Interaural Parameters Research Articles

Speech Sources Separation Based on Models of Interaural Parameters and Spatial Properties of Room

Speech Sources Separation Based on Models of Interaural Parameters and Spatial Properties of Room

Binaural Recordings in Natural Acoustic Environments: Estimates of Speech-Likeness and Interaural Parameters.

Binaural acoustic recordings were made in multiple natural environments, which were chosen to be similar to those reported to be difficult for listeners with impaired hearing. These environments include natural conversations that take place in the presence of other sound sources as found in restaurants, walking or biking in the city, and so on. Sounds from these environments were recorded binaurally with in-the-ear microphones and were analyzed with respect to speech-likeness measures and interaural difference measures. The speech-likeness measures were based on amplitude–modulation patterns within frequency bands and were estimated for 1-s time-slices. The interaural difference measures included interaural coherence, interaural time difference, and interaural level difference, which were estimated for time-slices of 20-ms duration. These binaural measures were documented for one-fourth-octave frequency bands centered at 500 Hz and for the envelopes of one-fourth-octave bands centered at 2000 Hz. For comparison purposes, the same speech-likeness and interaural difference measures were computed for a set of virtual recordings that mimic typical clinical test configurations. These virtual recordings were created by filtering anechoic waveforms with available head-related transfer functions and combining them to create multiple source combinations. Overall, the speech-likeness results show large variability within and between environments, and they demonstrate the importance of having information from both ears available. Furthermore, the interaural parameter results show that the natural recordings contain a relatively small proportion of time-slices with high coherence compared with the virtual recordings; however, when present, binaural cues might be used for selecting intervals with good speech intelligibility for individual sources.

Robust auditory localization using probabilistic inference and coherence-based weighting of interaural cues.

Robust sound source localization is performed by the human auditory system even in challenging acoustic conditions and in previously unencountered, complex scenarios. Here a computational binaural localization model is proposed that possesses mechanisms for handling of corrupted or unreliable localization cues and generalization across different acoustic situations. Central to the model is the use of interaural coherence, measured as interaural vector strength (IVS), to dynamically weight the importance of observed interaural phase (IPD) and level (ILD) differences in frequency bands up to 1.4 kHz. This is accomplished through formulation of a probabilistic model in which the ILD and IPD distributions pertaining to a specific source location are dependent on observed interaural coherence. Bayesian computation of the direction-of-arrival probability map naturally leads to coherence-weighted integration of location cues across frequency and time. Results confirm the model's validity through statistical analyses of interaural parameter values. Simulated localization experiments show that even data points with low reliability (i.e., low IVS) can be exploited to enhance localization performance. A temporal integration length of at least 200 ms is required to gain a benefit; this is in accordance with previous psychoacoustic findings on temporal integration of spatial cues in the human auditory system.

The impact of reference tones on the adjustment of interaural cues.

In the time-intensity trading paradigm, trading ratios are inconsistent in that they differ as a function of which cue is to be adjusted by the listener. Two explanations have been offered: First, the regression model assumes a regression to the interaural parameters of a reference tone played in alternation with the test tone to cause the phenomenon of inconsistent trading ratios. The second explanation is based on an inflated perceptual weighting of the to-be-adjusted cue. The perceptual-weight explanation has been supported by experimental results showing that the phenomenon of inconsistent trading ratios appears even in the absence of a reference tone. Those findings render regression as the sole explanation for inconsistent trading ratios implausible. The experiments presented in this paper address the question whether regression to the parameters of a reference tone plays a role if a reference tone is presented. Three experiments were conducted in which trials with and without reference tone were compared directly. Both within- and between-subject comparisons showed that a reference tone affects trading ratios and location judgments if present. Although regression cannot be the sole explanation for the phenomenon of inconsistent trading ratios it seems to play a role if a reference tone is presented.

Binaural spectral resolution as a function of interaural masker correlation.

