No description available.

While humans can separate a sound of interest amidst a cacophony of contending sounds in an echoic environment, machine-based methods lag behind in solving this task. This thesis thus aims at improving performance of audio separation algorithms when they are informed i.e. have access to source location information. These locations are assumed to be known a priori in this work, for example by video processing. Initially, a multi-microphone array based method combined with binary time-frequency masking is proposed. A robust least squares frequency invariant data independent beamformer designed with the location information is utilized to estimate the sources. To further enhance the estimated sources, binary time-frequency masking based post-processing is used but cepstral domain smoothing is required to mitigate musical noise. To tackle the under-determined case and further improve separation performance at higher reverberation times, a two-microphone based method which is inspired by human auditory processing and generates soft time-frequency masks is described. In this approach interaural level difference, interaural phase difference and mixing vectors are probabilistically modeled in the time-frequency domain and the model parameters are learned through the expectation-maximization (EM) algorithm. A direction vector is estimated for each source, using the location information, which is used as the mean parameter of the mixing vector model. Soft time-frequency masks are used to reconstruct the sources. A spatial covariance model is then integrated into the probabilistic model framework that encodes the spatial characteristics of the enclosure and further improves the separation performance in challenging scenarios i.e. when sources are in close proximity and when the level of reverberation is high. Finally, new dereverberation based pre-processing is proposed based on the cascade of three dereverberation stages where each enhances the twomicrophone reverberant mixture. The dereverberation stages are based on amplitude spectral subtraction, where the late reverberation is estimated and suppressed. The combination of such dereverberation based pre-processing and use of soft mask separation yields the best separation performance. All methods are evaluated with real and synthetic mixtures formed for example from speech signals from the TIMIT database and measured room impulse responses.

This paper explores how vision affects spatial audio perception in virtual reality. We created four virtual environments with different reverb and room sizes, and recorded binaural clicks in each one. We conducted two experiments: one where participants judged the audio-visual match, and another where they pointed to the click direction. We found that vision influences spatial audio perception and that congruent audio-visual cues improve accuracy. We suggest some implications for virtual reality design and evaluation.

Spatial hearing is crucial in real life but deteriorates in participants with severe sensorineural hearing loss or single-sided deafness. This ability can potentially be improved with a unilateral cochlear implant (CI). The present study investigated measures of sound localization in participants with single-sided deafness provided with a CI. Sound localization was measured separately at eight loudspeaker positions (4°, 30°, 60°, and 90°) on the CI side and on the normal-hearing side. Low- and high-frequency noise bursts were used in the tests to investigate possible differences in the processing of interaural time and level differences. Data were compared to normal-hearing adults aged between 20 and 83. In addition, the benefit of the CI in speech understanding in noise was compared to the localization ability. Fifteen out of 18 participants were able to localize signals on the CI side and on the normal-hearing side, although performance was highly variable across participants. Three participants always pointed to the normal-hearing side, irrespective of the location of the signal. The comparison with control data showed that participants had particular difficulties localizing sounds at frontal locations and on the CI side. In contrast to most previous results, participants were able to localize low-frequency signals, although they localized high-frequency signals more accurately. Speech understanding in noise was better with the CI compared to testing without CI, but only at a position where the CI also improved sound localization. Our data suggest that a CI can, to a large extent, restore localization in participants with single-sided deafness. Difficulties may remain at frontal locations and on the CI side. However, speech understanding in noise improves when wearing the CI. The treatment with a CI in these participants might provide real-world benefits, such as improved orientation in traffic and speech understanding in difficult listening situations.

People are usually directed towards each other in conversations, to make it easier to hear what is being said. Algorithms for voice and speech recognition works in a similar way, regarding the microphone direction towards the sound source. In this thesis in electronics has therefore a servo control with binaural sound localization been implemented on a microcontroller connected to two microphones. When people perceive sound, the brain can estimate the sound source direction by comparing the time taken by the sound reaching one ear to the other [1]. The difference in time is called the interaural time difference, and can be calculated using various techniques. An exploratory comparison between the techniques cross-correlation and cross-spectrum analysis was carried out before implementation. Advantages and disadvantages of each technique were evaluated at the same time. The result is a functioning servo control, that uses a cross correlation algorithm to calculate the interaural time difference, and controls a servo motor towards the sound source with a P-regulated error reduction method. The project was implemented on the microcontroller ATmega328P from Atmel without using floating point calculations. The thesis was carried out on behalf of the company Jetspark Robotics.

