Models Of Binaural Hearing Research Articles

The fMRI Data of Thompson et al. (2006) Do Not Constrain How the Human Midbrain Represents Interaural Time Delay.

This commentary provides an alternate interpretation of the fMRI data that were presented in a communication to the journal Nature Neuroscience (Thompson et al., Nat. Neurosci. 9: 1096-1098, 2006 ). The authors argued that their observations demonstrated that traditional models of binaural hearing which incorporate "internal delays," such as the coincidence-counting mechanism proposed by Jeffress and quantified by Colburn, are invalid, and that a new model for human interaural time delay processing must be developed. We argue that the fMRI data presented do not strongly favor either the refutation or the retention of the traditional models, although they may be useful in constraining the physiological sites of various processing stages. The conclusions of Thompson et al. are based on the locations of maximal activity in the midbrain in response to selected binaural signals. These locations are inconsistent with well-known perceptual attributes of the stimuli under consideration, as is noted by the authors, which suggests that further processing is involved in forming the percept of subjective lateral position.

Acoustic and perceptual effects of magnifying interaural difference cues in a simulated “binaural” hearing aid

Objective: To investigate the influence of an algorithm designed to enhance or magnify interaural difference cues on speech signals in noisy, spatially complex conditions using both technical and perceptual measurements. To also investigate the combination of interaural magnification (IM), monaural microphone directionality (DIR), and binaural coherence-based noise reduction (BC). Design: Speech-in-noise stimuli were generated using virtual acoustics. A computational model of binaural hearing was used to analyse the spatial effects of IM. Predicted speech quality changes and signal-to-noise-ratio (SNR) improvements were also considered. Additionally, a listening test was carried out to assess speech intelligibility and quality. Study sample: Listeners aged 65–79 years with and without sensorineural hearing loss (N = 10 each). Results: IM increased the horizontal separation of concurrent directional sound sources without introducing any major artefacts. In situations with diffuse noise, however, the interaural difference cues were distorted. Preprocessing the binaural input signals with DIR reduced distortion. IM influenced neither speech intelligibility nor speech quality. Conclusions: The IM algorithm tested here failed to improve speech perception in noise, probably because of the dispersion and inconsistent magnification of interaural difference cues in complex environments.

Applications of models of binaural hearing

Models of binaural hearing have been proposed for more than 60 decades, but it is only recently that they are ready for technological application. Recognizing this situation, 15 laboratories in Europe and in the United States have founded a research group (AABBA) with the aim of setting in place “Aural Assessment by Means of Binaural Algorithm.” Now, after its first 4-years term of activity, the group has published their recent results in a chapter book (Springer)—including a software package for building customized binaural models. In this talk, the group’s most relevant results will be described. The results concern the following areas of model application, among others: binaural analysis of aural scenes, binaural de-reverberation, binaural quality assessment of audio channels, loudspeakers and performance spaces, binaural perceptual coding, binaural processing in hearing aids and cochlea implants, binaural systems in robots, binaural/tactile human–machine interfaces, speech-intelligibility prediction in rooms, and/or multi-speaker scenarios. Current members of AABBA are research labs at Universities/Techn. Universities in Helsinki, Boston, Cardiff, Oldenburg, Lyon, Troy, NY (Rensselaer), Bochum, Berlin, Copenhagen, Dresden, Eindhoven, Munich, Paris (Pierre et Marie Curie), Toulouse, and at the Austrian Academy’s Acoustics Research Inst. in Vienna.

Using low-frequency threshold interaural time differences to test models of binaural hearing

All models for detecting an interaural time difference (ITD) begin with model cross-correlator cells. Models differ in their inputs (excitatory/inhibitory) and in the distribution of cells with respect to interaural delay and frequency. For instance, the Jeffress model postulates a broad distribution on delay. Also, alternative binaural displays are based on different moments of the cross-correlation. For instance, the centroid is based on the first moment. The different models predict different frequency dependences of the threshold ITD in the limit of low frequency. Limiting behavior was computed for the various models using different assumptions about the frequency dependence of the synchrony of inputs to cross-correlator cells and about the sharpness of the rate-ITD function from the cells. The results were compared with the measured low-frequency functional behavior of ITD thresholds for four human listeners, which ranged from ƒ-0.8 to ƒ-1.4. These measured exponents disagree with predictions from some combinations of models. In particular, the popular centroid display within the Jeffress model tends to lead to slopes that are steeper than observed experimentally. [Work supported by the National Natural Science Foundation of China grant 61175043 and the Air Force Office of Scientific Research grant 11NL002.]

