A signal detection theoretic analysis of the Carney model of auditory processing

Abstract

Computational models of human auditory processing provide a platform on which to apply signal detection theory (SDT) directly to the ‘‘output‘‘ of various peripheral processes. Theoretical performance predicted by standard SDT analyses which do not incorporate such models often substantially exceeds measured performance. These discrepancies are commonly compensated for by adding a physiologically unspecified source of ‘‘internal noise.’’ By applying SDT to auditory models it is possible to evaluate whether the performance predicted at various stages of the auditory system provides a more accurate estimate of human performance than standard approaches which incorporate internal noise. Our previous work indicates that integrating SDT with a deterministic model of the human auditory system affords theoretical performance predictions that match experimental measurements more closely than traditional approaches. In the work presented here, SDT is integrated with Carney’s model of auditory processing [Carney, J. Acoust. Soc. Am. 93 (1993)]. This model incorporates physiologically based models of the basilar membrane and inner hair cell mechanisms, as well as a nonhomogeneous Poisson process to model nerve discharge times. The performance predicted by this model is presented, and its relationship to previous results and to modeling assumptions is discussed. [Work supported by NIH.]