Spectral Shape Discrimination Research Articles

A new technique for measuring spectral shape discrimination

A new technique for measuring the ability of listeners to discriminate between sounds on the basis of spectral shape—called “auditory profile analysis”—is described. In this technique, the signal is a series of intensity increments and decrements to an equal‐level multitone reference spectrum. For any stimulus trial, the signal is either an intensity increment to the odd‐numbered tones and an intensity decrement to the even‐numbered tones, or vice‐versa. The advantage of the technique is that it reduces, by one‐half (in a two‐down, one‐up adaptive procedure), the range of the random, within‐trial rove in overall level needed to limit the usefulness of level cues. Using this technique, discrimination performance was measured for a group of normally hearing listeners for broadband, low‐pass, and high‐pass filtered conditions, and for individual hearing‐impaired listeners. The results indicated that the new technique appears to be well suited for studying auditory profile analysis in hearing‐impaired listeners where the range of sound intensities that may be presented is often quite limited. [Work supported by NIH/NIDCD.]

Spectral shape discrimination of narrow‐band spectra

Previously, it was demonstrated that spectral pitch is used to discriminate three‐component, narrow‐band spectra when tones are separated by 40 or 80 Hz, but not for a 10‐Hz separation. Here, discriminations are investigated based on the temporal envelope. The standard consists of three, equal‐intensity sinusoids, with respective frequencies 990, 1000, and 1010 Hz. Presentation level is a random variable with a range of 20 dB and mean 55 dB. Thresholds for an intensity increment of the central tone range from −10 dB to −1 dB re: the standard. Thus, performance is as good as that obtained in profile tasks for which components are separated by two or more octaves. Because performance is unaffected by (1) the addition of a modulated masker centered at 4000 Hz [W. A. Yost and S.S. Sheft, J. Acoust. Soc. Am. 85, 848–857 (1989)], or (2) randomization of each component's phase as opposed to the same, fixed phase for each, it does not appear that modulation depth is a critical variable. It is believed that discriminability is related to the spectrum of the envelope. [Work supported by AFOSR, ONR, and NIH.]

Spectral shape discrimination using three‐tone complexes

Thresholds were obtained for an increment in the intensity of a 100‐Hz, center component of a standard consisting of three equal intensity tones. For three conditions, the frequency separation between each tone was 80, 40, and 10 Hz, respectively. Overall intensity was varied over a 20‐dB range to preclude the use of absolute intensity as a cue. Unexpectedly, thresholds when tones were separated by 80 or 40 Hz differ little from thresholds when tones are separated by more than a critical hand (e.g., profile analysis). Moreover, thresholds increase by 10 dB when the digital‐to‐analog conversion rate is varied randomly between 38–42 μs/sample, providing evidence that listeners use pitch cues. If we assume that listeners use a strategy of off‐frequency listening, then the results are consistent with Feth's EWAIF model [Feth and Stover, in Auditory Processing of Complex Sounds, edited by W. A. Yost and C. S. Watson, pp. 76–86 (1987)]. When tones are separated by 10 Hz, thresholds increase by 5 dB but are unaf...

Context sensitivity in profile analysis by parallel distributed networks

Spectral shape discrimination with uncertain signal frequency was simulated using simple, three‐layer, feed‐forward, parallel networks. As in previous studies, pairs of multitonal sounds were used—one with all components at a uniform, background level and the other with a tonal signal added in‐phase to a single component [Howard et al., J. Acoust. Soc. Am. Suppl. 1 85, S37 (1989)]. Networks were trained to select the signal alternative in a 2 AFC paradigm using the generalized delta rule with error propagation and were then tested under various conditions. The signal frequency was either fixed or varied across trials, and stimulus pairs were presented either alone or with a cue that was correlated with the signal frequency. Network sensitivity was greater with the fixed signal frequency than with the varied signal frequency and this difference was reduced by the presence of a predictive cue. As in earlier experiments, these stimulation results mimic previous findings with human listeners under similar task conditions. [Work suported by ONR.]

The Psychoacoustics of Profound Hearing Impairment.

A group of profoundly hearing-impaired adults is being studied in order to determine their residual auditory capabilities. Most tests use acoustic contrasts that have relevance for the perception of speech, so that the results obtained may be used to guide the development of appropriate hearing aids. All the listeners are postlingually deafened, with losses >95 dB HL at and above 500 Hz, and least loss at low frequencies. Trends in detection thresholds, discomfort levels, dynamic ranges, intensity discrimination (both static and dynamic), frequency selectivity, spectral shape discrimination, gap detection, tone/noise discrimination, frequency discrimination and phase sensitivity are reported and discussed. Generally speaking, temporal resolving power seems to be more resistant to degradation than mechanisms of frequency selectivity. Furthermore, at least some of the differences in performance between normal and profoundly-impaired listeners may be attributable to the loss of frequency selectivity. Implications are drawn for hearing aid fitting and design, and comparisons made to the electro-auditory abilities of users of single-channel cochlear implants.

Profile analysis by parallel distributed networks

Several profile analysis (spectral shape discrimination) phenomena were simulated using simple, three‐layer, feed‐forward, parallel networks. As in previous studies with human listeners, pairs of multitonal sounds were used—one with all components at a uniform, background level and the other with a tonal signal added in‐phase to the central component. Networks were trained to select the signal alternative in a 2AFC paradigm using the generalized delta rule with error propagation and were then tested under various conditions. Background levels were varied randomly either between or within trials. In the former case, the network learned to compare energy at the signal frequency across the two sounds (intensity discrimination), whereas, in the latter case, it learned to compare the distribution of spectral energy within sounds (profile analysis). These results mimic previous findings with human listeners reported by Green and his colleagues [D. M. Green, Profile Analysis (Oxford U. P., New York, 1988)]. In follow‐up simulations, improved profile analysis have been demonstrated when the frequency range of tonal components is expanded and as the duration of the complex increases to 100 ms. [Work supported by ONR.]

