Corpus of deaf speech for acoustic and speech production research

Mendel, Lisa Lucks; Lee, Sung-Min; Pousson, Monique; Patro, Chhayakanta; McSorley, Skylar; Banerjee, Bonny; Najnin, Shamima; Kapourchali, Masoumeh Heidari

doi:10.1121/1.4994288

Cited by 12 publications

(8 citation statements)

References 5 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Other two spectral moments (spectral mean and skewness) indicated that those with highly intelligible speech yields more energy in the high-frequency. This findings were in line with Raghavendra et al, [15] in which they systematically analyzed the /s/, /ʃ/, /f/, and /θ/ of the same dataset in Mendel et al, [14] by cutting them as a short duration of fricative segments.…”

Section: Spectral Moment Analysissupporting

confidence: 89%

“…Recently, Mendel et al [14] analyzed four spectral moments (spectral mean, standard deviation, skewness, and kurtosis) for short passages spoken by three hearing impaired groups that were sorted based on perceptual intelligibility. In the results, spectral variance (standard deviation) increased with increases in intelligibility, while kurtosis decreased with increases in intelligibility, suggesting that speech that is more intelligible has a broader range of frequency distribution.…”

Section: Spectral Moment Analysismentioning

confidence: 99%

See 1 more Smart Citation

Spectral analysis of english fricatives /s/ and /ʃ/ produced by people with profound hearing loss

Lee¹

2020

Clin Arch Commun Disord

Self Cite

View full text Add to dashboard Cite

Results: Variance and WE for HI were significantly higher than those for NH group (p < 0.05). Mean and SP for /s/ were significantly greater in NH group, while Mean and SP for /ʃ/ were significantly greater in HI group (p < 0.05). This trends suggest greater Mean and SP differences between /s/ and /ʃ/ for NH group than HI group.Conclusions: Based on our finding, Variance and WE are spectral measures that are different between NH and HI individuals' spoken /s/ and /ʃ/. We assume that HI individuals likely pronounce /s/ and /ʃ/ without making clear distinction substituting by each other due probably to lack of auditory feedback or failure of motor control.

show abstract

Section: Spectral Moment Analysissupporting

confidence: 89%

Section: Spectral Moment Analysismentioning

confidence: 99%

Spectral analysis of english fricatives /s/ and /ʃ/ produced by people with profound hearing loss

Lee¹

2020

Clin Arch Commun Disord

Self Cite

View full text Add to dashboard Cite

show abstract

“…Two groups of 8 talkers each were selected from the “Corpus of deaf speech for acoustic and speech production research” database collected at the University of Memphis ( Mendel et al, 2017 ). This corpus is a pool of speech recordings digitally sampled at 44,100 Hz, spoken by NH talkers and hearing impaired (HI) talkers.…”

Section: Methodsmentioning

confidence: 99%

“…Table 2 presents the demographic details of the chosen CI talkers from the database ( Mendel et al, 2017 ). The CI talkers are aged between 16 and 77 years with an average age of 47.5 years.…”

Section: Methodsmentioning

confidence: 99%

Cortical entrainment to speech produced by cochlear implant talkers and normal-hearing talkers

Raghavendra

Lee²,

Chun³

et al. 2022

Front. Neurosci.

Self Cite

View full text Add to dashboard Cite

Cochlear implants (CIs) are commonly used to restore the ability to hear in those with severe or profound hearing loss. CIs provide the necessary auditory feedback for them to monitor and control speech production. However, the speech produced by CI users may not be fully restored to achieve similar perceived sound quality to that produced by normal-hearing talkers and this difference is easily noticeable in their daily conversation. In this study, we attempt to address this difference as perceived by normal-hearing listeners, when listening to continuous speech produced by CI talkers and normal-hearing talkers. We used a regenerative model to decode and reconstruct the speech envelope from the single-trial electroencephalogram (EEG) recorded on the scalp of the normal-hearing listeners. Bootstrap Spearman correlation between the actual speech envelope and the envelope reconstructed from the EEG was computed as a metric to quantify the difference in response to the speech produced by the two talker groups. The same listeners were asked to rate the perceived sound quality of the speech produced by the two talker groups as a behavioral sound quality assessment. The results show that both the perceived sound quality ratings and the computed metric, which can be seen as the degree of cortical entrainment to the actual speech envelope across the normal-hearing listeners, were higher in value for speech produced by normal hearing talkers than that for CI talkers. The first purpose of the study was to determine how well the envelope of speech is represented neurophysiologically via its similarity to the envelope reconstructed from EEG. The second purpose was to show how well this representation of speech for both CI and normal hearing talker groups differentiates in term of perceived sound quality.

show abstract

“…For example, the mean spectral energy of CI patients is lower than that of normal individuals, implying that they have difficulty in pronouncing high-frequency words. Patients with CI also pause (staccato) in the middle of the speech and often confuse voiceless with voiced speech [6,7]. The characteristics of CI patients' speech differ from those of standard speech.…”

Section: Introductionmentioning

confidence: 99%

An Effective Learning Method for Automatic Speech Recognition in Korean CI Patients’ Speech

et al. 2021

View full text Add to dashboard Cite

The automatic speech recognition (ASR) model usually requires a large amount of training data to provide better results compared with the ASR models trained with a small amount of training data. It is difficult to apply the ASR model to non-standard speech such as that of cochlear implant (CI) patients, owing to privacy concerns or difficulty of access. In this paper, an effective finetuning and augmentation ASR model is proposed. Experiments compare the character error rate (CER) after training the ASR model with the basic and the proposed method. The proposed method achieved a CER of 36.03% on the CI patient’s speech test dataset using only 2 h and 30 min of training data, which is a 62% improvement over the basic method.

show abstract

Corpus of deaf speech for acoustic and speech production research

Cited by 12 publications

References 5 publications

Spectral analysis of english fricatives /s/ and /ʃ/ produced by people with profound hearing loss

Spectral analysis of english fricatives /s/ and /ʃ/ produced by people with profound hearing loss

Cortical entrainment to speech produced by cochlear implant talkers and normal-hearing talkers

An Effective Learning Method for Automatic Speech Recognition in Korean CI Patients’ Speech

Contact Info

Product

Resources

About