| In this thesis the representation of speech sounds in the auditory system is modeled, in order to determine how the underlying mechanisms affect speech perception, and more specifically to understand the staged transformations at different levels of the auditory system that result in the ultimate percept of a linguistic message. The initial processing step models the transformation of acoustic signals into auditory-nerve (AN) representations, and the second stage models the central auditory system, particularly at the level of the auditory cortex. The ultimate goal is to predict human speech recognition performance for both normal-hearing and hearing impaired listeners.; A computational model of the auditory periphery has been developed to simulate normal and impaired AN fiber responses in cats. The model responses match physiological data over a wider dynamic range than previous auditory models. This is achieved by providing two modes of basilar membrane (BM) excitation to the inner hair cell (IHC) rather than one, which is motivated by recent physiological observations that support multiple resonances in the BM vibrations. The two modes are generated by two parallel filters, component 1 (C1) and component 2 (C2), which correspond to the active and passive modes of basilar membrane vibration, respectively. The outputs of the two filters are subsequently transduced by two separate functions, added together, and then low-pass filtered by the IHC membrane, which is followed by the IHC-AN synapse and discharge generator. The inclusion of separate transduction mechanisms in the IHC appears consistent with two important physiological observations: first, C1 response is highly affected by cochlear impairment, while C2 is resistant to trauma, and second, all components of a multi-component stimulus undergo C1/C2 transition simultaneously irrespective of their individual levels. The C1 filter is a narrow-band, chirp filter with the gain and bandwidth controlled by a nonlinear feed-forward control path. This filter is responsible for low and moderate level responses. A linear, static and broadly tuned C2 filter followed by a nonlinear, inverted and non-rectifying C2 transduction function is critical for producing transition region and high-level effects. Consistent with Kiang's two-factor cancelation hypothesis, the interaction between the two paths produces effects such as the C1/C2 transition and peak splitting in the period histogram. The model responses are consistent with a wide range of physiological data from both normal and impaired ears for stimuli (such as pure tone, two-tone, broadband noise and speech-like stimulus) presented at levels spanning the dynamic range of hearing.; A fall-off in speech intelligibility at higher-than-normal presentation levels has been observed for listeners with and without hearing loss. Speech intelligibility predictors based on the acoustic signal properties cannot directly account for the effects of presentation level or hearing impairment. This motivates the development of a model-based speech intelligibility metric that can directly address these deleterious effects on human speech recognition performance. The idea used in this thesis is extracted from a recently developed speech intelligibility predictor, the spectro-temporal modulation index (STMI), that is based on a model of how the auditory cortex analyzes the joint spectro-temporal modulations present in speech. By employing the physiologically-accurate model of the auditory-periphery developed earlier in this thesis, this model-based speech intelligibility metric can qualitatively account for the degradations on human speech recognition performance due to suprathreshold nonlinearities or cochlear impairment, providing a link between the physiological and psychophysical observations. The predictive accuracy of this model-based approach provides a strong basis for realizing effective speech processing systems and to design hearing aid al... |