Thresholds for an antiphasic 500-Hz sinusoid (Sπ) were measured in the presence of a notched-noise masker for three different interaural masker correlations: -1 (Nπ), 1 (N0), and 0.87. The difference between thresholds for the three masker correlations was largest for a notch width of zero and decreased continuously with increasing notch width. The bandwidth of a gammatone filter fitted to the data was 85 Hz for the interaural masker correlation of -1 (NπSπ) and 182 Hz for the interaural masker correlation of 1 (N0Sπ). For the intermediate correlation (0.87) the effective filter width was 134 Hz. This result is at odds with corresponding literature results of a bandwidening experiment where an effectively larger binaural bandwidth is only found with extreme interaural disparities of signal and masker, such as N0Sπ. The notched-noise thresholds were predicted if a detrimental across-channel process was included in the model. This approach failed to predict the effect of masker correlation in the bandwidening experiments. A beneficial across-channel process successfully used in the literature to simulate bandwidening data with extreme binaural parameters also failed to predict this effect. The effect may be due to interaural parameter fluctuations currently not used in the model framework.

Detection of simultaneous modulation of interaural time and level differences: Effects of modulation rate and relative phase (L)

The binaural system is known to be sluggish, i.e., unable to track modulations in interaural parameters even at a relatively slow rate. The present study evaluated the binaural system's sensitivity to modulation phase rather than to modulation magnitude. The detectability of simultaneous modulations in interaural time and level differences with various relative phases were measured. It was found that for modulation rates up to 10-20 Hz, the detectability varied with the relative phase. This indicates that information about higher rates is lost at or below the level of cue integration.

Binaural room acoustics

Room acoustics concerns the geometrical and materials properties of rooms as they determine measurable sound field parameters that are perceptually important to listeners. For example, the volume and absorption in a room determine the reverberation time, which is important to the comprehension of speech and the enjoyment of music. Binaural room acoustics concerns the properties of rooms as they affect measurable interaural parameters, important to the human binaural perceptual system. For example, room properties affect the short-term interaural cross-correlation, which is important to the perception of apparent source width. Our recent work emphasizes room effects on steady-state interaural level and phase differences, important to sound localization in the horizontal plane. Particularly, we have sought to give mathematical meaning to a “binaural critical distance.” If a sound source and a listener are separated by less than the binaural critical distance, there is good probability that the interaural differences correctly indicate whether the source is on the listener's left or right. Experimentally, we have focused on sine tones in the range 200–1200 Hz, but we expect the results to be more generally applicable. [Work supported by the AFOSR (grant 11NL002) and by the NSF REU program.]

Exploratory analysis on the binaural characteristics of the rowing sound

Differences between the signals captured at the entrances to the left and right ears are generally quantified by interaural time differences (ITDs), interaural level differences (ILDs) and interaural crosscorrelation (IC). Here, we attempt to gain knowledge on the changes of these interaural parameters during rowing and to evaluate their potential as source of information to determine rowing expertise exclusively from acoustic signals.

Model-Based Expectation-Maximization Source Separation and Localization

This paper describes a system, referred to as model-based expectation-maximization source separation and localization (MESSL), for separating and localizing multiple sound sources from an underdetermined reverberant two-channel recording. By clustering individual spectrogram points based on their interaural phase and level differences, MESSL generates masks that can be used to isolate individual sound sources. We first describe a probabilistic model of interaural parameters that can be evaluated at individual spectrogram points. By creating a mixture of these models over sources and delays, the multi-source localization problem is reduced to a collection of single source problems. We derive an expectation-maximization algorithm for computing the maximum-likelihood parameters of this mixture model, and show that these parameters correspond well with interaural parameters measured in isolation. As a byproduct of fitting this mixture model, the algorithm creates probabilistic spectrogram masks that can be used for source separation. In simulated anechoic and reverberant environments, separations using MESSL produced on average a signal-to-distortion ratio 1.6 dB greater and perceptual evaluation of speech quality (PESQ) results 0.27 mean opinion score units greater than four comparable algorithms.

Model‐based direction estimation of concurrent speakers in the horizontal plane.

A computational auditory model for the direction estimation of concurrent speakers in the horizontal plane is presented. The auditory signal processing is motivated by mammalian physiology and human psychoacoustic data, extending the binaural model by Dietz et al. [Brain Res., 1220, 234–245 (2008)]. The model exploits a wide range of interaural parameters, namely, the interaural level difference, the interaural phase difference of the fine structure, and the interaural coherence. The parameters are determined at the output of auditory filters allowing for a simultaneous estimation of several sound source directions in a multidimensional internal representation. By focusing on time segments with high interaural coherence [Faller and Merimaa, J. Acoust. Soc. Am. 116, 3075–3089 (2004)] in the auditory bands, which likely stem from a localized sound source, the robustness of the direction estimation of the model was considerably improved: for two well‐separated speakers in quiet, the estimation error is generally less than or equal to 10 deg. In noisy environments, the direction of a single speaker can be estimated for signal‐to‐noise ratios below 0 dB. Further improvements of the model by concurrent pitch estimation to reduce permutation errors between nearby speakers are investigated and discussed.

