Showing papers by "AG Armin Kohlrausch published in 1995"

Phase effects in masking related to dispersion in the inner ear. II. Masking period patterns of short targets

Abstract: This article investigates how the amplitude and phase characteristics of the inner ear influence the spectrotemporal representation of harmonic complex sounds. Five experiments are reported, in each of which three sets of maskers are compared that differ only in their phase spectra. The amplitude spectra of the complexes were flat and the phase choices were (a) zero phase, (b) Schroeder phases with a positive sign, and (c) Schroeder phases with a negative sign. In the first four experiments, the spectra contained all harmonics between 200 and 2000 Hz. In experiments 1 and 2, the signal frequency was fixed at 1100 Hz and the fundamental frequency of the maskers was varied. In experiments 3 and 4, the fundamental frequency of the maskers was fixed and the signal frequency varied between 200 and 2000 Hz. In experiments 1 and 3, the signal duration was long compared to the period of the maskers. In experiments 2 and 4, the signal duration was only 5 ms and thresholds were determined for different time points within the masker's period. The results show a strong correlation between the minima of the short signal's thresholds and the threshold of the long signal. In experiment 5, the spectral extent of the masker was shifted to values one octave lower (100 to 1000 Hz) or one or two octaves higher (400 to 4000 Hz and 800 to 8000 Hz, respectively). For each spectral region, masked thresholds of a long signal were obtained for three values of the fundamental frequency. In all five experiments the thresholds depended very much on the specific phase choices with differences of up to 25 dB. The masker with a negative Schroeder phase always led to the highest thresholds. The thresholds of the masker with a positive Schroeder phase, on the other hand, were for a wide range of parameters lower than the thresholds for the zero-phase masker. These phase effects are most likely caused by the phase characteristic of the basilar-membrane filter, which affects the flat envelopes of the two Schroeder-phase maskers in a very different way. For an appropriate choice of parameters, one of the two becomes even more strongly modulated than the zero-phase complex. This latter observation imposes some restrictions on the second derivative (curvature) of the phase-versus-frequency relation for the auditory filters.

Analytical expressions for the envelope correlation of certain narrow‐band stimuli

Abstract: Analytical solutions are presented for the correlation between the envelopes of certain narrow‐band stimuli which are typically used in binaural research. These stimuli are either the sum of an in‐phase masker and an out‐of‐phase signal, or they are two partially correlated noises. Because an envelope has a nonzero mean, the solutions differ depending on whether the envelope correlation is expressed as the normalized cross correlation or as the normalized cross covariance (Pearson product‐moment correlation). The envelope correlation depends on the statistics of the masker and the signal whereas the waveform correlation depends on neither. This influence only disappears for the normalized envelope cross correlation provided that the correlation is close to one. In this case, the normalized envelope cross correlation is equal to the square root of the waveform correlation. The results for two partially correlated noise bands are also of relevance for experiments dealing with monaural envelope discrimination and comodulation masking release.

The role of envelope fluctuations in spectral masking

Abstract: Two experiments are reported in this study. In the first experiment the masking effect of five different types of narrow‐band maskers was compared. The masker was either a tone, a narrow‐band Gaussian noise, or a multiplication noise obtained by multiplying a sinusoid with a low‐pass Gaussian noise. The noise maskers had a bandwidth of either 20 or 100 Hz. In all cases the masker had a center frequency of 1.3 kHz and a duration of 500 ms. Five‐point growth‐of‐masking functions were measured using a 2‐kHz tonal target with a duration of 400 ms, temporally centered in the masker. Six subjects participated in the experiment. Although considerable intersubject differences were observed, the data of all subjects showed several common trends. First, the tonal maskers produced more masking than the noise maskers. Second, Gaussian noise maskers produced more masking than multiplication noise maskers of the same bandwidth. Finally, 100‐Hz‐wide noise maskers produced more masking than 20‐Hz‐wide maskers of the same type. Differences in masked thresholds between the various masker types generally increased with masker level, and exceeded 25 dB in some conditions. The results are discussed in terms of masker envelope fluctuations. In the second experiment the masking effect was investigated for a bandpass noise at 1.3 kHz, with regular zero crossings, but with the envelope characteristics of a 100‐Hz‐wide Gaussian noise. Five‐point growth‐of‐masking functions were measured using a tonal target of 2 kHz. Masked thresholds produced by this hybrid masker were not significantly different from those produced by a 100‐Hz Gaussian masker, but differed significantly from those produced by 100‐Hz‐wide multiplication noise. This result indicates that differences in masking between Gaussian and multiplication noise are not due to their different fine structure, but to their different envelope statistics.

Comparison of auditory filter shapes obtained with notched-noise and noise-tone maskers.

