Spectral Weighting Underlies Perceived Sound Elevation

Age and Auditory Spatial Perception in Humans: Review of Behavioral Findings and Suggestions for Future Research

It has been well documented, and fairly well known, that concomitant with an increase in chronological age is a corresponding increase in sensory impairment. As most people realize, our hearing suffers as we get older; hence, the increased need for hearing aids. The first portion of the present paper is how the change in age apparently affects auditory judgments of sound source position. A summary of the literature evaluating the changes in the perception of sound source location and the perception of sound source motion as a function of chronological age is presented. The review is limited to empirical studies with behavioral findings involving humans. It is the view of the author that we have an immensely limited understanding of how chronological age affects perception of space when based on sound. In the latter part of the paper, discussion is given to how auditory spatial perception is traditionally conducted in the laboratory. Theoretically, beneficial reasons exist for conducting research in the manner it has been. Nonetheless, from an ecological perspective, the vast majority of previous research can be considered unnatural and greatly lacking in ecological validity. Suggestions for an alternative and more ecologically valid approach to the investigation of auditory spatial perception are proposed. It is believed an ecological approach to auditory spatial perception will enhance our understanding of the extent to which individuals perceive sound source location and how those perceptual judgments change with an increase in chronological age.

Spatial Audio Scene Characterization (SASC): Automatic Localization of Front-, Back-, Up-, and Down-Positioned Music Ensembles in Binaural Recordings

The automatic localization of audio sources distributed symmetrically with respect to coronal or transverse planes using binaural signals still poses a challenging task, due to the front–back and up–down confusion effects. This paper demonstrates that the convolutional neural network (CNN) can be used to automatically localize music ensembles panned to the front, back, up, or down positions. The network was developed using the repository of the binaural excerpts obtained by the convolution of multi-track music recordings with the selected sets of head-related transfer functions (HRTFs). They were generated in such a way that a music ensemble (of circular shape in terms of its boundaries) was positioned in one of the following four locations with respect to the listener: front, back, up, and down. According to the obtained results, CNN identified the location of the ensembles with the average accuracy levels of 90.7% and 71.4% when tested under the HRTF-dependent and HRTF-independent conditions, respectively. For HRTF-dependent tests, the accuracy decreased monotonically with the increase in the ensemble size. A modified image occlusion sensitivity technique revealed selected frequency bands as being particularly important in terms of the localization process. These frequency bands are largely in accordance with the psychoacoustical literature.

Adaptive Response Behavior in the Pursuit of Unpredictably Moving Sounds

Although moving sound-sources abound in natural auditory scenes, it is not clear how the human brain processes auditory motion. Previous studies have indicated that, although ocular localization responses to stationary sounds are quite accurate, ocular smooth pursuit of moving sounds is very poor. We here demonstrate that human subjects faithfully track a sound’s unpredictable movements in the horizontal plane with smooth-pursuit responses of the head. Our analysis revealed that the stimulus–response relation was well described by an under-damped passive, second-order low-pass filter in series with an idiosyncratic, fixed, pure delay. The model contained only two free parameters: the system’s damping coefficient, and its central (resonance) frequency. We found that the latter remained constant at ∼0.6 Hz throughout the experiment for all subjects. Interestingly, the damping coefficient systematically increased with trial number, suggesting the presence of an adaptive mechanism in the auditory pursuit system (APS). This mechanism functions even for unpredictable sound-motion trajectories endowed with fixed, but covert, frequency characteristics in open-loop tracking conditions. We conjecture that the APS optimizes a trade-off between response speed and effort. Taken together, our data support the existence of a pursuit system for auditory head-tracking, which would suggest the presence of a neural representation of a spatial auditory fovea (AF).

