CORTICAL REPRESENTATION OF SPEECH IN COMPLEX AUDITORY ENVIRONMENTS AND APPLICATIONS
Puvvada, Venkata Naga Krishna Chaitanya
Simon, Jonathan Z
MetadataShow full item record
Being able to attend and recognize speech or a particular sound in complex listening environments is a feat performed by humans effortlessly. The underlying neural mechanisms, however, remain unclear and cannot yet be emulated by artificial systems. Understanding the internal (cortical) representation of external acoustic world is a key step in deciphering the mechanisms of human auditory processing. Further, understanding neural representation of sound finds numerous applications in clinical research for psychiatric disorders with auditory processing deficits such as schizophrenia. In the first part of this dissertation, cortical activity from normal hearing human subjects is recorded, non-invasively, using magnetoencephalography in two different real-life listening scenarios. First, when natural speech is distorted by reverberation as well as stationary additive noise. Second, when the attended speech is degraded by the presence of multiple additional talkers in the background, simulating a cocktail party. Using natural speech affected by reverberation and noise, it was demonstrated that the auditory cortex maintains both distorted as well as distortion-free representations of speech. Additionally, we show that, while the neural representation of speech remained robust to additive noise in absence of reverberation, noise had detrimental effect in presence of reverberation, suggesting differential mechanisms of speech processing for additive and reverberation distortions. In the cocktail party paradigm, we demonstrated that primary like areas represent the external auditory world in terms of acoustics, whereas higher-order areas maintained an object based representation. Further, it was demonstrated that background speech streams were represented as an unsegregated auditory object. The results suggest that object based representation of auditory scene emerge in higher-order auditory cortices. In the second part of this dissertation, using electroencephalographic recordings from normal human subjects and patients suffering from schizophrenia, it was demonstrated, for the first time, that delta band steady state responses are more affected in schizophrenia patients compared with healthy individuals, contrary to the prevailing dominance of gamma band studies in literature. Further, the results from this study suggest that the inadequate ability to sustain neural responses in this low frequency range may play a vital role in auditory perceptual and cognitive deficit mechanisms in schizophrenia. Overall this dissertation furthers current understanding of cortical representation of speech in complex listening environments and how auditory representation of sounds is affected in psychiatric disorders involving aberrant auditory processing.