The Neural Response at the Fundamental Frequency of Speech is Modulated by Word-level Acoustic and Linguistic Information DOI Creative Commons
Mikolaj Kegler, Hugo Weissbart, Tobias Reichenbach

et al.

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2022, Volume and Issue: unknown

Published: April 8, 2022

ABSTRACT Spoken language comprehension requires rapid and continuous integration of information, from lower-level acoustic to higher-level linguistic features. Much this processing occurs in the cerebral cortex. Its neural activity exhibits, for instance, correlates predictive processing, emerging at delays a few hundred milliseconds. However, auditory pathways are also characterized by extensive feedback loops cortical areas ones as well subcortical structures. Early can therefore be influenced cognitive processes, but it remains unclear whether such contributes processing. Here, we investigated early speech-evoked that emerges fundamental frequency. We analyzed EEG recordings obtained when subjects listened story read single speaker. identified response tracking speaker’s frequency occurred delay 11 ms, while another elicited high-frequency modulation envelope higher harmonics exhibited larger magnitude longer latency about 18 ms. Subsequently, determined these responses each individual word story. then quantified context-independent used model compute context-dependent surprisal precision. The represented how predictable is, given previous context, precision reflected confidence predicting next past context. found word-level were predominantly features: average its variability. Amongst features, only showed weak significant modulation. Our results show is already suggesting top-down response.

Language: Английский

Predictors for estimating subcortical EEG responses to continuous speech DOI Creative Commons
Joshua P. Kulasingham, Florine L. Bachmann, Kasper Eskelund

et al.

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2023, Volume and Issue: unknown

Published: March 1, 2023

Abstract Perception of sounds and speech involves structures in the auditory brainstem that rapidly process ongoing stimuli. The role these processing can be investigated by measuring their electrical activity using scalp-mounted electrodes. However, typical analysis methods involve averaging neural responses to many short repetitive stimuli bear little relevance daily listening environments. Recently, subcortical more ecologically relevant continuous were detected linear encoding models. These estimate temporal response function (TRF), which is a regression model minimises error between measured signal predictor derived from stimulus. Using predictors highly non-linear peripheral system may improve TRF estimation accuracy peak detection. Here, we compare both simple complex models for estimating TRFs on electroencephalography (EEG) data 24 participants speech. We also discuss length required with clear peaks. Interestingly, resulted similar those estimated models, much faster compute. This work paves way efficient modelling detection speech, lead improved diagnosis metrics hearing impairment assistive technology.

Language: Английский

Citations

3
Cortical representation of speech temporal information through high gamma-band activity and its temporal modulation DOI Creative Commons
Shunsuke Tamura, Yoji Hirano

Cerebral Cortex, Journal Year: 2023, Volume and Issue: 33(13), P. 8773 - 8782

Published: May 11, 2023

Abstract Numerous studies have investigated low-frequency (theta-band) and high-frequency (gamma-band) neural activities that are phase-locked to temporal structures, including the envelope fine structure (TFS) of speech signals. Nonetheless, mechanisms underlying interaction between TFS processing remain elusive. Here we examined high gamma-band their amplitude modulations while listening monotone (MS) with a fundamental frequency (F0) 80 Hz non-speech sounds similar characteristics MS, namely an amplitude-modulated click train (AMC). Additionally, utilized noise-vocoded (NVS) evaluate impact eliminating from MS on activity. We observed discernible activity at same as F0 AMC (80 Hz). Furthermore, source localization analysis revealed exhibited left hemisphere dominance in both conditions. Finally, amplitude-modulation rate stimulus (5 Hz), though such modulation was not NVS. Our findings indicate is pivotal information processing, regardless nature being or non-speech.

Language: Английский

Citations

2
Extending Subcortical EEG Responses to Continuous Speech to the Sound-Field DOI Creative Commons
Florine L. Bachmann, Joshua P. Kulasingham,

Kasper Eskleund

et al.

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2023, Volume and Issue: unknown

Published: Nov. 13, 2023

Abstract The auditory brainstem response (ABR) is a valuable clinical tool for objective hearing assessment, which conventionally detected by averaging neural responses to thousands of short stimuli. Progressing beyond these unnatural stimuli, continuous speech presented via earphones have been recently using linear temporal functions (TRFs). Here, we extend earlier studies measuring subcortical in the sound-field, and assess amount data needed estimate TRFs. Electroencephalography (EEG) was recorded from 24 normal participants while they listened clicks stories loudspeakers. Subcortical TRFs were computed after accounting non-linear processing periphery either stimulus rectification or an nerve model. Our results demonstrated that could be reliably measured sound-field. estimated models outperformed simple rectification, 16 minutes sufficient all show clear wave V peaks both sound-field highly consistent earphone conditions, with click ABRs. However, required slightly more (16 minutes) achieve compared (12 minutes), possibly due effects room acoustics. By investigating this study lays groundwork bringing assessment closer real-life may lead improved evaluations smart technologies.