The interaural level difference (ILD) is a sound localization cue first computed in the lateral superior olive (LSO) by comparing the loudness of sounds between the two ears. In the auditory cortex, one class of neurons is excited by contralateral but not ipsilateral monaural sounds. These "EO" neurons prefer ILDs where contralateral sounds are louder than ipsilateral sounds. Another class, the "PB" neurons, are unresponsive to monaural sounds but respond predominantly to binaural ILDs, when both ears receive simultaneous sounds of roughly equal loudness (0 ILD). Behavioral studies show that ILD sensitivity is invariant to increasing sound levels. However, in the LSO, ILD response functions shift towards the excitatory ear as sound level increases, indicating level-dependence. Thus, changes in firing rate can indicate either a change in sound location or sound level, or both. This suggests a transformation in level-sensitivity between the LSO and the perception of sound sources, yet the location of this transformation remains unknown. I performed recordings in the auditory cortex of the rat to test whether neurons were invariant to overall sound level. I found that with increasing sound levels, ILD responses were level-dependent, suggesting that level invariance of ILD sensitivity is not present in the rat auditory cortex. In general, neurons follow one of two processing strategies. The tuning of cortical cells typically follows the "inheritance strategy", such that the spiking output of the cell matches that of the excitatory synaptic input. However, cortical tuning can be modified by inhibition in the "local processing strategy". In this case, neurons are prevented from spiking at non-preferred stimuli by inhibition that overwhelms excitation. The tuning strategy of cortical neurons to ILD remains unknown. I performed whole-cell recordings in the anesthetized rat and compared the spiking output with synaptic inputs to ILDs within the same neurons. I found that the PB neurons showed evidence of the local processing strategy, which is a novel role for cortical inhibition, whereas the EO neurons utilized the inheritance strategy. This result suggests that an auditory cortical circuit computes sensitivity for midline ILDs. This dissertation includes previously published/unpublished co-authored material.

Purpose: To establish normative data for children and to characterize developmental differences in performance with the free recall version of the Randomized Dichotic Digits Test. Research Design: Group comparison of behavioral data derived from administration of the Randomized Dichotic Digits Test. Study Sample: Children from 10 to 18 years of age (167) and young adults from 19 to 28 years of age (50). Results: Performance improved with age across all types of digit pairs, especially in the left ear, leading to smaller interaural asymmetries among older participants. A left-ear advantage was produced by 39 subjects (18%), only two of whom were left-handed. Normative data are reported for right and left ear scores and for interaural asymmetry (percent correct difference between the two ears) under one-, two-, and three-pair conditions of the test and for interaural asymmetry across the entire test. A unilateral deficit was identified in children (15.5%) and young adults (12%) for the left ear and in children (11.3%) and young adults (6%) for the right ear. A bilateral deficit was also identified in children (6.5%) and young adults (6%). Conclusions: This test may be useful as part of the clinical battery for identifying binaural integration weaknesses and referring individuals for auditory rehabilitation for interaural asymmetry (ARIA).

Many neurons in the auditory midbrain are tuned to binaural cues. Two prominent binaural cues are the interaural intensity difference (IID) and the interaural time difference (ITD). The ITD cue can further be classified as either an ongoing ITD, which compares the phase difference in the waveform of low frequency stimuli present at either ear, or an onset ITD, which compares the onset time of arrival of two stimuli at either ear. Little research has been done on the sensitivity of single neurons to onset ITDs in the auditory system, particularly in bats. The current study examines the response properties of neurons in the inferior colliculus (IC) of the big brown bat, Eptesicus fuscus, to onset ITDs in response to high frequency pure tones. Measures of neurons’ dynamic response—the segment of the ITD function exhibiting the highest rate of change in activity—revealed an average change of 36% of its maximum response within the estimated behaviorally relevant range of ITDs. Time-intensity trading describes the ability of the brain to compensate the binaural time cue (ITD) cue for the binaural intensity cue (IID) and can be measured as the horizontal shift of an ITD function at various IIDs. Across all IC neurons, an average time-intensity trading ratio of 30 μs/dB was calculated to measure the sensitivity of IC neurons’ ITD response to changing IIDs. Minimum and maximum ITD responses were found to be clustered within a narrow range of ITDs. The average peak ITD response occurred at 268 μs and is consistent with findings in other mammals. All results in ITD tuning, time-intensity trading, and response maximum were invariant to stimulus frequency, confirming that IC neurons responded to onset ITDs and not ongoing ITDs. These results suggest the potential for high frequency onset cues to assist in the azimuthal localization of sound in echolocating bats. / Thesis / Master of Science (MSc)

Mémoire numérisé par la Division de la gestion de documents et des archives de l'Université de Montréal.

Virtual three-dimensional (3D) auditory displays utilize signal-processing techniques to alter sounds presented through headphones so that they seem to originate from specific spatial locations around the listener. In some circumstances bone-conduction headsets (bonephones) can provide an alternative sound presentation mechanism. However, existing 3D audio rendering algorithms need to be adjusted to use bonephones rather than headphones. This study provided anchor points for a function of shift values that could be used to adapt virtual 3D auditory displays for use with bonephones. The shift values were established by having participants adjust phase and amplitude of two waves in order to cancel out the signal and thus produce silence. These adjustments occurred in a listening environment consisting of air-conducted and bone-conducted tones, as well as air- conducted masking. Performance in the calibration condition suggested that participants understood the task, and could do this task with reasonable accuracy. In the bone-to-air listening conditions, the data produced a clear set of anchor points for an amplitude shift function. The data did not reveal, however, anchor points for a phase shift function the data for phase were highly variable and inconsistent. Application of shifts, as well as future research to establish full functions and better understand phase are discussed, in addition to validation and follow-up studies.