Stochastic interpolation model of the medial superior olive neural circuit

This article presents a stochastic model of binaural hearing in the medial superior olive (MSO) circuit. This model is a variant of the slope encoding models. First, a general framework is developed describing the elementary neural operations realized on spike trains in individual parts of the circuit and how the neurons converging onto the MSO are connected. Random delay, coincidence detection of spikes, divergence and convergence of spike trains are operations implemented by the following modules: spike generator, jitter generator, and coincidence detector. Subsequent processing of spike trains computes the sound azimuth in the circuit. The circuit parameters that influence efficiency of slope encoding are studied. In order to measure the overall circuit performance the concept of an ideal observer is used instead of a detailed model of higher relays in the auditory pathway. This makes it possible to bridge the gap between psychophysical observations in humans and recordings taken of small rodents. Most of the results are obtained through numerical simulations of the model.

Lateralization of Huggins pitch

The Huggins pitch is a sensation of pitch generated from a broadband noise having a narrowband boundary region where the interaural phase difference varies rapidly as a function of frequency. Models of binaural hearing predict that the pitch image should be well lateralized. A direct psychophysical experimental method was used to estimate the lateral positions of Huggins pitch images with two different forms of phase boundaries, linear phase and stepped phase. A third experiment measured the lateral positions of sine tones with controlled interaural phase differences. The results showed that the lateralization of Huggins pitch stimuli was similar to that of the corresponding sine tones and that the lateralizations of the two forms of Huggins pitch phase boundaries were even more similar to one another. Both Huggins pitches and sine tones revealed strong laterality compression (exponent approximately 0.5). Ambiguous stimuli, with an interaural phase difference of 180 degrees , were consistently lateralized on one side or the other according to individual asymmetries-an effect called "earedness." An appendix to this article develops a new first-order lateralization model, the salient phase density model, which combines attributes of previous models of dichotic pitch lateralization.

Computational Auditory Scene Analysis: Principles, Algorithms and Applications

First Page

Binaural Hearing and Systems for Sound Reproduction

Recent developments in models of binaural hearing can be usefully adapted and extedned to provide design tools for engineers engaged in the design of systems for sound reproduction. The particular focus of the work described is upon the development of models that give good statistical predictions of human sound localisation, based upon knowledge of the fluctuating acoustic pressures at the ears. Such models can be applied successfully to the prediction of stereophonic image localisation and reveal a number of important features of localisation relevant to audio system design. Developments will also be described in loudspeaker based systems for binaural reproduction that are finding their way into practical use. Binaural hearing models can be used to provide a preliminary evaluation of the performance of alternative designs. Finally, a brief review will be presented of multi‐channel loudspeaker‐based systems aimed at “full field” sound reproduction. Again, models of localisation provide some useful guidance for the designers of such systems.

How does the mammalian brain code for binaural information?

We will summarize new physiological results which contradict established models of binaural hearing, namely (1) the A-limited range of encoded interaural time difference, ITD, (2) a correlation between ipsilateral inhibition and best ITD, and (3) a frequency dependent non-linear mapping of ITD which reflects distribution of naturally occurring ITDs. Furthermore, we will outline a new concept for binaural models which incorporates these new results, and will propose a new illustrative format in which to present the triplet of binaural parameters: interaural phase difference, interaural intensity difference, and interaural coherence (redefined).

Virtual auditory localization in noise

Interfering sounds have different interaural cross correlations (IACCs) depending on the location and environment in which they are produced. IACCs, in addition to signal bandwidth and signal-to-noise ratio, influence localization acuity. Most localization experiments use stimuli presented in the free-field without maskers or with one directional masker [Good and Gilkey, J. Acoust. Soc. Am. 99, 1108–1117 (1996)]. Hirsh [J. Acoust. Soc. Am. 22, 196–200 (1950)] and Mershon and Lin [Ergonomics 30, 1161–1173 (1987)] have measured auditory localization in reverberant environments. However, in these studies the signals and masker(s) were presented in the reverberant environment. No studies have addressed the effects of diffuse noise on auditory localization over headphones. The current study examines the effect of band passing a directional noise signal presented over headphones masked by noises of different interaural cross correlations. The signal-to-noise ratio is varied from +10 to −10 dB, and the signal is low-pass filtered at 1.6, 4, 8, and 15 kHz. Findings from this study have implications for models of binaural hearing and the design of directional audio displays.