Language: Английский

Citations

2
The effect of phoneme-based auditory training on speech intelligibility in hearing-aid users DOI Creative Commons
Aleksandra Koprowska, Jérémy Marozeau, Torsten Dau

et al.

International Journal of Audiology, Journal Year: 2022, Volume and Issue: 62(11), P. 1048 - 1058

Published: Oct. 27, 2022

Hearing loss commonly causes difficulties in understanding speech the presence of background noise. The benefits hearing-aids terms intelligibility challenging listening scenarios remain limited. present study investigated if phoneme-in-noise discrimination training improves phoneme identification and sentence noise hearing-aid users.Two groups participants received either a two-week program or control intervention. Three categories were trained: onset consonants (C1), vowels (V) post-vowel (C2) C1-V-C2-/i/ logatomes from Danish nonsense word corpus (DANOK). Phoneme test hearing (HINT) administered before after respective interventions and, for group only, three months.Twenty 63-to-79 years old individuals with mild-to-moderate sensorineural at least one year experience using hearing-aids.The provided an improvement scores consonants, which was retained over months. No significant performance HINT found.The demonstrates that induced robust refinement auditory perception level but provides no evidence generalisation to untrained task.

Language: Английский

Citations

3
Level Dependent Subcortical EEG Responses to Continuous Speech DOI Creative Commons
Joshua P. Kulasingham, Hamish Innes-Brown, Martin Enqvist

et al.

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: April 2, 2024

Abstract The auditory brainstem response (ABR) is a measure of subcortical activity in to stimuli. wave V peak the ABR depends on stimulus intensity level, and has been widely used for clinical hearing assessment. Conventional methods estimate average electroencephalography (EEG) responses short unnatural stimuli such as clicks. Recent work moved towards more ecologically relevant continuous speech using linear deconvolution models called Temporal Response Functions (TRFs). Investigating whether TRF waveform changes with crucial step use natural assessments involving responses. Here, we develop level-dependent TRFs EEG data collected from 21 participants listening presented at 4 different levels. We find that can be detected almost all participants, are consistent click-ABR V. also investigate most suitable peripheral model generate predictors simple gammatone filterbanks perform best. Additionally, around 6 minutes may sufficient detecting effects peaks above noise floor segments higher intensity. Finally, show proof-of-concept level dependent even inherent fluctuations speech. Visual abstract Significance statement Subcortical sound depend provide window into early human pathway. However, current detect transient clicks or chirps. stimuli, which several advantages over Critically, patterns an individual directly comparable those seen conventional click Our lays foundation future applications assessment assistive technology.

Language: Английский

Citations

0
The Neural Response at the Fundamental Frequency of Speech is Modulated by Word-level Acoustic and Linguistic Information DOI Creative Commons
Mikolaj Kegler, Hugo Weissbart, Tobias Reichenbach

et al.

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2022, Volume and Issue: unknown

Published: April 8, 2022

ABSTRACT Spoken language comprehension requires rapid and continuous integration of information, from lower-level acoustic to higher-level linguistic features. Much this processing occurs in the cerebral cortex. Its neural activity exhibits, for instance, correlates predictive processing, emerging at delays a few hundred milliseconds. However, auditory pathways are also characterized by extensive feedback loops cortical areas ones as well subcortical structures. Early can therefore be influenced cognitive processes, but it remains unclear whether such contributes processing. Here, we investigated early speech-evoked that emerges fundamental frequency. We analyzed EEG recordings obtained when subjects listened story read single speaker. identified response tracking speaker’s frequency occurred delay 11 ms, while another elicited high-frequency modulation envelope higher harmonics exhibited larger magnitude longer latency about 18 ms. Subsequently, determined these responses each individual word story. then quantified context-independent used model compute context-dependent surprisal precision. The represented how predictable is, given previous context, precision reflected confidence predicting next past context. found word-level were predominantly features: average its variability. Amongst features, only showed weak significant modulation. Our results show is already suggesting top-down response.

Language: Английский

Citations

2