Auditory Filter Shapes Research Articles

Improving Auditory Filter Estimation by Incorporating Absolute Threshold and a Level-dependent Internal Noise.

Auditory filter (AF) shape has traditionally been estimated with a combination of a notched-noise (NN) masking experiment and a power spectrum model (PSM) of masking. However, there are several challenges that remain in both the simultaneous and forward masking paradigms. We hypothesized that AF shape estimation would be improved if absolute threshold (AT) and a level-dependent internal noise were explicitly represented in the PSM. To document the interaction between NN threshold and AT in normal hearing (NH) listeners, a large set of NN thresholds was measured at four center frequencies (500, 1000, 2000, and 4000 Hz) with the emphasis on low-level maskers. The proposed PSM, consisting of the compressive gammachirp (cGC) filter and three nonfilter parameters, allowed AF estimation over a wide range of frequencies and levels with fewer coefficients and less error than previous models. The results also provided new insights into the nonfilter parameters. The detector signal-to-noise ratio () was found to be constant across signal frequencies, suggesting that no frequency dependence hypothesis is required in the postfiltering process. The ANSI standard "Hearing Level-0dB" function, i.e., AT of NH listeners, could be applied to the frequency distribution of the noise floor for the best AF estimation. The introduction of a level-dependent internal noise could mitigate the nonlinear effects that occur in the simultaneous NN masking paradigm. The new PSM improves the applicability of the model, particularly when the sound pressure level of the NN threshold is close to AT.

Auditory filter shapes derived from forward and simultaneous masking at low frequencies: Implications for human cochlear tuning

Behavioral forward-masking thresholds with a spectrally notched-noise masker and a fixed low-level probe tone have been shown to provide accurate estimates of cochlear tuning. Estimates using simultaneous masking are similar but generally broader, presumably due to nonlinear cochlear suppression effects. So far, estimates with forward masking have been limited to frequencies of 1 kHz and above. This study used spectrally notched noise under forward and simultaneous masking to estimate frequency selectivity between 200 and 1000 Hz for young adult listeners with normal hearing. Estimates of filter tuning at 1000 Hz were in agreement with previous studies. Estimated tuning broadened below 1000 Hz, with the filter quality factor based on the equivalent rectangular bandwidth (QERB) decreasing more rapidly with decreasing frequency than predicted by previous equations, in line with earlier predictions based on otoacoustic-emission latencies. Estimates from simultaneous masking remained broader than those from forward masking by approximately the same ratio. The new data provide a way to compare human cochlear tuning estimates with auditory-nerve tuning curves from other species across most of the auditory frequency range.

An adaptive transmission line cochlear model based front-end for replay attack detection

The cochlea is a remarkable spectrum analyser with desirable properties including sharp frequency tuning and level-dependent compression and the potential advantages of incorporating these characteristics in a speech processing front-end are investigated. This paper develops a framework for an active transmission line cochlear model employing adaptive notch and resonant filters. The proposed model reproduces the observed asymmetric auditory filter shape with a sharp high-frequency roll-off and level-dependent nonlinear dynamic range compression characteristics. Experimental analysis demonstrates that sharp frequency tuning and dynamic range compression of the proposed model lead to an enhanced spectral representation compared with other spectral analysis methods. The proposed model was employed in the front-end of replay spoofing attack detection systems, and experiments on the ASVspoof 2017 version 2.0 and ASVspoof 2019 databases demonstrate that the proposed model outperforms linear and nonlinear level-dependent parallel filter bank auditory models and classical spectro-temporal front-ends. The use of the proposed model leads to relative improvements of 45.6%, 51.9% and 60.8% over the baseline feature CQCCs of ASVspoof version 2.0 and CQCCs and LFCCs of ASVspoof2019 on evaluation datasets, respectively.

A Bayesian active-learning test for remote hearing-aid fitting

A major problem of remote or online hearing tests is the calibration of the equipment and control of the acoustic environment: For example, an audiogram requires calibrated headphones with a wide dynamic range and a silent room. Instead, a notched-noise test alleviates these problems and gives valuable information about auditory filters that can be used to fit a hearing aid. Although absolute levels still play a role, relative levels between signal and noise are more robust as the environment does not need to be quieter than the noise. Using traditional methods, determination of auditory-filter shapes across the whole frequency range would take hours. We developed a notched-noise test that uses Gaussian Processes to maintain a probabilistic estimate of the outcome, and to select stimulus parameters based on mutual information. This active-learning method gives an estimate of auditory-filter shapes in about 30 minutes across a wide range of frequency. We demonstrate that these estimates, even when obtained with headphones for which the frequency response and exact calibration is unknown, are sufficient to give a good initial fit of a hearing aid.

