columns is defined as the onset and offset of the fricative [following Skarnitzl and Machac (2011)].…”

Section: Automatic and Manual Segmentationmentioning

confidence: 99%

Identifying the Russian voiceless non-palatalized fricatives /f/, /s/, and /ʃ/ from acoustic cues using machine learning

Ulrich

Allassonnière-Tang

Pellegrino

et al. 2021

The Journal of the Acoustical Society of America

View full text Add to dashboard Cite

This paper shows that machine learning techniques are very successful at classifying the Russian voiceless nonpalatalized fricatives [f], [s], and [S] using a small set of acoustic cues. From a data sample of 6320 tokens of read sentences produced by 40 participants, temporal and spectral measurements are extracted from the full sound, the noise duration, and the middle 30 ms windows. Furthermore, 13 mel-frequency cepstral coefficients (MFCCs) are computed from the middle 30 ms window. Classifiers based on single decision trees, random forests, support vector machines, and neural networks are trained and tested to distinguish between these three fricatives. The results demonstrate that, first, the three acoustic cue extraction techniques are similar in terms of classification accuracy (93% and 99%) but that the spectral measurements extracted from the full frication noise duration result in slightly better accuracy. Second, the center of gravity and the spectral spread are sufficient for the classification of [f], [s], and [S] irrespective of contextual and speaker variation. Third, MFCCs show a marginally higher predictive power over spectral cues (<2%). This suggests that both sets of measures provide sufficient information for the classification of these fricatives and their choice depends on the particular research question or application.

show abstract

“…A number of automatic instruments have been developed, most frequently based on HMMs (Hidden Markov Models) [6]. Unfortunately, these methods are at present not accurate enough for phonetic research and they need manual correction [7]. The part of speech corpus (the utterances of 6 speaker-women) was manually annotated using SFSWin program providing HTK-based labels [8] of segments.…”

Section: Speech Corpusmentioning

confidence: 99%

Advanced Recognition of Lithuanian Digit Names Using Hybrid Approach

Ratkevicius¹,

Paškauskaitė²,

Bartisiute³

2018

ElAEE

View full text Add to dashboard Cite

The paper deals with the recognition of digits and with the hybrid recognition technology. By the hybrid approach, we assume the combination of two or more different recognizers have to achieve higher recognition accuracy. Two Lithuanian recognizers using the word based and phonemebased hidden Markov models (HMM) together with the Spanish language recognizer 8.0 (Spanish-US) and Microsoft Speech Server Spanish language recognizer 9.0 (Spanish-US) were investigated. Using data mining package Weka, classification research was carried out with five different recognizer combining scenarios. The results of connecting two or three recognizers showed that the suggested method of using machine learning method to connect different recognizers greatly improved the recognition accuracy of digits speech corpus in all five cases. Manual annotation of the part of speech corpus enables to increase the recognition accuracy of Lithuanian digits names about 40 % using sub-word-based recognizer. SAMPA_LT set of phonemes is redundant for the digits recognition.

show abstract

Principles of Phonetic Segmentation

Cited by 26 publications

References 11 publications

Acoustic correlates of word stress and focus marking in Udmurt

Acoustic correlates of word stress and focus marking in Udmurt

Identifying the Russian voiceless non-palatalized fricatives /f/, /s/, and /ʃ/ from acoustic cues using machine learning

Advanced Recognition of Lithuanian Digit Names Using Hybrid Approach

Contact Info

Product

Resources

About