Page 57 - Fall2021

Figure 7. Four spectral profiles (see Figure 6) are shown, each at a different overall intensity. Two are “flat” spectral profiles (top right and bottom left) and two are “target-incremented” spectral profiles. (top left and bottom right).
a profile could that judgment be made. Dave showed in several experiments described in his book (Green, 1988) that trained listeners were sensitive to the relative inten- sity changes in a spectral profile.
In addition to spectral changes between two profiles, the time-domain waveform will also differ, and it could be a basis for distinguishing between profiles. To investigate this possibility, Green and Mason (1985) randomly varied the phases of the spectral components in several ways, which changes the time-domain waveforms but leaves the amplitude spectra unchanged. Phase variations made little difference in the trained listeners’ ability to discrimi- nate one profile from another. Thus, profile analysis is most likely due to differences in the stimuli’s amplitude spectra as opposed to the time-domain waveforms.
Of all the aspects of profile stimuli that Dave studied, he was clearly impressed with one general finding: that when there were many background tones that were very different in frequency from the target tone, there was a large effect on profile discrimination performance. This is in contrast to what was often found in masking and discrimination experiments, where target detection or discrimination per- formance in these cases was affected mainly by spectral components close in frequency to the target component.
The general idea is that stimuli that are close together in frequency directly interact in the biomechanical inner ear transduction of sound into neural action potentials that
flow to the brainstem via the auditory nerves. Each audi- tory nerve is tuned to a particular frequency range (each auditory nerve has a tuning curve) such that if stimuli have frequencies within that range, the nerve responds, but if components have frequencies outside the range, the nerve is unresponsive. A perceptual consequence of the tuning curve is the critical band. A critical band is a spec- tral region such that only components with frequencies within the critical band affect detection or discrimina- tion performance of a target component. Thus, the profile analysis finding that components with frequencies well outside the target’s critical band affected discrimination performance was not consistent with “traditional” criti- cal-band accounts of detection and discrimination.
A conclusion of research on profile analysis is that auditory spectral processing is not necessarily limited by critical- band processing but can be “wideband.” Although at the time of writing Profile Analysis: Auditory Intensity Dis- crimination (Green, 1988) there were only a few examples of such wideband spectral processing in auditory detec- tion and discrimination experiments, Dave did study the detection of tones of different frequencies in a noisy back- ground early in his career (Green, 1958). In this study, he concluded that a wideband approach of integrating across critical bands could account for his results. In his profile analysis research, Dave pointed out that wideband process- ing is consistent with what must be required to perceive complex sounds such as speech and music, and profile analysis provides a means of investigating wideband per- ceptual processing of real-world sounds.
Shortly after the publication of Profile Analysis: Auditory Intensity Discrimination (Green, 1988), several authors pointed out that the perceptual parsing of complex sounds is likely based on how sources produce sound, particularly when the sources produce nearly simulta- neous sounds. Bregman’s book “Auditory Scene Analysis” (1990) brought this view to the forefront. Auditory Scene Analysis describes the acoustic world as a scene of sound sources, and auditory perception involves determining the sound sources in such a scene. In Bregman’s view, to do so requires not only an ability to process sounds pro- duced by sources but also requires information gained from experience that has been stored in memory and then accessed through attentional processes. Perceiving sounds in an auditory scene often requires wideband processing, and profile analysis, along with several other
Fall 2021 • Acoustics Today 57