Application of Bayesian Active Learning to the Estimation of Auditory Filter Shapes Using the Notched-Noise Method.

Time-efficient hearing tests are important in both clinical practice and research studies. This particularly applies to notched-noise tests, which are rarely done in clinical practice because of the time required. Auditory-filter shapes derived from notched-noise data may be useful for diagnosis of the cause of hearing loss and for fitting of hearing aids, especially if measured over a wide range of center frequencies. To reduce the testing time, we applied Bayesian active learning (BAL) to the notched-noise test, picking the most informative stimulus parameters for each trial based on nine Gaussian Processes. A total of 11 hearing-impaired subjects were tested. In 20 to 30 min, the test provided estimates of signal threshold as a continuous function of frequency from 500 to 4000 Hz for nine notch widths and for notches placed both symmetrically and asymmetrically around the signal frequency. The thresholds were found to be consistent with those obtained using a 2-up/1-down forced-choice procedure at a single center frequency. In particular, differences in threshold between the methods did not vary with notch width. An independent second run of the BAL test for one notch width showed that it is reliable. The data derived from the BAL test were used to estimate auditory-filter width and asymmetry and detection efficiency for center frequencies from 500 to 4000 Hz. The results agreed with expectations for cochlear hearing losses that were derived from the audiogram and a hearing model.

Toward Routine Assessments of Auditory Filter Shape.

Purpose A Bayesian adaptive procedure, that is, the quick auditory filter (qAF) procedure, has been shown to improve the efficiency for estimating auditory filter shapes of listeners with normal hearing. The current study evaluates the accuracy and test-retest reliability of the qAF procedure for naïve listeners with a variety of ages and hearing status. Method Fifty listeners who were naïve to psychophysical experiments and exhibit wide ranges of age (19-70 years) and hearing threshold (-5 to 70 dB HL at 2 kHz) were recruited. Their auditory filter shapes were estimated for a 15-dB SL target tone at 2 kHz using both the qAF procedure and the traditional threshold-based procedure. The auditory filter model was defined using 3 parameters: (a) the sharpness of the tip portion of the auditory filter, p; (b) the prominence of the low-frequency tail of the filter, 10log( w); and (c) the listener's efficiency in detection, 10log( K). Results The estimated parameters of the auditory filter model were consistent between 2 qAF runs tested on 2 separate days. The parameter estimates from the 2 qAF runs also agreed well with those estimated using the traditional procedure despite being substantially faster. Across the 3 auditory filter estimates, the dependence of the auditory filter parameters on listener age and hearing threshold was consistent across procedures, as well as consistent with previously published estimates. Conclusions The qAF procedure demonstrates satisfactory test-retest reliability and good agreement to the traditional procedure for listeners with a wide range of ages and with hearing status ranging from normal hearing to moderate hearing impairment.

No Effect of Musical Training on Frequency Selectivity Estimated Using Three Methods.

It is widely believed that the frequency selectivity of the auditory system is largely determined by processes occurring in the cochlea. If so, musical training would not be expected to influence frequency selectivity. Consistent with this, auditory filter shapes for low center frequencies do not differ for musicians and nonmusicians. However, it has been reported that psychophysical tuning curves (PTCs) at 4000 Hz were sharper for musicians than for nonmusicians. This study explored the origin of the discrepancy across studies. Frequency selectivity was estimated for musicians and nonmusicians using three methods: fast PTCs with a masker that swept in frequency, "traditional" PTCs obtained using several fixed masker center frequencies, and the notched-noise method. The signal frequency was 4000 Hz. The data were fitted assuming that each side of the auditory filter had the shape of a rounded-exponential function. The sharpness of the auditory filters, estimated as the Q10 values, did not differ significantly between musicians and nonmusicians for any of the methods, but detection efficiency tended to be higher for the musicians. This is consistent with the idea that musicianship influences auditory proficiency but does not influence the peripheral processes that determine the frequency selectivity of the auditory system.

The dynamic gammawarp auditory filterbank.

