EasyCall Corpus: A Dysarthric Speech Dataset

Turrisi, Rosanna; Braccia, Arianna; Emanuele, Marco; Giulietti, Simone; Pugliatti, Maura; Sensi, Mariachiara; Fadiga, Luciano; Badino, Leonardo

doi:10.21437/interspeech.2021-549

Cited by 20 publications

(11 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Compared with more widely available normal speech corpora, such as Switchboard and Fisher conversational telephone speech [52] or LibriSpeech [53] containing from hundreds to thousands of hours of audio data, existing dysarthric and elderly speech corpora are much smaller in size. Similar data scarcity is also found not only among dysarthric speech datasets collected for non-English languages such as Dutch [54], Italian [55], Cantonese [56] and Korean [49], but also in elderly speech corpora.…”

Section: Introductionsupporting

confidence: 66%

Personalized Adversarial Data Augmentation for Dysarthric and Elderly Speech Recognition

Jin¹,

Deng²,

Wang³

et al. 2022

Preprint

View full text Add to dashboard Cite

Despite the rapid progress of automatic speech recognition (ASR) technologies targeting normal speech, accurate recognition of dysarthric and elderly speech remains highly challenging tasks to date. It is difficult to collect large quantities of such data for ASR system development due to the mobility issues often found among these users. To this end, data augmentation techniques play a vital role. In contrast to existing data augmentation techniques only modifying the speaking rate or overall shape of spectral contour, fine-grained spectro-temporal differences between dysarthric, elderly and normal speech are modelled using a novel set of speaker dependent (SD) generative adversarial networks (GAN) based data augmentation approaches in this paper. These flexibly allow both: a) temporal or speed perturbed normal speech spectra to be modified and closer to those of an impaired speaker when parallel speech data is available; and b) for non-parallel data, the SVD decomposed normal speech spectral basis features to be transformed into those of a target elderly speaker before being re-composed with the temporal bases to produce the augmented data for state-of-the-art TDNN and Conformer ASR system training. Experiments are conducted on four tasks: the English UASpeech and TORGO dysarthric speech corpora; the English DementiaBank Pitt and Cantonese JCCOCC MoCA elderly speech datasets. The proposed GAN based data augmentation approaches consistently outperform the baseline speed perturbation method by up to 0.91% and 3.0% absolute (9.61% and 6.4% relative) WER reduction on the TORGO and DementiaBank data respectively. Consistent performance improvements are retained after applying LHUC based speaker adaptation.

show abstract

Section: Introductionsupporting

confidence: 66%

Personalized Adversarial Data Augmentation for Dysarthric and Elderly Speech Recognition

Jin¹,

Deng²,

Wang³

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…Similar to the PC-GITA corpus, the EasyCall experiments include a training session with only control speakers and another session with control and dysarthric speakers (no speaker in the test set in the train set). The train and test split was selected based on the split from the EasyCall article [23]. The test set contains 12 dysarthric speakers (7 male, 5 female) ranging from mild to severe dysarthria.…”

Section: Methodsmentioning

confidence: 99%

“…The EasyCall corpus consists of 24 healthy (10 females, 14 males) and 31 dysarthric (11 females, 20 males) Italian speakers [23]. A range of disorders causing dysarthria includes Parkinson's Disease, Huntington's Disease, Amyotrophic Lateral Sclerosis, peripheral neuropathy, myopathic or myasthenic lesions.…”

Section: Easycallmentioning

confidence: 99%

Cross-lingual Self-Supervised Speech Representations for Improved Dysarthric Speech Recognition

Hernandez¹,

Pérez-Toro²,

Nöth³

et al. 2022

Preprint

View full text Add to dashboard Cite

State-of-the-art automatic speech recognition (ASR) systems perform well on healthy speech. However, the performance on impaired speech still remains an issue. The current study explores the usefulness of using Wav2Vec self-supervised speech representations as features for training an ASR system for dysarthric speech. Dysarthric speech recognition is particularly difficult as several aspects of speech such as articulation, prosody and phonation can be impaired. Specifically, we train an acoustic model with features extracted from Wav2Vec, Hubert, and the cross-lingual XLSR model. Results suggest that speech representations pretrained on large unlabelled data can improve word error rate (WER) performance. In particular, features from the multilingual model led to lower WERs than filterbanks (Fbank) or models trained on a single language. Improvements were observed in English speakers with cerebral palsy caused dysarthria (UASpeech corpus), Spanish speakers with Parkinsonian dysarthria (PC-GITA corpus) and Italian speakers with paralysis-based dysarthria (EasyCall corpus). Compared to using Fbank features, XLSR-based features reduced WERs by 6.8%, 22.0%, and 7.0% for the UASpeech, PC-GITA, and EasyCall corpus, respectively.

show abstract

“…For example, the PRAUTOCAL corpus of speech in Down syndrome (Escudero-Mancebo et al, 2021) contains sentences obtained from speakers with Down syndrome during a video game and qualitatively assessed by several experts. The EasyCall is a dysarthric speech dataset of commands most likely to be used in a voice-controlled contact application (Turrisi et al, 2021). In the Atlanta Motor Speech Disorders Corpus (Laures-Gore et al, 2016) the data include single vowels, single words, sentences and discourse passages from people with motor speech disorders who speak different dialects of English.…”

Section: Oral Speech Corpora In Clinical Linguistics: An Overviewmentioning

confidence: 99%

Discourse Diversity Database (3D) for Clinical Linguistics Research: Design, Development, and Analysis

Mariya

Antonova

Nelubina

et al. 2023

Bakhtiniana, Rev. Estud. Discurso

View full text Add to dashboard Cite

Discourse Diversity Database (3D) is a corpus designed for clinical linguistics research. It consists of oral speech samples of three different genres: picture-elicited narratives, personal stories, and picture-based instructions. The sub-sections of 3D include recordings by Russian speakers from three independent groups: people with brain tumors before and after tumor removal, people with schizophrenia, and neurologically healthy individuals. This article is devoted to the description of the data collection, the annotation scheme, and the specific characteristics of each sub-section of the corpus.

show abstract

EasyCall Corpus: A Dysarthric Speech Dataset

Cited by 20 publications

References 7 publications

Personalized Adversarial Data Augmentation for Dysarthric and Elderly Speech Recognition

Personalized Adversarial Data Augmentation for Dysarthric and Elderly Speech Recognition

Cross-lingual Self-Supervised Speech Representations for Improved Dysarthric Speech Recognition

Discourse Diversity Database (3D) for Clinical Linguistics Research: Design, Development, and Analysis

Contact Info

Product

Resources

About