Human auditory spectrotemporal processing in noise and its relationship to speech perception

Abstract

[EN] The auditory system adapts to the varying acoustic environment to ensure a reliable and efficient representation of perceptually relevant sound features. Physiological mechanisms such as the medial olivocochlear reflex (MOCR), the middle-ear muscle reflex (MEMR) or neural synaptic adaptation can be activated by background noise, and may change the neural encoding of sound over time. Therefore, auditory frequency selectivity, modulation sensitivity, and speech perception may change dynamically in noisy settings. The overall aim of this thesis was to investigate potential noise-induced changes in frequency selectivity and modulation sensitivity, and their possible relationship with adaptation to noise in word recognition. The thesis comprises three studies. The first study explored whether the stimulus duration affects estimates of human cochlear tuning at low and high frequencies. Cochlear tuning was estimated using a forward-masking, notched-noise (NN) method. Masker levels at masking threshold were measured for masker durations of 30 and 400 ms. The hypothesis was that tuning estimates may be different when the maskers are long enough to activate the MEMR and/or the MOCR. Masker duration had a negligible effect on tuning at 4 kHz. In contrast, the use of short masker produced broader tuning estimates at 500 Hz. The second study investigated the effect of noise precursors on human psychoacoustical tuning curves (PTCs) at 500 Hz and 4 kHz. The stimuli used to measure the PTCs were too short (30- ms maskers and 10-ms probes) to activate the MEMR or the MOCR by themselves. The noise precursors (300 ms broadband noise at 60 dB SPL) were designed to activate the MOCR with minimal activation of MEMR. Contralateral precursors had a negligible effect on the PTCs. Ipsilateral and bilateral noise precursors sharpened the PTCs at 500 Hz but broadened the PTCs at 4 kHz. An existing physiologically inspired computer model of forward masking with efferent control was used to simulate and interpret the results. The simulated PTCs were broadly consistent with the experimental PTCs and suggested that, while seemingly different, the pattern of results at the two test frequencies is consistent with the precursors inhibiting the gain of the cochlear amplifier by activation of the MOCR. The third study investigated if adaptation to noise occurs in spectral (SM), temporal (TM), and spectrotemporal modulation (STM) detection as well as in speech recognition. Vocoded-word recognition, TM and SM sensitivity in noise improved when the word or the modulated signal were delayed 800 ms from the noise onset. In contrast, this improvement did not occur in natural-word recognition or STM detection. Combined, these findings suggest that adaptation to noise in speech recognition is unlikely mediated by improvements in the encoding of STM. Together, the findings show that human auditory frequency selectivity and the sensitivity to spectral and temporal modulations can change in the presence of noise. This has implications for both understanding human auditory frequency selectivity and designing optimal methods and stimuli to estimate it. Additionally, these results highlight the importance of incorporating noise adaptation mechanisms into models of the auditory periphery to accurately simulate auditory spectrotemporal processing in noisy environments. Keywords: adaptation to noise, dynamic range adaptation, auditory frequency selectivity, cochlear tuning, medial olivocochlear reflex, middle-ear muscle reflex, speech-in-noise perception, spectrotemporal modulation perception.