Processing interaural differences in lateralization and binaural signal detection

Humans perceive their acoustic environment by means of two ears located at opposite positions on the sides of their head. This specific configuration gives the possibility to simultaneously sense the surroundings with two, slightly different, inputs. The differences between the waveforms reaching the two ears are called interaural differences and are measured in the form of interaural level differences (ILD) and interaural time differences (ITD). While it is known from the literature that humans rely on the assessment of these two characteristics of sound waves to perform auditory tasks, knowledge about of the actual processing of these interaural differences in the peripheral and central parts of the human auditory system is still limited. In this thesis, we conducted experiments and used models to study the mechanisms underlying the perception and the processing of interaural time and level differences. In chapter 2, the model proposed by Breebaart et at. (2001a) for predicting binaural signal detection thresholds was used to analyze the internal representations of stimuli carrying ITDs. Several strategies were considered in order to simulate lTD detection conditions. In particular, issues regarding the characteristics and the amount of binaural cells used to perform the simulations have been discussed. In chapter 3, lTD detection thresholds were determined experimentally as a function of the stimulus duration. These thresholds were then compared with theoretical predictions that were derived under the assumption that the cue used by listeners was optimally integrated over time. The results showed that the temporal integration of lTD is optimal for stimulus durations up to about 20 ms. For longer stimuli, temporal integration was less than theoretically predicted. It has been suggested in the literature that an important mechanism involved in the processing of interaural differences is the emphasis of information near the stimulus onsets. Our results showed that such a mechanism does not play a major role in the temporal integration of ITDs for the considered stimulus durations (10-400 ms). We were also able to predict lTD detection thresholds as a function of the stimulus duration with a slightly modified version of the binaural model proposed by Breebaart et at. (2001a). The perception and the processing of ITDs was further investigated in chapter 4 in an experimental configuration where the part of the stimuli that carried the lTD (probe) was preceded, followed or surrounded by stimulus parts (fringes) that had an lTD of 0 ms. Such experimental conditions had been previously investigated for stimuli shorter than 100 ms. The results of these studies showed that a forward (preceding) fringe induces a larger decrease in performance than a backward (following) fringe. These findings had been accounted for by models relying on an emphasis of the information near the stimulus onsets and a filtering of the stimuli with a temporal window that diluted the lTD within the diotic fringes. We conducted several experiments with similar conditions with the exception that both fringes and probes had much longer durations. Our results showed that, while a forward fringe had a minor effect, a backward fringe impaired greatly the listener’s performance. Our results, which could not be predicted by any of the models proposed for lTD detection, suggested that for stimulus durations larger than about 100 ms, an eventual onset emphasis does not play a dominant role. From our present understanding, it seems doubtful that these experimental results can be understood solely by bottom-up processes. The processing and the influence of ILDs on binaural detection conditions was investigated in a third study described in chapter 5. We conducted experiments where listeners had to detect sinusoids with an interaural phase difference of ~r presented within a diotic noise masker. Thresholds were measured as a function of an overall ILD and for varying masker bandwidths. These results were compared with binaural detection thresholds measured with a reduction of the masker correlation. A partial phenomenological agreement was found between the two data sets. The binaural model proposed by Breebaart et at. (2001a) could account for the results obtained for both types of experimental conditions for subcritical bandwidths. Analyses of internal representations brought insights regarding the partial agreement of the two data sets. It was found that an ILD induces an interaural decorrelation solely in the form of instantaneous interaural level differences in the envelope. In contrast, a direct decorrelation of the masker leads to instantaneous interaural differences in both the level and the phase of the internal representations. Our findings on the perception and the processing of interaural differences can be summarized in a few points. Interaural time differences are not necessarily optimally processed if presented in isolation. This certainly seems to be the case for stimuli exceeding a certain duration (20-50 ms). When presented together with other, non-informative signal parts, interaural time differences are processed differently whether the stimuli are longer or shorter than about 100 ms. While a mechanism like onset emphasis has been reported in the literature to be relevant for short duration stimuli (<10 ms), it is unclear what the dominant aspects of the processing of the lTD are for longer stimuli. The data from this thesis suggest a possible influence of more central processes. Our results support the view that an interaural level difference in a binaural stimulus leads to a specific decorrelation in the internal representation. It is, however, still unclear, how much this decorrelation is caused by the initial overshoot and how much by the late, more stationary, parts of a stimulus.