The influence of masker fringe on sound localization in noise.

Several studies have shown that detection of a dichotic target signal in a concurrent diotic noise masker improves when the masker onset occurs prior to that of the target and/or the masker duration extends beyond the target offset (so‐called masker fringe; McFadden, 1966; Trahiotis et al. 1972). It has been argued that this improved performance results from the fact that the masker fringe provides a baseline set of interaural parameters against which a change in parameters, associated with the signal onset or offset, can be detected. As the duration of the fringe increases, more time is available to estimate these baseline parameters, thus improving signal detectability. However, little is known about the relation between masker fringe and spatial hearing. In this study, the localization of a 250‐ms click train signal presented in a broadband directional noise masker was measured as a function of the duration of the masker fringe and the signal‐to‐noise ratio. The results indicate that the presence of masker fringe improves localization performance relative to the case in which the masker and signal are pulsed on and off simultaneously. Moreover, the benefit of masker fringe increases with decreasing signal‐to‐noise ratio. [Work supported by a grant from AFOSR.]

Binaural room acoustics - cross-correlation

The subject of ''Binaural room acoustics" combines the disciplines of room acoustics with facts and models of human binaural physiology and psychology. It considers the linear effects of rooms on the inputs to the two ears of a listener, and it considers the nonlinear effects of binaural hearing processes on perceptually relevant acoustical characterizations of rooms. Linear effects of interest are binaural differences in amplitude and phase spectra, which are combined in the cross-correlation function (CCF). A psychophysically relevant CCF is computed over frequency bands approximately corresponding to auditory filters. Because the bands are narrow, the effects on the CCF of frequency-dependent interaural parameters, most importantly dispersion, can be modeled analytically. The roles of amplitude incoherence and dispersion can also be independently modelled. A systematic experimental approach begins with the dependence of the CCF on head related transfer functions (HRTF) as measured at different azimuths in a free field. Although there are notable individual differences, the effects of HRTFs appear to be almost entirely below detection threshold. An empirical relationship between the CCF for a waveform and for its envelope has been established numerically and in extensive room experiments. [Work supported by the NIDCD, grant DC00181.]

Sound source localization in real sound fields based on empirical statistics of interaural parameters

The role of temporal fluctuations and systematic variations of interaural parameters in localization of sound sources in spatially distributed, nonstationary noise conditions was investigated. For this, Bayesian estimation was applied to interaural parameters calculated with physiologically plausible time and frequency resolution. Probability density functions (PDFs) of the interaural level differences (ILDs) and phase differences (IPDs) were estimated by measuring histograms for a directional sound source perturbed by several types of interfering noise at signal-to-noise ratios (SNRs) between -5 and +30 dB. A moment analysis of the PDFs reveals that the expected values shift and the standard deviations increase considerably with decreasing SNR, and that the PDFs have non-Gaussian shape at medium SNRs. A d' analysis of the PDFs indicates that elevation discrimination is possible even at low SNRs in the median plane by integrating information across frequency. Absolute sound localization was simulated by a Bayesian maximum a posteriori (MAP) procedure. The simulation is based on frequency integration of broadly tuned "detectors." Confusion patterns of real and estimated sound source directions are similar to those of human listeners. The results indicate that robust processing strategies are needed to exploit interaural parameters successfully in noise conditions due to their strong temporal fluctuations.

Investigation of the relationship among three common measures of precedence: fusion, localization dominance, and discrimination suppression.