Auditory filterbanks are an integral part of many metrics designed to predict speech intelligibility and speech quality. Considerations in these applications include accurate reproduction of auditory filter shapes, the ability to reproduce the impact of hearing loss as well as normal hearing, and computational efficiency. This paper presents an alternative method for implementing a dynamic compressive gammachirp (dcGC) auditory filterbank [Irino and Patterson (2006). IEEE Trans. Audio Speech Lang. Proc. 14, 2222-2232]. Instead of using a cascade of second-order sections, this approach uses digital frequency warping to give the gammawarp filterbank. The set of warped finite impulse response filter coefficients is constrained to be symmetrical, which results in the same phase response for all filters in the filterbank. The identical phase responses allow the dynamic variation in the gammachirp filter magnitude response to be realized as a sum, using time-varying weights, of three filters that provide the responses for high-, mid-, and low-intensity input signals, respectively. The gammawarp filterbank offers a substantial improvement in execution speed compared to previous dcGC implementations; for a dcGC filterbank, the gammawarp implementation is 24 to 38 times faster than the dcGC Matlab code of Irino.

Frequency selectivity in macaque monkeys measured using a notched-noise method

The auditory system is thought to process complex sounds through overlapping bandpass filters. Frequency selectivity as estimated by auditory filters has been well quantified in humans and other mammalian species using behavioral and physiological methodologies, but little work has been done to examine frequency selectivity in nonhuman primates. In particular, knowledge of macaque frequency selectivity would help address the recent controversy over the sharpness of cochlear tuning in humans relative to other animal species. The purpose of our study was to investigate the frequency selectivity of macaque monkeys using a notched-noise paradigm. Four macaques were trained to detect tones in noises that were spectrally notched symmetrically and asymmetrically around the tone frequency. Masked tone thresholds decreased with increasing notch width. Auditory filter shapes were estimated using a rounded exponential function. Macaque auditory filters were symmetric at low noise levels and broader and more asymmetric at higher noise levels with broader low-frequency and steeper high-frequency tails. Macaque filter bandwidths (BW3dB) increased with increasing center frequency, similar to humans and other species. Estimates of equivalent rectangular bandwidth (ERB) and filter quality factor (QERB) suggest macaque filters are broader than human filters. These data shed further light on frequency selectivity across species and serve as a baseline for studies of neuronal frequency selectivity and frequency selectivity in subjects with hearing loss.

A likelihood-based method for point and interval estimation on auditory filter data

A novel method for determining auditory filter (AF) shapes given a set of n-alternative forced choice (nAFC) responses from a single human subject or set of subjects is discussed. The method works by developing a function which maps individual nAFC responses into likelihood values—either supporting or conflicting with a proposed model AF shape. The aggregate of these likelihoods is then used as an objective function for optimization schemes for point estimation, or as the basis function for Metropolis Hastings-like algorithms for interval estimation, both of either parameters of the AF model or of the entire AF shape. The method is demonstrated on simulated up-down staircase data. The consistency of the method is discussed in the context of canonical methods for AF data analysis, some of which are shown to produce systematic errors. Other possible benefits of this approach are discussed including the ability of the method to: combine data from heterogeneous nAFC tasks (e.g., notched-noise maskers with ton...

Effects of Low- and High-Frequency Side Bands of Notched Noise on Masking and Auditory Filter Shape at Very High Frequencies

Abstract This paper is concerned with the determination of the auditory filter shape using the notched noise method with noise bands symmetrically located above and below a probe frequency of 10 kHz. Unlike in the classical experiments conducted with the use of Patterson method the levels as well as power spectrum densities of the lower and upper component bands of the notched noise masker were not the same and were set such as to produce the same amount of masking at the 10-kHz frequency. The experiment consisted of three conditions in which the following values were determined: (I) the detection threshold for a 10-kHz probe tone in the presence of a noise masker presented below the tone’s frequency; (II) the level of a noise masker presented above the 10-kHz probe tone frequency, at which the masker just masked the probe tone, (III) the detection threshold for a probe tone in the presence of a notched-noise masker. The data show a considerable amount of variability across the subjects, however, the resulting frequency characteristics of the auditory filters are consistent with those presented in the literature so that the Equivalent Rectangular Bandwidth is less than 11% of their centre frequency.

Rapid estimation of high-parameter auditory-filter shapes.