Abstract: The notched‐noise method has been widely used to estimate the shape of the auditory filter. Results obtained using this method may be influenced by combination bands produced by the interaction of components within the upper band of noise in the notched‐noise masker. To assess the possible effect of such combination bands, results were compared for two types of masker: A notched noise, as used in previous experiments; and a masker in which the upper band of noise was replaced by a sinusoid with a frequency corresponding to the lower edge frequency of that band. This is referred to as the noise‐tone masker. The signal frequency was 2 kHz, and measurements were obtained for two different spectrum levels of the noise masker, 30 and 45 dB. Auditory filter shapes derived using the two maskers were similar on their low‐frequency sides, as expected. The low‐frequency sides were less steep at the higher masker level. The high‐frequency sides of the auditory filters derived using the noise‐tone masker were sometimes slightly steeper than those obtained using the notched‐noise masker, but the effect was generally small. Changes with level on the high‐frequency sides were not consistent across subjects. An analysis of the notched‐noise data taking into account the effects of the combination bands suggests that the maximal spectrum level of the combination bands, in the region just below the lower spectral edge of the primary noise band, is about 20 to 30 dB below the spectrum level of the primary band. At this relative level, the combination bands have only a very small influence on the high‐frequency sides of the derived auditory filters. The influence on the estimated equivalent rectangular bandwidths (ERBs) of the auditory filters is usually negligible.

Comparison of monaural (CMR) and binaural (BMLD) masking release

Abstract: Release of masking for a sinusoidal signal of 5 kHz masked by a 25-Hz-wide noise band centered around 5 kHz was measured. The masking release was provided by a second noise band that was comodulated with the on-frequency masker band. For CMR configurations the second noise band was centered at 3 kHz and presented to the ipsi-lateral or to the contra-lateral ear. For BMLD configurations the second band was centered at 5 kHz and presented to the contra-lateral ear. In another condition the second noise band also contained the signal presented with such a phase that maximal differences in the envelope resulted. For both the CMR and the BMLD paradigm, the masking release for the latter condition was larger than for the former condition. To assess further the similarity between monaural and binaural masking release, a sinusoidal masker and either a noise or a sinusoidal signal were used. The data indicate that, at high frequencies, envelope correlation may be a valuable cue for CMR as well as for the BMLD.

Modulation detection as a function of carrier frequency and level

Abstract: This paper describes recent experiments investigating temporal processing in the auditory system. Subjects had to discriminate sinusoidal signals with a flat temporal envelope from those with a sinusoidal amplitude modulation. The modulation depth at threshold was measured for a wide range of carrier frequencies, carrier levels and modulation rates. For modulation rates up to 100Hz, thresholds remained constant and were the same for all centre frequencies tested. For higher rates, modulation thresholds initially increased and then decreased rapidly, reflecting the subjects' ability to resolve the sidebands spectrally. Detection thresholds generally improved with increasing carrier level. Most of these observations differ from results obtained for modulated broadband noises, and some are not consistent with published results for sinusoidal carriers.

Model simulations of masked thresholds for tones in dichotic noise maskers

Abstract: The study of masked thresholds in dichotic noise maskers is important for understanding the processing in binaural hearing. To simulate these thresholds a psychoacoustically motivated perception model was used [T. Dau et al. (1995). ‘‘A quantitative model of the ‘‘effective’’ signal processing in the auditory system: I. Model structure,’’ submitted to J. Acoust. Soc. Am.]. This model, which has been successfully applied to several monaural psychoacoustical experiments, was extended by an additional binaural processing unit. It consists of a filterbank, half‐wave rectifier, low‐pass filter, and adaptation loops, which model the temporal processing. The binaural processing unit detects the interaural correlation and makes decisions based on the difference between the signals from both ears. Masked thresholds in the NoSπ and NπSo configurations, obtained as a function of noise masker frequency and bandwidth, were simulated and compared to new experimental measurements. The dependence on interaural delay and ...

The role of distortion products in masking by single bands of noise

Abstract: Masking experiments with frozen‐noise maskers were conducted to investigate the role of distortion products generated by the interaction between the components of a bandpass noise masker. In the first experiment, thresholds of a 900‐Hz sinusoidal signal with a duration of 50 ms (10‐ms ramps included) were measured in the presence of bandpass noise maskers ranging from 1 to 2 kHz. In all measurements the same 500‐ms noise sample was used (frozen noise), presented at overall sound‐pressure levels of 35, 50, 65, or 80 dB. The signal was temporally centered in the masker. Four subjects participated in the experiment. Threshold variations of more than 10 dB were observed on varying the signal phase. The pattern of threshold versus signal phase resembled a sinusoid; the signal phases at the minimum and maximum thresholds differed by about 180 deg. The phase pattern shifted with increasing masker level. The individual shift for the masker range of 45 dB was between 100 and 200 deg. The direction of this shift ag...

Modeling modulation perception: Modulation low‐pass filter or modulation filter bank?

Abstract: In current models of modulation perception, the stimuli are first filtered and nonlinearly transformed (mostly half‐wave rectified). In order to model the low‐pass characteristic of measured modulation transfer functions, the next stage in the models is a first‐order low‐pass filter with a typical cutoff frequency of 50 to 60 Hz. From physiological studies in mammals it is known that many neurons in, e.g., the inferior colliculus, show a bandpass characteristic in their sensitivity to amplitude modulation. Results from psychophysical studies of modulation masking also suggest some kind of bandpass analysis of modulation frequencies. Results of two experiments on modulation detection that allow discrimination between models incorporating a low‐pass filter and those using a modulation filterbank are presented. In the first experiment, modulation detection thresholds were measured for noise carriers of bandwidths between 3 and 6000 Hz. In the second experiment, modulation detection for a sinusoidal carrier w...