Listeners have a remarkable ability to localize and identify sound sources in reverberant environments. The term "precedence effect" (PE; also known as the "Haas effect," "law of the first wavefront," and "echo suppression") refers to a group of auditory phenomena that is thought to be related to this ability. Traditionally, three measures have been used to quantify the PE: (1) Fusion: at short delays (1-5 ms for clicks) the lead and lag perceptually fuse into one auditory event; (2) Localization dominance: the perceived location of the leading source dominates that of the lagging source; and (3) Discrimination suppression: at short delays, changes in the location or interaural parameters of the lag are difficult to discriminate compared with changes in characteristics of the lead. Little is known about the relation among these aspects of the PE, since they are rarely studied in the same listeners. In the present study, extensive measurements of these phenomena were made for six normal-hearing listeners using 1-ms noise bursts. The results suggest that, for clicks, fusion lasts 1-5 ms; by 5 ms most listeners hear two sounds on a majority of trials. However, localization dominance and discrimination suppression remain potent for delays of 10 ms or longer. Results are consistent with a simple model in which information from the lead and lag interacts perceptually and in which the strength of this interaction decreases with spatiotemporal separation of the lead and lag. At short delays, lead and lag both contribute to spatial perception, but the lead dominates (to the extent that only one position is ever heard). At the longest delays tested, two distinct sounds are perceived (as measured in a fusion task), but they are not always heard at independent spatial locations (as measured in a localization dominance task). These results suggest that directional cues from the lag are not necessarily salient for all conditions in which the lag is subjectively heard as a separate event.

Binaural sluggishness in the perception of tone sequences and speech in noise.

The binaural system is well-known for its sluggish response to changes in the interaural parameters to which it is sensitive. Theories of binaural unmasking have suggested that detection of signals in noise is mediated by detection of differences in interaural correlation. If these theories are correct, improvements in the intelligibility of speech in favorable binaural conditions is most likely mediated by spectro-temporal variations in interaural correlation of the stimulus which mirror the spectro-temporal amplitude modulations of the speech. However, binaural sluggishness should limit the temporal resolution of the representation of speech recovered by this means. The present study tested this prediction in two ways. First, listeners' masked discrimination thresholds for ascending vs descending pure-tone arpeggios were measured as a function of rate of frequency change in the NoSo and NoSpi binaural configurations. Three-tone arpeggios were presented repeatedly and continuously for 1.6 s, masked by a 1.6-s burst of noise. In a two-interval task, listeners determined the interval in which the arpeggios were ascending. The results showed a binaural advantage of 12-14 dB for NoSpi at 3.3 arpeggios per s (arp/s), which reduced to 3-5 dB at 10.4 arp/s. This outcome confirmed that the discrimination of spectro-temporal patterns in noise is susceptible to the effects of binaural sluggishness. Second, listeners' masked speech-reception thresholds were measured in speech-shaped noise using speech which was 1, 1.5, and 2 times the original articulation rate. The articulation rate was increased using a phase-vocoder technique which increased all the modulation frequencies in the speech without altering its pitch. Speech-reception thresholds were, on average, 5.2 dB lower for the NoSpi than for the NoSo configuration, at the original articulation rate. This binaural masking release was reduced to 2.8 dB when the articulation rate was doubled, but the most notable effect was a 6-8 dB increase in thresholds with articulation rate for both configurations. These results suggest that higher modulation frequencies in masked signals cannot be temporally resolved by the binaural system, but that the useful modulation frequencies in speech are sufficiently low (<5 Hz) that they are invulnerable to the effects of binaural sluggishness, even at elevated articulation rates.

Application of localization models to noise suppression in hearing aids

One major problem of hearing-impaired people is the cocktail-party effect, i.e., the reduced ability to understand speech in noisy environments. Normal-hearing listeners use binaural information, i.e., interaural level and phase/time differences, in addition to monaural cues to localize sound sources in the environment. Based on the localization result and on the monaural cues, the listener performs a significant noise reduction by concentrating on one of the sources. Noise reduction algorithms have been proposed that try to imitate the binaural information processing in the hearing system by using two microphones at both ears and a central processor. This setup allows for the use of interaural parameters to estimate the direction of sound sources and to attenuate unwanted sound source directions. However, in noisy environments the binaural parameters become noisy as well, reducing the accuracy of the sound source estimation and therefore reducing the maximum performance of the noise reduction algorithm. In this presentation, a localization algorithm is introduced, which uses a statistical description of the binaural parameters in order to perform a robust estimation of sound source directions in noisy environments. Noise reduction schemes based on this localization algorithm are discussed.

