The long-term objective of this work is to reach an understanding of the auditory system that will allow modeling of the perception of tones, noise, and speech in listeners with normal hearing and hearing losses. The overarching hypothesis to be tested in this project is that cochlear hearing losses alter auditory processing differently at different frequencies and that alterations at a specific frequency will affect only the speech information transmitted by a band of speech around that frequency. To this end, listeners with normal hearing and with hearing losses of primary cochlear origin will be tested to identify auditory- processing deficits in different spectral regimens. These deficits will be related to the listeners' abilities to understand speech in noise by applying the new Speech Recognition-Sensitivity. Contrary to traditional Articulation-Index models, the SRS model can accurately predict the intelligibility of speech in disjoint spectral bands, making it possible to model 'dead' regions in cochlear losses. It predicts speech intelligibility by, among other things, integrating an importance function the Congruence-Index Density Function. CIDF across the bandwidth of the speech.
The specific aims are to test four hypotheses: (1) Speech - recognition performance can be predicted from the CIDF, which depends on frequency-specific psychoacoustic abilities. (2) The CIDF for normal listeners is similar to the band-importance function for nonsense syllables in the Speech Intelligibility Index. (3) Conditional-on-Single- Stimulus (COSS) functions obtained by varying the signal-to-noise ratio randomly in distinct frequency bands of speech do not reflect the contribution that each band makes to the intelligibility. Rather, as the SRS model can predict recognition performance for different speech materials by varying a single parameter that reflects the linguistic entropy of the speech. To test these hypotheses, the CIDF obtained with a new COSS techniques will be compared to data for a targeted set of frequency-specific psychoacoustical tests for temporal, frequency, and intensity processing. The SRS-model predictions will be compared to measured speech-recognition scores. All the experiments are motivated by model predictions or specific hypotheses about psychoacoustic abilities that are likely to be important for speech recognition. Regardless of the success of the model, are data are highly likely to reveal auditory processes that are important for recognition of speech and will add significantly to our understanding of auditory perception and speech recognition in listeners with normal hearing and with hearing losses.
Showing the most recent 10 out of 24 publications