A Bayesian adaptive procedure, the quick-auditory-filter (qAF) procedure, was used to estimate auditory-filter shapes that were asymmetric about their peaks. In three experiments, listeners who were naive to psychoacoustic experiments detected a fixed-level, pure-tone target presented with a spectrally notched noise masker. The qAF procedure adaptively manipulated the masker spectrum level and the position of the masker notch, which was optimized for the efficient estimation of the five parameters of an auditory-filter model. Experiment I demonstrated that the qAF procedure provided a convergent estimate of the auditory-filter shape at 2 kHz within 150 to 200 trials (approximately 15 min to complete) and, for a majority of listeners, excellent test-retest reliability. In experiment II, asymmetric auditory filters were estimated for target frequencies of 1 and 4 kHz and target levels of 30 and 50 dB sound pressure level. The estimated filter shapes were generally consistent with published norms, especially at the low target level. It is known that the auditory-filter estimates are narrower for forward masking than simultaneous masking due to peripheral suppression, a result replicated in experiment III using fewer than 200 qAF trials.

Dolphin biosonar target detection in noise: wrap up of a past experiment.

The target detection capability of bottlenose dolphins in the presence of artificial masking noise was first studied by Au and Penner [J. Acoust. Soc. Am. 70, 687-693 (1981)] in which the dolphins' target detection threshold was determined as a function of the ratio of the echo energy flux density and the estimated received noise spectral density. Such a metric was commonly used in human psychoacoustics despite the fact that the echo energy flux density is not compatible with noise spectral density which is averaged intensity per Hz. Since the earlier detection in noise studies, two important parameters, the dolphin integration time applicable to broadband clicks and the dolphin's auditory filter shape, were determined. The inclusion of these two parameters allows for the estimation of the received energy flux density of the masking noise so that the dolphin target detection can now be determined as a function of the ratio of the received energy of the echo over the received noise energy. Using an integration time of 264 μs and an auditory bandwidth of 16.7 kHz, the ratio of the echo energy to noise energy at the target detection threshold is approximately 1 dB.

The relationship between speaker size perception and the auditory filter

When we hear a new voice on the radio, we can tell whether the speaker is an adult or a child. We can also extract the message of the communication without being confused by the size information. This shows that auditory signal processing is scale invariant, automatically segregating information about vocal tract shape from information about vocal tract length. Patterson and colleagues have performed a series of experiments to measure the characteristics of size/shape perception [e.g., Smith et al., J. Acoust. Soc. Am. 117(1), 305–318 (2005)], and provided a mathematical basis for auditory scale invariance in the form of the stabilized wavelet-Mellin transform (SWMT) [Irino and Patterson, Speech Commun. 36(3–4), 181–203 (2002)]. The mathematics of the SWMT dictates the optimal form of the auditory filter, insofar as it must satisfy minimal uncertainty in a time-scale representation [Irino and Patterson, J. Acoust. Soc. Am. 101(1), 412–419 (1997)]. The resulting gammachirp auditory filter is an asymmetric extension of the earlier gammatone auditory filter—one which can explain the level dependence of notched-noise masking. Thus, although it is not immediately intuitive, speaker size perception and auditory filter shape are both aspects of a larger, unified framework for auditory signal processing.

Bayesian estimation of high-parameter models of the auditory filter

The Bayesian adaptive procedure proposed by Shen and Richards [J. Acoust. Soc. Am. 134(2), 1134–1145 (2013)] was extended to allow the rapid estimation of auditory filters that were asymmetric about their peak frequencies. The estimation of the auditory-filter shape (five free parameters) was achieved using single Bayesian adaptive tracks of 150–200 trials (approximately 15 min to complete). During the experimental track, listeners detected a tonal signal presented with either simultaneous or forward maskers. Both types of maskers consisted of two bands of noises, one on each side of the signal frequency. The Bayesian adaptive procedure iteratively updated the parameter estimates following each experimental trial and determined the stimulus that would maximize the gain of information on the following trial. The stimuli were adaptively manipulated along three dimensions: the masker level and the spectral location of the upper and lower masker bands. The proposed procedure allowed the reliable estimation of the auditory-filter shape for naïve normal-hearing listeners. The model predictions replicated the known effect of the masker-signal simultaneity on the auditory filter.

Dolphin biosonar target detection in noise, auditory filter shape, and temporal integration

The biosonar target detection in noise capability of the Atlantic bottlenose dolphin (Tursiops truncatus) has been studied as early as 1981 by Au and Penner. At that time, they presented results of the performance of two dolphins as a function of the signal energy to the density of the noise intensity (as in human psycho-acoustics) and not the signal energy to the noise energy ratio. Subsequently, two important pieces of information were determined, that is the auditory filter shape of the bottlenose dolphin in 2012 along with the temporal integration time for the reception of broadband biosonar echoes in 1988. In all but one experiment, it has been shown that the auditory system of odontocetes has a constant-Q characteristic. The biosonar detection capability of the bottlenose dolphin is revisited, and the target detection performance is now determined as a function of the received energy in the echo to the received noise energy. This presentation is an example of how some biosonar questions can only be ...