A Comparison of Human against Machine-Classification of Spatial Audio Scenes in Binaural Recordings of Music

The purpose of this paper is to compare the performance of human listeners against the selected machine learning algorithms in the task of the classification of spatial audio scenes in binaural recordings of music under practical conditions. The three scenes were subject to classification: (1) music ensemble (a group of musical sources) located in the front, (2) music ensemble located at the back, and (3) music ensemble distributed around a listener. In the listening test, undertaken remotely over the Internet, human listeners reached the classification accuracy of 42.5%. For the listeners who passed the post-screening test, the accuracy was greater, approaching 60%. The above classification task was also undertaken automatically using four machine learning algorithms: convolutional neural network, support vector machines, extreme gradient boosting framework, and logistic regression. The machine learning algorithms substantially outperformed human listeners, with the classification accuracy reaching 84%, when tested under the binaural-room-impulse-response (BRIR) matched conditions. However, when the algorithms were tested under the BRIR mismatched scenario, the accuracy obtained by the algorithms was comparable to that exhibited by the listeners who passed the post-screening test, implying that the machine learning algorithms capability to perform in unknown electro-acoustic conditions needs to be further improved.

Differential Adaptation in Azimuth and Elevation to Acute Monaural Spatial Hearing after Training with Visual Feedback

Sound localization in the horizontal plane (azimuth) relies mainly on binaural difference cues in sound level and arrival time. Blocking one ear will perturb these cues, and may strongly affect azimuth performance of the listener. However, single-sided deaf listeners, as well as acutely single-sided plugged normal-hearing subjects, often use a combination of (ambiguous) monaural head-shadow cues, impoverished binaural level-difference cues, and (veridical, but limited) pinna- and head-related spectral cues to estimate source azimuth. To what extent listeners can adjust the relative contributions of these different cues is unknown, as the mechanisms underlying adaptive processes to acute monauralization are still unclear. By providing visual feedback during a brief training session with a high-pass (HP) filtered sound at a fixed sound level, we investigated the ability of listeners to adapt to their erroneous sound-localization percepts. We show that acutely plugged listeners rapidly adjusted the relative contributions of perceived sound level, and the spectral and distorted binaural cues, to improve their localization performance in azimuth also for different sound levels and locations than those experienced during training. Interestingly, our results also show that this acute cue-reweighting led to poorer localization performance in elevation, which was in line with the acoustic–spatial information provided during training. We conclude that the human auditory system rapidly readjusts the weighting of all relevant localization cues, to adequately respond to the demands of the current acoustic environment, even if the adjustments may hamper veridical localization performance in the real world.

Automatic Spatial Audio Scene Classification in Binaural Recordings of Music

The aim of the study was to develop a method for automatic classification of the three spatial audio scenes, differing in horizontal distribution of foreground and background audio content around a listener in binaurally rendered recordings of music. For the purpose of the study, audio recordings were synthesized using thirteen sets of binaural-room-impulse-responses (BRIRs), representing room acoustics of both semi-anechoic and reverberant venues. Head movements were not considered in the study. The proposed method was assumption-free with regards to the number and characteristics of the audio sources. A least absolute shrinkage and selection operator was employed as a classifier. According to the results, it is possible to automatically identify the spatial scenes using a combination of binaural and spectro-temporal features. The method exhibits a satisfactory classification accuracy when it is trained and then tested on different stimuli but synthesized using the same BRIRs (accuracy ranging from 74% to 98%), even in highly reverberant conditions. However, the generalizability of the method needs to be further improved. This study demonstrates that in addition to the binaural cues, the Mel-frequency cepstral coefficients constitute an important carrier of spatial information, imperative for the classification of spatial audio scenes.

Learning to localise weakly-informative sound spectra with and without feedback

How the human auditory system learns to map complex pinna-induced spectral-shape cues onto veridical estimates of sound-source elevation in the median plane is still unclear. Earlier studies demonstrated considerable sound-localisation plasticity after applying pinna moulds, and to altered vision. Several factors may contribute to auditory spatial learning, like visual or motor feedback, or updated priors. We here induced perceptual learning for sounds with degraded spectral content, having weak, but consistent, elevation-dependent cues, as demonstrated by low-gain stimulus-response relations. During training, we provided visual feedback for only six targets in the midsagittal plane, to which listeners gradually improved their response accuracy. Interestingly, listeners' performance also improved without visual feedback, albeit less strongly. Post-training results showed generalised improved response behaviour, also to non-trained locations and acoustic spectra, presented throughout the two-dimensional frontal hemifield. We argue that the auditory system learns to reweigh contributions from low-informative spectral bands to update its prior elevation estimates, and explain our results with a neuro-computational model.