In Thrust area II(B), we proposed a variety of signal processing algorithms for the detection of changes in features of acoustic signals (e.g., their spectra) reflecting mechanical faults (e.g., helicopter gear boxes or in manufacturing processes) and underwater transients. Here we shall characterize human psychophysical performance on the detection of increments and decrements of carefully controlled time- and band-limited signals in the presence of a wideband masker. Hypotheses for the mechanisms of detection are tested with physiologically based models for the early and central auditory stages. Decision variables for these psychophysiological models are based on the representation of stimulus cues, such as temporal onsets or spectral edges, in the nonlinear responses of auditory nerve fibers and subsequent structures. We plan to achieve these goals following four intermediate objectives:
(i) To develop a nonlinear filter bank model with parameters appropriate for the human auditory periphery that includes real-time changes in the bandwidths representing the frequency tuning of human auditory nerve fibers in response to arbitrary stimulus waveforms [. Carney 1993, Glasberg Moore 1990.]. This model will be compared with those in Thrust area I(A). Once in place, this auditory model will allow simulation of psychophysical experiments, such as threshold tracking procedures, that are implemented in exactly the same way for the computer as they are for human listeners. Psychophysical tests will be designed and carried out to test assumptions used in the development of the nonlinear model.
(ii) To identify cues in the filter bank response that are candidates for psychophysical performance on detection of increments/decrements [.Formby 1995.]. Possible cues are related to the spectral profiles and/or edges, temporal profiles and/or edges and overall level. Model performance on simulated psychophysical experiments based on these decision variables will be compared to data available for human listeners on tasks related to detection of increments and decrements.
(iii) To develop physiologically-based models for detection and processing of the spectral and temporal cues. Various existing models will be adapted for this task, including simple models for the convergence of auditory nerve fibers on cells in the auditory brainstem, or cross-channel-correlation and related algorithms [. Joris Carney 1991, Joris Carney 1994, Shamma 1985 spee2.] ( Thrust area I(A)), or multiscale cortical representations ( Thrust area I(C)).
(iv) To perform psychophysical studies designed to test hypotheses based on the decision variables and physiological mechanisms described above. The nature of the decisions variables will dictate the nature of the experiments performed to test those hypotheses. For example, decision variables that depend strongly upon edge-related cues should be resistant to certain manipulations that would affect performance based on profile cues, and vice versa.