Phoneme Boundary Detection Using Learnable Segmental Features

Kreuk, Felix; Sheena, Yaniv; Keshet, Joseph; Adi, Yossi

doi:10.1109/icassp40776.2020.9053053

Cited by 17 publications

(9 citation statements)

References 22 publications

(33 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For phone segmentation on TIMIT, we simply hypothesize a boundary when a code changes. Our results do not rely on peak detection [24,22] or any segmentation algorithms [26]. We find that fewer codes leads to fewer transitions, showing good results on phone segmentation.…”

Section: Analysis Of Learned Codesmentioning

confidence: 70%

See 1 more Smart Citation

Learning Dependencies of Discrete Speech Representations with Neural Hidden Markov Models

Yeh¹,

Tang²

2022

Preprint

View full text Add to dashboard Cite

While discrete latent variable models have had great success in self-supervised learning, most models assume that frames are independent. Due to the segmental nature of phonemes in speech perception, modeling dependencies among latent variables at the frame level can potentially improve the learned representations on phonetic-related tasks. In this work, we assume Markovian dependencies among latent variables, and propose to learn speech representations with neural hidden Markov models. Our general framework allows us to compare to self-supervised models that assume independence, while keeping the number of parameters fixed. The added dependencies improve the accessibility of phonetic information, phonetic segmentation, and the cluster purity of phones, showcasing the benefit of the assumed dependencies.

show abstract

Section: Analysis Of Learned Codesmentioning

confidence: 70%

“…Following [5], we also evaluate phone cluster purity on WSJ si284. For phone segmentation on TIMIT, we do not follow the setting in other studies [24,22]. The published results themselves are not entirely comparable due to subtle differences, such as data set split and preprocessing protocols.…”

Section: Methodsmentioning

confidence: 99%

Learning Dependencies of Discrete Speech Representations with Neural Hidden Markov Models

Yeh¹,

Tang²

2022

Preprint

View full text Add to dashboard Cite

show abstract

“…Although the simple DGP does not capture all dynamics of the speech signal, 7 out of 9 phoneme boundaries were correctly identified, with a time tolerance of 20 ms. A baseline detector that predicts segment boundaries from a uniform distribution was as good or better only in 69 out of 10000 runs (< 1%). This minimal experiment suggests that relaxed segmented models, when combined with more powerful DGPs, may be useful for discrete representation learning [43,46,14], in particular for learning segmental embeddings [27,48,9,29]. We consider this a fruitful direction for future work.…”

Section: Phoneme Segmentationmentioning

confidence: 98%

Differentiable Segmentation of Sequences

Scharwächter¹,

Lennartz²,

Müller³

2020

Preprint

View full text Add to dashboard Cite

Segmented models are widely used to describe non-stationary sequential data with discrete change points. Their estimation usually requires solving a mixed discretecontinuous optimization problem, where the segmentation is the discrete part and all other model parameters are continuous. A number of estimation algorithms have been developed that are highly specialized for their specific model assumptions. The dependence on non-standard algorithms makes it hard to integrate segmented models in state-of-the-art deep learning architectures that critically depend on gradient-based optimization techniques. In this work, we formulate a relaxed variant of segmented models that enables joint estimation of all model parameters, including the segmentation, with gradient descent. We build on recent advances in learning continuous warping functions and propose a novel family of warping functions based on the two-sided power (TSP) distribution. TSP-based warping functions are differentiable, have simple closed-form expressions, and can represent segmentation functions exactly. Our formulation includes the important class of segmented generalized linear models as a special case, which makes it highly versatile. We use our approach to model the spread of COVID-19 by segmented Poisson regression, perform logistic regression on Fashion-MNIST with artificial concept drift, and demonstrate its capacities for phoneme segmentation.Preprint. Under review.

show abstract

“…When the orthographic or phonetic information is not given (and the speech recognition is not employed), we speak about unsupervised speech segmentation 37,38 or phone/phoneme boundary detection. 39,40 This task usually considers phones/phonemes as segments with nearly invariable acoustic features, whereas the boundaries between them are points where these features are significantly changed. 41 The resulting segmentation may contain artificial phone-like units that may not correspond to actual phones.…”

Section: Related Workmentioning

confidence: 99%

“…The model can be used directly or after adaptation to the new language. 56,57 The cross-lingual modeling is also important for under-resourced languages, for example, Vietnamese 58 or Hebrew, 40 where a cross-lingual system often performs better than a simple monolingual one.…”

Section: Related Workmentioning

confidence: 99%

Using LSTM neural networks for cross‐lingual phonetic speech segmentation with an iterative correction procedure

Hanzlíček,

Matoušek,

Vít

2023

Computational Intelligence

View full text Add to dashboard Cite

This article describes experiments on speech segmentation using long short‐term memory recurrent neural networks. The main part of the paper deals with multi‐lingual and cross‐lingual segmentation, that is, it is performed on a language different from the one on which the model was trained. The experimental data involves large Czech, English, German, and Russian speech corpora designated for speech synthesis. For optimal multi‐lingual modeling, a compact phonetic alphabet was proposed by sharing and clustering phones of particular languages. Many experiments were performed exploring various experimental conditions and data combinations. We proposed a simple procedure that iteratively adapts the inaccurate default model to the new voice/language. The segmentation accuracy was evaluated by comparison with reference segmentation created by a well‐tuned hidden Markov model‐based framework with additional manual corrections. The resulting segmentation was also employed in a unit selection text‐to‐speech system. The generated speech quality was compared with the reference segmentation by a preference listening test.

show abstract

Phoneme Boundary Detection Using Learnable Segmental Features

Cited by 17 publications

References 22 publications

Learning Dependencies of Discrete Speech Representations with Neural Hidden Markov Models

Learning Dependencies of Discrete Speech Representations with Neural Hidden Markov Models

Differentiable Segmentation of Sequences

Using LSTM neural networks for cross‐lingual phonetic speech segmentation with an iterative correction procedure

Contact Info

Product

Resources

About