Evidence for sensory-trace comparisons in spectral shape discrimination.

The ability of experienced observers to discriminate changes in the shapes of complex sound spectra was studied for three conditions. In one condition, the reference spectrum or "background" was fixed in spectral shape across each block of trials; in a second condition, the reference spectrum differed in spectral shape from trial to trial within each block of trials but was the same for the two presentations within a trial; and, in a third condition, the reference spectrum differed in spectral shape on every stimulus presentation. The variation in the reference spectrum was a random perturbation in the amplitudes of the tonal components comprising the complex sounds. The signal was an intensity increment to the center component (1000 Hz) of the complex and was present in one interval of each two-interval, two-alternative, forced-choice trial. The principal experimental manipulations were the degree of amplitude perturbation of the reference spectrum and the interval of time between the two stimulus presentations of each trial (interstimulus interval, ISI). The theory proposed by Durlach and Braida [J. Acoust. Soc. Am. 46, 372-383 (1969)] describing memory processes involved in the perception of sound intensity was used to explain the experimental results. As that theory was applied in this study, only the condition in which the stimuli were perturbed between trials showed evidence for comparisons based on "sensory traces," while the conditions in which the stimuli were fixed or were perturbed within trials were best explained by comparisons based on judgments relative to the stimulus context.

Spectral-shape discrimination. I. Results from normal-hearing listeners for stationary broadband noises.

This research is concerned with the ability of normal-hearing listeners to discriminate broadband signals on the basis of spectral shape. The signals were six broadband noises whose spectral shapes were modeled after the spectra of unvoiced fricative and plosive consonants. The difficulty of the discriminations was controlled by the addition of noise filtered to match the long-term speech spectrum. Two-interval discrimination measurements were made in which loudness cues were eliminated by randomizing (roving) the overall stimulus level between presentation intervals. Experimental results, examined as a function of intensity rove width, stimulus duration, and stimulus pair, were related to the predictions of a simple filter-bank model whose fitting parameter provides an estimate of internal noise. Most results, with the notable exception of duration effects, were predicted by the model. Estimates of internal noise in each frequency channel averaged roughly 7 dB for long-duration stimuli and 13 dB for short-duration stimuli. Results and predictions are compared to results of other studies concerned with the discrimination of spectral shape.

Profile analysis and background noise.

Spectral shape discrimination, or profile analysis, of complex waveforms (21 components) in the presence of broadband noise and special sinusoidal maskers of random amplitude was studied. The first experiment involved the discrimination between a standard flat spectrum and a "rippled" spectrum in broadband noise of different spectrum levels. Thresholds obtained under control conditions, without noise or without a standard, were used to estimate constants of an equation that predict thresholds where standard and noise are both present. The model assumes an external variance, produced by the noise, is added linearly to an internal variance caused by the flat standard. The mean squared error is less than 2 dB. The second experiment involved the detection of an increment on the center component of the 21-component standard. Added to the standard was an additional masking sinusoid of random amplitude. Both the frequency and the range of the random amplitude were varied and both showed a systematic influence on the detectability of the 1000-Hz increment.

Towards a model for discrimination of broadband signals.

The conventional model for broadband discrimination assumes that resolution is limited by peripheral internal noise that is statistically independent across channels. In this paper, we extend this model in a number of directions. In particular, we compute, compare, and discuss the effects of interchannel correlation and central noise on the sensitivity index d', for discrimination of overall level and discrimination of spectral shape.

Spectral‐shape discrimination of synthetic plosive bursts

The objective of this research is to measure the discriminability of speech‐like sounds on the basis of spectral shape alone and to relate these measurements to underlying auditory abilities by means of psychoacoustic experiments and quantitative models. Preliminary measurement using normal‐hearing listeners have been obtained on: (a) the discriminability of two synthetic plosive bursts (corresponding to /p/ and /t/ bursts in the context C /a/) and (b) the discriminability of incremental changes in the spectral shape os each synthetic stimulus in various frequency regions. Loundness and duration cues were eliminated by roving intensities and equating durations, and the difficulty of the tasks was controlled by adding low‐pass noise at various S/N ratios. The talk will include a description of the data and of initial attempts to develop a quantitative theoretical model of spectral shape discrimination. [Work supported by NIH.]

Discrimination of spectral shape by blinded veterans

The analysis of the acoustic environment for the purpose of gleaning spatial information is a very important sensory skill potentially available to an individual lacking sight. The purpose of the experimental procedure outlined in this paper was twofold: (1) to provide the means for predicting the level of auditory skills of individual clients and (2) to enhance the auditory skills of blinded clients in a manner more efficient than techniques presently available. The task involved the detection of a notch filtered out of a broad-band noise background. A standard (low pass-900 Hz, high pass-1300 Hz) and three different notch widths (850–1450, 800–1600, 750–1750) filtered by means of a General Radio 1952 Universal filter were used, providing the experimenters with three levels of difficulty in the comparison task. It was found that the test results correlated with performance in real-world acoustic assessment situations to a much greater degree than previously possible with pure tone audiometric results. Significant learning effects generalized to the real world situation will also be discussed.