Phoneme recognition with a model of binaural hearing

This work presents a method for phoneme recognition based on a model of the human binaural auditory system. Performance of the binaural system is compared to that of an equivalent monaural system across a range of signal-to-noise ratios (SNRs) and simulated azimuth separations of speech and noise sources. Comparison of binaural and monaural phoneme recognition accuracy for both speech and noise at 0/spl deg/ suggests an effective increase in SNR of 5 dB for the binaural system. Experiments are also conducted to investigate binaural system performance for speech and noise at differing azimuth angles. Two binaural speech and noise conditions are considered: i) speech from 0/spl deg/ plus noise from 0/spl deg/; and ii) speech from 0/spl deg/ plus noise from 90/spl deg/. A comparison of relative performance differences for each binaural system with the monaural system shows the greatest increase in phoneme recognition accuracy when the speech and noise are simulated using different azimuth angles (0/spl deg/ speech plus noise from 90/spl deg/).

Lateralization of comodulated bands of noise

Lateralization thresholds were measured in a 2IFC task for two conditions. The 400-ms signal consisted of two high-frequency narrow bands of noise with common interaural delays (cf.=2.55 and 3.35 kHz; bandwidths=50,100,200,400 Hz). In the first condition, the temporal envelopes of the two bands were the same (comodulated) and, in the second condition, they were independent. Results of this experiment showed improvements in lateralization when the bands were comodulated compared to independently modulated bands. In a second experiment, the two noise bands were replaced by two sinusoidally amplitude modulated (SAM) tones (cf.=2.55 and 4 kHz; modulation rate=100 Hz) and lateralization thresholds were measured as a function of the relative phase of modulation between the two SAMs. Thresholds were lowest when the envelope of the two SAMs were in phase and monotonically increased as the envelopes were increasingly shifted out of phase up to 180°. In a third experiment, binaural interference was examined. Results are discussed in terms of cross-correlation models of binaural hearing. [Work supported by NIH and AFOSR.]

Speech enhancement based on physiological and psychoacoustical models of modulation perception and binaural interaction.

A novel approach for analyzing and filtering speech is described and evaluated which utilizes the "modulation spectrogram," i.e., the two-dimensional representation of modulation frequencies versus center frequency as a function of time. This approach is based on physiological findings of a tonotopical organization of modulation frequencies perpendicular to carrier frequencies as well as psychoacoustical findings of "modulation tuning curves." In addition, an interaction is assumed between the representation of modulation frequencies and the representation of auditory space as described by physiological and psychological models of binaural hearing. A noise-reduction algorithm based on this approach was implemented and tested which enhances or suppresses each combination of modulation frequency and center frequency according to its phase and intensity relation between the two input signals (i.e., both stereo channels of a dummy-head recording). When tested in several situations with interfering speakers and background noise both in anechoic and reverberant environment, the algorithm provided a small but a very robust increase in speech intelligibility which corresponds to approximately 2 dB in signal-to-noise ratio. Possible applications of this algorithm are noise reduction in adverse acoustical situations, digital hearing aids, processing schemes and preprocessing for speech recognition.

Analog VLSI model of binaural hearing

The stereausis model of biological auditory processing is proposed as a representation that encodes both binaural and spectral information in a unified framework. A working analog VLSI chip that implements this model of early auditory processing in the brain is described. The chip is a 100000-transistor integrated circuit that computes the stereausis representation in real time, using continuous-time analog processing. The chip receives two audio inputs, representing sound entering the two ears, computes the stereausis representation, and generates output signals that can directly drive a color CRT display. Outputs from the chips for a variety of artificial and speech stimuli are shown.

Lateralization of bands of noise: Effects of bandwidth and interaural differences of intensity, time, and/or phase

Listeners indicated the intracranial position of narrow-band noises that varied in terms of interaural differences of intensity (IDI), time (IDT), and/or phase (IDP). Extents of laterality produced by IDIs depended greatly on particular combinations of bandwidth and the two other interaural cues. This dependence is inconsistent with the notions that IDIs serve as “weights” (scalars) or act via a (peripheral) time/intensity trade. For example, with narrow-band stimuli, changing IDI by only 3 dB could result in very large changes of laterality (including cases where the image moved from near one ear to near the other). The results will be discussed in terms of current, cross-correlation-based, models of binaural hearing that incorporate “pattern analysis” schemes. [Research supported by NIH and AFOSR.]

Effects of masker waveform and signal-to-masker phase relation on diotic and dichotic masking by reproducible noise.

