Neural tracking of the fundamental frequency in natural speech

Hearing loss is a common health issue that, if left untreated, may have far-reaching consequences on education, social functioning and mental well-being. In order to select the appropriate rehabilitation strategy (hearing aids, middle ear surgery, etc.), the hearing loss needs to be carefully evaluated. Various hearing evaluation techniques are available for use in clinic. Most of them are behavioural tests that depend on feedback from the participant about presented sounds, but this feedback can be unreliable (e.g. malingering) or challenging to obtain (e.g. young children). For this reason, among others, there is growing interest for objective measures that evaluate hearing based on bodily signals without requiring active participation of the person being tested. Electro-encephalography (EEG) is a bodily signals that reflects the electrical activity that originates from the nervous system. In this thesis, we considered several types of EEG responses for objective hearing evaluation and took steps to facilitate their implementation as objective measures to evaluate hearing in both research and clinical practice. In the first study, we investigated the effect of modulation envelope shape on the auditory steady-state response (ASSR). We found that the best-performing custom envelopes provided responses that were several decibels larger than the commonly used sinusoidal envelope shape. In the second study, we explored the effect of stimulus type, envelope frequency and envelope frequency contour on the envelope following response (EFR). One of the main findings was that responses evoked by artificial vowels were on average 4-6 decibels larger than responses evoked by modulated tones, natural vowels or syllables. Understanding these stimulus-response relations is beneficial because it provides insight in the underlying response mechanisms, but also because it allows to select the most optimal evoking stimulus for each application. To obtain ASSRs and EFRs with sufficient signal-to-noise ratio, hundreds of repetitions of the response need to be averaged. To study auditory processing in a way that is representative for daily life, it is important to use a natural meaningful stimulus without repetition. A novel method based on linear modelling, named f0-tracking, allows to analyse EFR-like for continuous speech stimuli relevant to daily life (e.g stories, audiobooks, podcasts etc.). F0-tracking investigates the neural coding of the fundamental frequency of the voice (f0) in the brain during continuous speech presentation. In the third study we employed this method to study the f0-tracking response for continuous speech by four different speakers. We found that the properties of the f0-tracking response strongly depended on the f0 and the rate-of-f0-change of the narrator's voice. In a fourth study, we improved the f0-tracking technique by modelling the expected neural response of the auditory system. The new technique nearly doubled the obtained response strengths. Stronger responses are more easily measured in noisy environments, which facilitates clinical implementation. Finally, in a 5th study we studied how the optimized f0-tracking technique is affected by important subject-related factors like age and hearing loss. This type of characterization is essential for clinical implementation. The results indicated a response decrease with advancing age, but a response enhancement with increasing severity of hearing loss. This effect was likely driven by a cortical compensation mechanism for hearing loss.

Publication year:2021

Accessibility:Open