CRCNS: Predictive Models of Human Speech Recognition in Natural Environmental Noise - The human ability to carry a conversation in natural auditory scenes with competing talkers and environmental noise is a critical cognitive skill. Although normal human listeners effortlessly perform this task, for instance in a crowded restaurant, it challenges the elderly and individuals with hearing loss since even moderate levels of environmental noise can severely compromise their speech recognition abilities. Understanding the neural mechanisms that underlie speech recognition in real sources of environmental noise is thus highly clinically relevant. Different types of background noise vary in acoustic structure and complexity and can interfere with speech perception in fundamentally different ways. Yet, how the acoustic structure of the background interferes with comprehension is not well understood. This study leverages recent advances in modeling sound textures , such as wind , rain, street noise or speech babble, using a biologically grounded generative auditory model to study human speech recognition (HSR)in the presence of natural environmental noises. Texture synthesis will be used to manipulate and control the background sound complexity and its influence on HSR behavior, with a wide range of natural acoustic backgrounds and variants with perturbed statistical sound structure. The study includes (Aim 1) behavioral testing to assess which elements of acoustic structure in natural backgrounds interfere with phonetic, lexical, and segmental structure in speech during HSR, (Aim 2) human EEG studies to map out how the different stages of the auditory system, from brainstem to cortex, cope with natural environment noise during speech comprehension. Finally, the study will (Aim 3) develop biologically inspired computational models of HSR that are tolerant to background environmental noise and which account for human HSR behavior and electrophysiology. This interdisciplinary approach will contribute towards a comprehensive theory of HSR in real-world natural acoustic environments. The combined use of human behavior, electrophysiology, and computational models, which are tested using identical acoustic paradigms, will identify neural mechanisms and behavioral strategies for HSR and allow us to link behavior, cognition, perception, and neurobiology. Collectively, the findings could lead to diagnostic tools that can identify auditory deficits for speech perception, which are not easily identified in clinical settings, and new biologically inspired auditory prosthetics that can operate effectively in a wide range of real-world settings. RELEVANCE (See instructions): Human speech recognition is remarkably resilient to competing environmental noise, such as in a crowded restaurant, yet this ability is severely compromised for the elderly and populations with hearing loss. This study proposes to stud y behavioral, neural, and computational mechanisms that humans use to segregate and recognize speech in natural acoustic environments. This knowledge is critical to identify neural mechanisms that endow normal hearing abilities and to develop future interventions to diagnose and treat hearing loss. P ROJ EC / P E R FO R M AN C E SI T E(S) (if ad ditional space is need ed , use