Binaural modulation detection interference

The ability to detect amplitude modulation (AM) of a tonal probe can be disrupted by the presence of modulated masking tones. Two experiments examined whether a disparity in the interaural parameters of the probe and masker can reduce the amount of interference. In the first experiment, the effects of interaural time and intensity differences were studied in separate sets of conditions. With low-frequency carriers, the detection of 10-Hz probe modulation in the presence of 10-Hz masker modulation was not significantly affected by interaural time differences. With higher-frequency carriers, dichotic stimuli were generated through combinations of diotic, dichotic, or monotic probe and masker presentations in which the probe and masker did not share a common interaural intensity difference. In these conditions, the amount of interference was affected by the interaural configuration. However, monotic level differences between the probe and masker may have contributed to the effect of interaural configuration. In the second experiment, the probe and masker were presented through separate speakers in an enclosed listening environment. Spatial separation between the sources for the probe and masker led to a small reduction in the amount of interference. When the masker modulation rate was varied with the probe AM rate fixed at 10 Hz, the extent of tuning in the modulation domain in the sound-field conditions was similar to that obtained with diotic stimulus presentation over headphones.

The effects of randomizing values of interaural disparities on binaural detection and on discrimination of interaural correlation.

The purpose of this study was to assess whether randomizing (roving) baseline interaural parameters of binaural stimuli would adversely affect performance in masking-level differences and in interaural correlation-discrimination paradigms. Listeners' detection thresholds were measured in NoSo and NoS pi configurations for both broadband (100-3000 Hz) and narrow-band (450-550 Hz) maskers centered at 500 Hz. In addition, thresholds of interaural decorrelation (from a reference correlation of 1.0) were measured for 100-Hz-wide bands of noise centered at 500 Hz. Data were obtained under conditions in which either values of ITD or values of IID were roved both within and across trials. Data were also collected in the absence of a rove. The range of the rove was +/- 500 microseconds for ITDs and was +/- 6 dB for IIDs. The duration of the stimuli was varied between 17 and 310 ms. Overall, the results indicate that, independent of duration, roving the interaural cues produced small degradations of performance as compared to data obtained in the absence of a rove. The data are inconsistent with the notion that binaural detection depends upon reliable changes in laterality produced by adding S pi signals to diotic maskers. On the other hand, the data are consistent with modern models of binaural processing.

Localization interference between free-field signals

With headphones, the perceived lateral position of one frequency band (the target) can be affected by the presence of a second band with different interaural parameters (the interferer). This binaural interference persists despite a stimulus onset asynchrony (SOA) from interferer to target of up to 320 ms [L. R. Bernstein and C. Trahiotis, J. Acoust. Soc. Am. 94, 735–742 (1993)]. The current study presented targets and interferers in anechoic space from pairs of speakers chosen from an array of 12. A visual pointer was used to indicate the perceived location of the target. Localization of a sinusoidally amplitude-modulated (SAM) target (carrier = 500 Hz, modulator = 250 Hz) was unaffected by the presence of an SAM interferer (carrier = 4000 Hz). When the target was the higher frequency, however, there were large individual differences, with some listeners showing interference while others did not. Unlike the studies with headphones, when interference was seen a distractor to target SOA of only 40 ms was sufficient to eliminate the interference. [Work supported by a grant from the NIH, NIDCD 00087.]

Effects of interaural configuration of wideband forward and backward fringes on detection of an interaural phase shift of a narrow portion of wideband noise

Threshold interaural phase shifts were measured for a 20-Hz-wide section (CF=500 Hz) of wideband noise. Each interval of a 2-afc task consisted of an observation interval (either 100 or 250 ms), and forward and backward temporal fringes, all having identical amplitude spectra. A spectral band (250, 500, or 750 Hz wide) surrounding the target band was diotic throughout the entire stimulus presentation. Interaural phases of frequencies above and below the diotic band (flanking bands) differed for the temporal fringes and the observation interval. For one set of conditions, the flanking bands were diotic during the temporal fringes and interaurally phase-shifted ±90° during the observation interval. In other conditions, the flanking bands contained the 90° interaural phase shift in the temporal fringes, and were diotic in the observation interval. In general, thresholds were higher for conditions in which the temporal fringes were completely diotic, compared to conditions in which the interaural parameters of the flanking bands initially differed from the diotic band surrounding the target. This difference in thresholds decreased as the width of the diotic band and the duration of the observation interval increased. [Work supported by NIH.]