The proportions of hits and false alarms were estimated for the detection of a 500-Hz sinusoidal signal in each of 25, reproducible samples of wideband, white, Gaussian noise. The effects of signal phase were investigated under diotic (MoSo) and dichotic (MoS pi) conditions and compared to the predictions of two major models of binaural hearing. Averaging the data over samples obscured important across-sample and across-subject differences in performance. The proportions of hits and false alarms for individual noise samples presented under the MoSo condition were highly correlated with those for the same noise samples under the dichotic MoS pi condition, suggesting that the cues determining performance under these conditions are related. Signal-to-masker phase had a large effect on the proportion of hits under the MoSo condition, but only a small effect under the MoS pi condition. The Vector model predicts a large effect of signal phase under the MoS pi condition, and is, therefore, imcompatible with this aspect of the data. The expected value of the decision variable of the EC model is independent of signal phase. However, when the variance of the decision variable is also considered, the EC model does predict changes in the proportion of hits with the phase of the signal, comparable to those observed here. Further, it was shown that, if minor changes in the form of the EC model's decision variable or in the distribution of the internal noise parameters are assumed, the expected value of the decision variable also changes with the phase of the signal.

Theory of binaural interaction based on auditory-nerve data. I. General strategy and preliminary results on interaural discrimination.

We discuss initial research on a model of binaural hearing in which the peripheral transduction from acoustical waveforms to firing patterns on the auditory nerves is explicitly described. In most of this initial research, attention is focused on interaural time discrimination, and the processing that follows the peripheral transduction (the central processing) is assumed to be ideal. By imposing limitations on the central processing and computing the consequences for performance, we find that performance at least as good as experimentally observed performance can be achieved with central processing that is substantially restricted.

Model of Binaural Hearing Based on Auditory Nerve Patterns

A model for binaural hearing is presented in which the firing patterns of the auditory nerve are regarded as input to a central processor. The model is a continuation of work by W. M. Siebert on monaural phenomena, and the random-process description of the firing patterns is based on data from N. Y. S. Kiang's studies in the cat. Early work centered on analysis by an ideal central processor and indicated that further restrictions on the use of detailed timing information were necessary. Restrictions on the permissible interaural comparisons of event times result in a model for the central processor in the form of a set of statistics that are used optimally. Predictions of this model are compared with data from binaural unmasking and interaural discrimination. A distinctive feature of the present model is the central rôle played by the observed randomness inherent in the firing patterns of auditory nerve fibers. [Work supported primarily by a grant from the National Institutes of Health and in part by the Joint Services Electronics Program.]

Effect of masker spectrum level on masking-level differences at low signal frequencies.

Most, if not all, models of binaural hearing predict a dependence of the magnitude of the masking-level difference (MLD) on signal frequency. Specifically, the size of the MLD is expected to increase as the signal frequency is decreased. The results of many experiments in which signal frequency was varied, however, do not support the above prediction. The magnitude of the MLD appears to decrease at frequencies below 300 cps. The present experiment was designed to re-examine the decrease in the size of the MLD at low frequencies and to test a qualitative explanation of this phenomenon based on the concept of internal noise. To do so, the size of the MLD in the N0Sπ masking condition and the N0S0 masking condition was estimated relative to the NMSM masking condition at signal frequencies of 150 and 300 cps at several masker spectrum levels. The results indicated that the magnitude of the MLD at both frequencies is dependent upon spectrum level of the masker. The results further indicate that the dependence of the size of the MLD on spectrum level varies as a function of the signal frequency. The size of the MLD at 150 cps is much more sensitive to changes in masker spectrum level than is the MLD at 300 cps. Both results agree with predictions based on the internal noise hypothesis.

New Model of Binaural Hearing

Preliminary results are presented on a model of binaural hearing in which the firing patterns on the auditory nerve are regarded as inputs to a central processor. The model is a continuation of work by W. M. Siebert on monaural phenomena, and the random-process description of the firing patterns is based on data from the cat. Initially, it is assumed that the processes for the two ears are statistically independent and the central processor is ideal. Predictions are derived for jnd's in interaural time and amplitude for tone bursts as a function of the amplitudes in the two ears, the interaural time delay, and the frequency and duration of the tone burst. These results define the limitations on performance imposed by the peripheral transformation. A more realistic model is achieved by placing restrictions on the central processor. The theoretical results are compared with available data. [Work supported principally by a grant from the National Institutes of Health and in part by the Joint Services Electronics Program.]