Bayesian adaptive estimation of the auditory filter

A Bayesian adaptive procedure for estimating the auditory-filter shape was proposed and evaluated using young, normal-hearing listeners at moderate stimulus levels. The resulting quick-auditory-filter (qAF) procedure assumed the power spectrum model of masking with the auditory-filter shape being modeled using a spectrally symmetric, two-parameter rounded-exponential (roex) function. During data collection using the qAF procedure, listeners detected the presence of a pure-tone signal presented in the spectral notch of a noise masker. Dependent on the listener's response on each trial, the posterior probability distributions of the model parameters were updated, and the resulting parameter estimates were then used to optimize the choice of stimulus parameters for the subsequent trials. Results showed that the qAF procedure gave similar parameter estimates to the traditional threshold-based procedure in many cases and was able to reasonably predict the masked signal thresholds. Additional measurements suggested that occasional failures of the qAF procedure to reliably converge could be a consequence of incorrect responses early in a qAF track. The addition of a parameter describing lapses of attention reduced the likelihood of such failures.

Binaural notched-noise masking and auditory-filter shape

Thresholds for sinusoids interaurally in phase (S0) and antiphase (Sπ) were measured in the presence of a diotic notched-noise masker (N0) as a function of notch width. The signal frequency was 250, 500, 1000, or 2000 Hz. For all signal frequencies, the difference between N0S0 and N0Sπ thresholds (binaural masking-level difference, BMLD) decreased continuously as the notch width increased. Model simulations showed that this result cannot be accounted for by a model that only processes the output of the auditory filter centered at the signal frequency, even if the nonlinear behavior of the monaural frequency selectivity or interaural differences in the filter shape are considered. The data were predicted well if a detrimental across-channel process was included, either by an addition of portions of the output of adjacent filters to the output of the on-frequency filter or by a notch-width dependent adverse shift in interaural phase in the binaural stage. The strength of this detrimental across-channel process tends to decrease with increasing signal frequencies.

Rapid measurement of auditory filter shape in mice using the auditory brainstem response and notched noise

The notched noise method is an effective procedure for measuring frequency resolution and auditory filter shapes in both human and animal models of hearing. Briefly, auditory filter shape and bandwidth estimates are derived from masked thresholds for tones presented in noise containing widening spectral notches. As the spectral notch widens, increasingly less of the noise falls within the auditory filter and the tone becomes more detectible until the notch width exceeds the filter bandwidth. Behavioral procedures have been used for the derivation of notched noise auditory filter shapes in mice; however, the time and effort needed to train and test animals on these tasks renders a constraint on the widespread application of this testing method. As an alternative procedure, we combined relatively non-invasive auditory brainstem response (ABR) measurements and the notched noise method to estimate auditory filters in normal-hearing mice at center frequencies of 8, 11.2, and 16 kHz. A complete set of simultaneous masked thresholds for a particular tone frequency were obtained in about an hour. ABR-derived filter bandwidths broadened with increasing frequency, consistent with previous studies. The ABR notched noise procedure provides a fast alternative to estimating frequency selectivity in mice that is well-suited to high through-put or time-sensitive screening.

High-frequency auditory filter shape for the Atlantic bottlenose dolphin

High-frequency auditory filter shapes of an Atlantic bottlenose dolphin (Tursiops truncatus) were measured using a notched noise masking source centered on pure tone signals at frequencies of 40, 60, 80 and 100 kHz. A dolphin was trained to swim into a hoop station facing the noise/signal transducer located at a distance of 2 m. The dolphin's masked threshold was determined using an up-down staircase method as the width of the notched noise was randomly varied from 0, 0.2, 04, 0.6, and 0.8 times the test tone frequency. The masked threshold decreased as the width of the notched increased and less noise fell within the auditory filter associated with the test tone. The auditory filter shapes were approximated by fitting a roex (p,r(r)) function to the masked threshold results. A constant-Q value of 8.4 modeled the results within the frequency range of 40 to 100 kHz relatively well. However, between 60 and 100 kHz, the 3 dB bandwidth was relatively similar between 9.5 and 10 kHz, indicating a constant-bandwidth system in this frequency range The mean equivalent rectangular bandwidth calculated from the filter shape was approximately 16.0%, 17.0%, 13.6% and 11.3% of the tone frequencies of 40, 60, 80, and 100 kHz.